Repository navigation
Unified Inbox
Every conversation with every customer and supplier in one place: email, live chat, contact form enquiries, phone calls, voicemail, text messages and WhatsApp, with your own records sitting beside them. One screen instead of five, and a shared history instead of whatever happens to be on somebody's phone.
It is not a CRM. There are no pipelines, no deals and no lead scoring. People exist here so that two emails, a live chat and a phone call from the same human collapse into one story - and, since the Contacts tab arrived, so that you have somewhere to write down their mobile number and where to post things. That is an address book. It stops there.
You will find it under Inbox in the admin sidebar, as a tab called Unified Inbox. Opening it folds the admin sidebar down to its icon rail, because a mail program wants the width more than it wants a list of links, and leaving the Inbox unfolds it again. If you would rather keep the sidebar out, click the ‹ / › toggle while you are in there and it stays put. Its settings live under Settings → Unified Inbox, split across a row of tabs so that changing a folder name no longer means scrolling past a retention policy.
- Your site needs its email set up already (Settings → Email), because that is what sends your replies. See the configuration reference.
- You need the app password for whichever mailbox you want to read. Most providers call it an "app password" and make you generate one; searching for "[your provider] app password" gets you there. iCloud's is at appleid.apple.com.
- If you already run Reply Catcher, read the section at the bottom of this page first. The two must never be pointed at the same mailbox.
Settings → Unified Inbox opens on a row of tabs, one job each. Whichever you are on is remembered in the address bar, so a refresh or a bookmark comes back to it rather than dropping you at the beginning.
| Tab | What is on it |
|---|---|
| Overview | Whether it is working, how much has been collected, and a short list of anything wanting your attention - with a button beside each one that takes you to where it is fixed. On a site with nothing set up yet it is a three-step start instead. |
| Mail accounts | The mailboxes it reads. Test connection, Check now, and how far the first collection has got. |
| Inboxes | The addresses people write to: where each is collected from, how its replies go out, what they are signed with, and who may read them. |
| Collecting | How far back to go, how long to keep things, attachments, and which way round a conversation reads. |
| Sent replies | Whether to find out what became of a reply after it left. |
| People | Telling colleagues from customers, whether shop customers are added to your contacts (only where the shop is installed), the categories you file contacts under, and the shape of your order and quote numbers. |
| Campaigns | Your business name and address for the unsubscribe footer, how long to leave between writing to the same person twice, and how long a finished campaign's log is kept. |
| Other apps | Telling something else on the internet when the post arrives, or when a colleague asks for somebody. |
Start a form on one tab, wander off to another and come back, and what you had typed is still there. Nothing is saved until you press the Save button on that tab, mind.
This is the one idea worth getting straight, because everything else follows from it.
A mail account is a connection to a real mailbox: a server, a username and an app password. It is the thing that does the collecting.
An inbox is an address people write to: hi@yourcompany.co.uk, accounts@yourcompany.co.uk, marcus@yourcompany.co.uk. It is the thing conversations get filed into, and it carries its own name on replies, its own signature, and its own list of who may read it.
One mail account can feed several inboxes. That is the normal case: a single mailbox collecting several addresses at your domain, with the enquiries and the invoices kept apart from each other once they land. You can also run several mail accounts if you genuinely have several mailboxes, and an inbox can exist with no mail account at all for an address you only ever send from.
Settings → Unified Inbox → Mail accounts → Add a mail account.
- Give it a name you will recognise (What to call it).
- Fill in the Mail server, your Username (usually the full email address) and the app password.
- Save, then press Test connection. It either lists the folders it can see or tells you in plain English what went wrong.
The folders it finds are remembered against the account, so the folder boxes on your addresses become menus rather than something to type from memory. They are only ever refreshed when you ask - by Test connection here, or by Update folders on an address.
It reads the main inbox, your Sent folders, your Archive folders, and anything else you name under Other folders to read.
Reading Sent matters more than it sounds. If you answer a customer from your phone, that reply is in your Sent folder and nowhere else, and without it the conversation here would show your customer talking to themselves.
Junk, Trash and Drafts are never read. Spam would mint a conversation and a person out of every message in it, and a draft is not a message.
Nothing in your mailbox is ever changed by the collecting: nothing is marked read, moved or deleted. The single exception is the Sent-folder copy described below, which you switch on yourself.
The rules above assume the mail account exists to serve the site. Plenty do not. If you have pointed this at the account you already had, where the shop's post is filed into one folder and the main inbox is your own bank, your doctor and your online shopping, then reading the main inbox is the last thing you want. Left alone, all of it ends up in here, where your staff can read it.
Two tick boxes on the mail account deal with that. Both are off to begin with, so an account you set up before they existed carries on exactly as it did.
- Read only the folders named here and on the addresses below. The main inbox, the Archive and the Sent folder stop being read automatically. What gets read is what you have actually named: the folders under Other folders to read, plus the Folder to read on each of the addresses this account collects. Name your Sent folder yourself if you want your phone replies, as described above.
- Ignore mail that is not addressed to one of your addresses. Normally that post is kept out of the way under Not filed, in case somebody writes to an address you have not set up yet. Tick this and it is not kept at all - it is passed over and never stored. Replies to conversations already here still arrive either way, so nobody's thread stops halfway through because their reply happened to be addressed to somebody else.
The pair go together on a personal account: the first stops it reading your private folders, the second catches anything of yours that has been filed into the shop's folder by mistake.
Settings → Unified Inbox → Inboxes → Add an inbox. The form asks for a lot at once, so it is grouped: the address itself, where it is collected from, how its replies go out, its signature, and who may read it.
- Address is what people write to. Post arriving for that address is filed here.
- Kind of inbox is the one answer on this form you cannot take back by reading it, so it is asked early. See "Shared and individual inboxes" just below.
- Mail account is which connection collects it. Leave it as "not collected from a mailbox" for a send-only address.
- Folder to read is normally the main inbox. Once the mail account has been tested, this is a menu of the folders that account actually has, spelled the way the mail server spells them - which saves guessing between "Sent", "Sent Items" and "Sent Messages". Update folders beside it asks the mail server again, for a folder you have made since. Made one this minute and would rather not wait? Choose Type a folder name myself and write it in.
- Sent folder is the same menu, and optional: left as "work it out from the mail account", the server is asked which folder it treats as Sent.
- Everything in that folder belongs to this address: for a folder you fill yourself rather than one the mail server sorts. With it on, anything sitting in the folder above is collected and filed here, even when it was sent to an old address of yours that this site has never heard of - drag it in and it turns up. Post that names one of your other addresses on its To or Cc line still goes to that address; this only settles what would otherwise have been filed under Not filed. Two addresses cannot both own the same folder: point two at one folder with this on and neither claims it, and filing goes back to the addresses.
- Catch-all: one shared inbox can be nominated to take anything that arrives at the mail account but matches no address you have listed. An individual inbox is not offered it, because post nobody could place is everybody's problem and filing it into one colleague's private mailbox puts it where the person who has to sort it out cannot see it. Without one, that post is filed under Not filed, which only an administrator can see, and the settings page tells you how much is sitting there.
- Name on replies and Signature are what your customer sees when you answer. Signatures have a section of their own below.
- Who can read this inbox is covered under "Who can see what" below.
Post is filed by the address it was delivered to, then by the To line, then by Cc, then by the folder it was found in where that folder has been told it owns its post, then by the catch-all. There are no rules for sender or subject in this version.
Every address is one of two things, and you choose which when you add it.
Shared is the business's: sales@, accounts@, hello@. Your team reads it between them, conversations get handed round, and the guest list under Who can read it says who is on it - empty meaning everybody who can see the hub at all. This is what every address was before the two kinds existed, and every address you already have is one.
Individual is one colleague's own post at work: marcus@, chris@. It belongs to the person you name, and nobody else opens it - not their colleagues, and not whoever looks after the site - unless somebody is deliberately named on it under Who else can read it. Leave that empty, which is how every one of them starts, and it is theirs alone.
That last part is worth being plain about, because it is the only place on the whole site where being an administrator does not get you in:
- Whoever looks after the site can rename an individual inbox, change which mailbox it is collected from, hand it to somebody else or delete it altogether. That is the configuration, and it stays with the administrator, where it belongs.
- They cannot read a word of what is in it unless they put themselves on its list, which is a deliberate act that shows on the form for anyone to see afterwards. Not by opening it, not through All, and not through search - a snippet from somebody's individual inbox never appears in the results of anybody who is not on it.
- A promise of privacy that the site owner can quietly read anyway is not a promise, it is a label. So this one is real: the only way in is a name on a list, and there is no way to be on that list by accident.
Sharing one, for when somebody is away. Tick a colleague under Who else can read it and that address joins their inbox under Team inboxes - see Covering somebody's post. It is meant for the ordinary reasons: covering the post while its owner is on leave, or an assistant who works somebody's diary. Tick Can reply as this address beside their name as well and they can answer from it; leave it off and they can read and nothing more. The owner is always on their own list and always able to answer - it is their post - so they do not appear among the tick boxes.
A few consequences worth knowing:
- Nobody has one until you say so. Every address on an existing site is shared, and updating does not change a single one of them. Turning one into somebody's own is a deliberate act, because it takes an address away from people who can read it today.
- A conversation in an individual inbox can only be handed to somebody who can open it - its owner, and anybody named under Who else can read it.
- It becomes the address that person opens on, and its signature goes on their replies, exactly as ticking Their own inbox on a shared address does. Naming somebody who already had an address of their own moves them here.
-
Post at it is handed to them, both ways. Anything that lands at somebody's own address is put on that person's desk as it arrives, and so is any conversation they start from it - so it stops sitting in the "nobody has picked this up" pile, which nobody else could have picked up anyway. It never takes a conversation off somebody who already has it, it never touches shared addresses like
sales@, and it is a tick you can turn off: Settings → Unified Inbox → Collecting → Who gets it. - If their staff account is deleted, the post is not. The address stops belonging to anybody and falls back to whoever looks after the site, the same answer this module gives for mail it could not place. Nothing is lost, and nobody is locked out of it for good.
- A suspended account cannot be given one, since that would leave an address only a suspended person may read, which is an address nobody may read.
Mail between two of your own addresses lands in both of them. Write from chris@ to marcus@ and there are two conversations afterwards: one in Marcus's inbox, unread, where he can answer it exactly as he would a customer's, and one in yours, showing as something you sent. His reply joins both of them. Each side is a conversation in its own right - mark one done, snooze it or hand it to somebody, and the other is left alone, because the two of you are not finished with it at the same moment.
A customer copying in a colleague lands in both of your tabs. Somebody writes to hi@ and copies in emma@, and the mail server puts it in both mailboxes - so does the hub. It stays ONE conversation, living at the first address the email names (the To line before the Cc line), and it is listed in every other address the email named as well: it shows in General Enquiries and in Emma's tab, with one unread mark, one done, one snooze and one reply between you, so two people do not answer the same customer twice. Anybody who can read either address can open it. A copy only Emma was sent blind turns up the same way. It only happens when the email names two or more of your addresses: a later reply to just the one address never pulls a conversation you have moved somewhere else back out again, and colleague-to-colleague mail is still split into a conversation each, as above. Conversations collected before this arrived are sorted out by the update, except ones somebody has since moved.
Copy a third address in and it gets a conversation too. Nobody has to go looking through a colleague's tab for a message that was addressed to them.
This is not how it used to behave. Until this update a colleague email was filed once, on whichever of the two addresses the site happened to read first, and a reply followed it there - so answering a colleague could put your answer somewhere they could not see it. Conversations already collected are put right on the update: the site reads that mail again the next time it checks, and fills in the sides that were missing. Nothing is collected twice.
Each inbox has one signature, and it goes below a dividing line at the foot of replies sent from that address. That is deliberate: accounts@ signs off as the accounts department rather than as whoever happened to be typing - unless the person typing has an address of their own, which is the exception below.
There are three ways to write one, and they are the same three the contact form offers, so you only ever learn this once:
- Rich text - type it, with buttons for bold, links and lists. Right for four lines and a name.
- HTML - paste the signature your organisation already uses, tables, logo and all. It comes through as written; anything that runs on its own is removed when you save, because this markup ends up in a customer's inbox.
- Page builder - build it out of the same blocks your site's emails are built from, with your own colours and fonts already in the list.
All three are kept, so trying the builder out never loses the one you typed first. Switching back is a matter of pressing the other button.
Pasted and block-built signatures can carry fill-in tags, so one design works for every address: {{FROM_NAME}} becomes the name replies go out under, {{INBOX_NAME}} what the inbox is called, and {{EMAIL}} its address. A tag with nothing behind it becomes nothing at all rather than a pair of curly brackets in front of a customer.
Show me how it will look renders the signature exactly as the email will, before you save it, which is a cheaper way to discover that the logo is enormous.
One exception, and it is a useful one. If somebody has been given an inbox of their own (see Giving somebody an inbox of their own), the signature they wrote there is the one that goes out on their replies wherever they send from - so a purchasing manager answering out of sales@ still signs off in their own name. If their own address has no signature, the address the reply is leaving from provides it, exactly as before.
So a shared address's signature is the fallback, not the rule. On a team where everybody has been given an address of their own, accounts@'s own signature is what signs for anybody who has not - and it is what a campaign sent from that address uses, whoever set the campaign up. It is worth writing one even if you rarely see it. The other way round, an address that is somebody's own signs as them outright, whoever presses Send: a reply leaving sam@ is from Sam even when a colleague sent it out for him.
If you already had a signature before this arrived, it is still there and still goes out unchanged, sitting under HTML.
Replies go out through your site's usual email service, as the inbox they belong to. A customer who wrote to hi@ is answered by hi@, and a supplier who wrote to marcus@ is answered by marcus@. That is rather the point of having more than one inbox.
Your email service has to be willing to send as each of those addresses. With Brevo that means the address, or the whole domain, has to be authenticated in your Brevo account. Cactus checks when you save an inbox and puts a line on the settings page if Brevo will not have it, so you find out with five minutes' work in front of you rather than when a colleague is trying to answer somebody.
If one address needs to go out through a different account altogether, How replies are sent lets that inbox carry its own Brevo key or its own outgoing mail server. Most sites will never need it.
Not everything starts with somebody writing to you. Write a message sits at the right-hand end of the row with the search box, just past it, and opens a blank message in a box over the top of the list: who it is going to, a Cc or a Bcc if you want one, a subject, what you want to say, and anything from your media library attached to it - or dragged straight onto the box.
Cc and Bcc are links at the end of the To line, and each opens a line of its own when you press it. A Bcc goes to somebody without anybody else on the message being told - that is the whole of what it is for, and it is why it is a separate line rather than more names in the Cc box. An empty line has a Remove beside it to put it away again; a line with somebody on it has not, because a button that quietly dropped a recipient would be worse than one more click.
The To and Cc lines suggest people for you. Click into either one before typing a word and the addresses that inbox has been dealing with are already listed, most recent first, with the name and company where the site knows them. Type and the list narrows. Arrow keys move through it, Return or Tab takes the highlighted one, Escape puts the list away without closing the message. Pick somebody and only the part you were typing is replaced, so a line that already has two people on it keeps them.
The suggestions are that address's own correspondents rather than everybody the site has ever met, which is usually the shorter and better list. Your own addresses are never suggested, and neither is anybody already on the line. Nobody is offered an address out of an inbox they are not allowed to read. The conversations stay where they were underneath, and closing the box - the cross at the top right, or the Escape key - puts you straight back to them, offering to keep what you had written as a draft on the way. There is no Cancel button any more: a button whose whole job was to lose a screenful of typing was not worth the row it sat on.
It goes out as whichever inbox you were looking at. Standing in accounts@ and pressing Write means writing as accounts@, which is almost always what you meant. The From menu at the top of the message is there for the times it is not: pick any of your addresses and the message leaves as that one instead, with that inbox's name on replies and its signature, and the answer comes back to that inbox. Only addresses you are allowed to send from appear in the menu, so nothing on it can turn round and refuse you.
Once it has gone, you are standing in the new conversation, which behaves exactly like one somebody started for you: it can be assigned, snoozed, noted on and replied to, and their answer joins it rather than arriving as something unrelated.
If you are looking at All, at Not filed, or at one of the other channels, the menu opens on your first address rather than guessing. And if none of your addresses is one you may write from, the button is not there at all, rather than there and disappointing.
There is a small arrow beside the write button. It opens three other things you might have meant.
Discussion is a conversation with your colleagues that no customer ever sees. The box opens with the cursor already in the To line, because who it is to is the first thing about it: start typing a name, pick them off the list, and each one sits there as a tag you can take off again. Give it a subject, say your piece, and it appears alongside the real post - except that nothing on it is ever sent anywhere.
It lands in your inbox and in the inbox of everybody you put it to. One conversation, not a copy each, so a reply on it is a reply the others see rather than three half-conversations that never meet. Anybody you put it to who has not been given an address of their own still gets it on their Asked me list, which is how they were reached before. In the list it reads as your name, then theirs, the way every other conversation reads as who it is with. You are not asked which address to start it in: it starts in your own, or, if nobody has given you one, in whichever address you happen to be looking at. There is no reply arrow on a discussion, only the note line at the foot of it and, behind the dots on any message, the full note box for one with a file on it - because there is nobody outside it to reply to.
Done is done for you. Ticking a discussion Done for me takes it off your own list and nobody else's - the colleagues still working on it keep it open. The next time anybody writes in it - you included - it comes back for everyone who had closed it, the same way an answered email reopens. Putting it to sleep is still shared, as it is on any conversation.
Marking a conversation done settles your ask about it. If a colleague asked you to look at something and you mark that conversation done, the ask goes to Done in your Mentioned folder as well - only yours, not anybody else's who was asked about the same thing.
Deleting one is for you, too. A discussion you delete goes into your own Bin and disappears from your lists only - everybody else in it keeps it, even if you were covering their post when you pressed the button. Emptying your Bin takes it out of the Bin for good, but never destroys it for the others; an ordinary email is still destroyed for everyone, exactly as before.
It is the same internal note the inbox has always had, with one thing added: somewhere to put the first one. "A word about the Henderson order" no longer has to wait for the Hendersons to write in.
Call rings somebody. Your phone rings first, from whichever of your numbers you picked, tells you who is about to be rung, and puts you through when you press a key - so you are never listening to a phone ringing in an empty room. The number you dial from is the one they see, and the one they ring back, which is why it is a menu rather than a fixed value. The call turns up afterwards in the phone conversation with that number, filed by the part of the site that placed it.
You can type a name instead of a number. Start writing one into the Ring box and the contacts who have a number on them appear underneath, with the number beside each so you can see which one you are about to dial. Click one, or pick it with the arrow keys and Return, and the number goes into the box. Nobody remembers a phone number; they remember whose it is, and it is already written down here from the last time that person got in touch. Typing an actual number still works exactly as it did - the list only appears once there is a letter in the box.
Call me at is filled in from your own account, so you are not typing your mobile in from memory every time. Put your number on it once - Your account > Profile > Your phone number - and it is there whenever you place a call. It is still a box rather than a fixed setting, because the day you are sitting at a customer's desk you want the call to reach you there instead, and changing it here changes it for that call only.
Numbers are typed the way you would say them. 020 8138 0512 is a whole number, and so is 07700 900123, and so is 2081380512 - all three reach the same place. Anything without a country code is taken as your site's, which is +44 unless somebody has changed it in Settings > General > Dialling code. Type a number from anywhere else in full, with its + and country code, and it is left exactly as it is.
SMS sends a text. A number, a message, Send. The To box behaves exactly like Ring does: type a name and the contacts who have a number on them appear underneath to be clicked, and a number typed the way you would say it - 07700 900123, or 7445164234 - is taken as your site's country unless you write one in full with its + and country code. It goes out through whatever the site sends its texts with, and lands in the same phone conversation the calls do. The line under the box counts what it will cost you: 160 characters is one text, and a single curly quote or emoji - the sort of thing a word processor puts in without asking - drops the whole message to 70 characters a text for the rest of its length.
The menu only offers what your site can actually do. No texts without something to send them with, no calls without something to place them with, and no arrow at all when it can do neither - in which case the button is exactly what it always was. Both of those come from the Twilio module today; the inbox knows nothing about it beyond "there is something here that makes calls".
The arrow you press is the message you answer. Press Reply on the fourth message of a long conversation and that fourth message is what gets quoted underneath, folded away under Show the earlier messages the way every mail program does it. It used to quote whatever was at the bottom of the conversation instead, so an answer to something said on Tuesday went out with Friday's message tucked under it - confusing for the customer and no help to you either.
The same goes for Reply to all and Forward on that message's own dots, and for the copy filed alongside: the reply is threaded against the message you answered, so it lands in the right place in the customer's own mail program rather than at the end.
The subject follows the message too. A conversation keeps the subject it started with, but the messages in it often do not: a supplier answers "Quote" with "PO 1234 - revised delivery date". Any message whose subject is not the conversation's now shows it, on a Subject: line under who it is from and to ("Re:" and "Fwd:" are ignored for the comparison, so ordinary replies do not repeat the heading). Reply to that message and your reply is "Re: PO 1234 - revised delivery date", not "Re: Quote" - whether you leave the Subject line shut or open it to look. It used to be right when you left it shut and wrong when you opened it, which is the sort of thing nobody should have to know.
Nothing changes for the ordinary case. A reply started from the conversation itself, rather than from a particular message, still answers the newest one - which is what you want nine times out of ten and what the box has always done.
You can see what will be quoted before you send it. Under the writing box there is a Show the earlier messages line - the same words the conversation above uses. Open it and the message you are answering is there in full, earlier messages and all, exactly as it will appear beneath your reply. On a forward it says Show what you are forwarding and shows the From, Date, Subject and To lines that go with it. It is there to be read rather than edited: what gets quoted is decided by the arrow you pressed, so changing your mind means pressing a different one. Nothing is fetched until you open it.
A saved reply remembers which message it was answering, so one finished tomorrow morning, or set to go out on Monday, still quotes the message you wrote it against. If that message has been deleted in the meantime, the reply still goes: it simply quotes the newest one instead, because losing what you wrote would be the worse of the two.
The To line on a reply is a box you can type in. Cactus still works out who a plain reply goes to and fills it in for you - the sender, or whoever their Reply-To names - and that is what you will send nine times out of ten. But it is now the answer in a box rather than a sentence, so the one address that wants taking off, or the colleague who wants adding, is a click rather than a change of plan.
Switching between Reply and Reply to all on the message's own menu refills the line, which is exactly what those two mean. Once you have edited it yourself, your answer stands and nothing rewrites it again.
The box no longer wears a title strip. It used to spend a whole row along the top saying Reply - naming the button you had just pressed, on the screen where you are trying to read a customer's message. Which one you are writing is decided on the message itself, with the arrow or the dots beside it, and changing your mind means pressing one of those rather than anything in here.
Cc, Bcc and Subject are links at the end of the To line. None of them takes up a line until you ask for it, because most replies want none of the three. Press Subject and the line opens on what the subject would have been - Re: whatever it was - so you can change it rather than start it. Leave it alone and nothing changes.
The two ways out of the box sit at the right-hand end of that same line, where the strip along the top used to carry them. The arrow opens the reply in a window of its own, over the whole screen, for the reply that turns out to be a letter. It is the same box either way - nothing you have typed is lost going out or coming back - and the arrow in the corner of that window, or the Escape key, puts it back under the conversation. On an internal note, which has no To line, the pair sit at the end of the sentence saying nobody outside can see it.
The cross beside it closes the reply, and it is the one place the draft is asked about: Save it as a draft, Throw it away, or Keep writing - the same three answers the new-message box gives. It asks whether you have just typed something or came back to a draft you saved yesterday; only a genuinely empty box shuts without a word. There used to be a Throw the draft away button on the strip along the bottom as well, which put a destructive press a thumb away from Send and asked the same question in a second place. See Drafts.
With the Google AI Studio module installed and a key pasted into it, there is a Suggest reply button under the writing box.
Press it and the conversation - everything on it, including the internal notes you and your colleagues have left - goes off and comes back as three drafts. Click one to try it in the box, where you can see how it actually reads. The tick keeps it; the cross takes it back out and puts back whatever you had written before you started looking. Click a different one to try that instead.
From the moment you tick, it is your draft: edit it, add to it, attach something to it, and send it like anything else.
It is on a plain reply and on a reply to everybody, and on nothing else - a forward is a covering note to a colleague and an internal note goes to nobody at all, so neither is a thing to have an answer to a customer written for.
Chasing an email nobody answered. If the last message on the conversation is your own, the button says Suggest a follow-up and writes one - a chase, referring back to what you already said, with an easy way out for the customer. Without that it would write a reply to your own email, which reads perfectly well right up until you notice it is answering you. A note left by a colleague does not count as speaking to the customer, so it stays a follow-up.
Notes are read but never repeated. A note is handed over marked as something the customer has never seen, so a draft can know the chair is the last one in stock without telling them you said so.
No button? No key, or no module. Nothing is sent anywhere until you press it.
Both writing boxes carry seven buttons on the strip along the bottom, next to the paperclip: bold, italic, strikethrough, a colour, a link, a bullet list and a numbered list. Select some words and press one, or press it and carry on typing. Ctrl+B and Ctrl+I work as they always have. They used to sit in a band of their own above the words; everything you can do to a message is on the one strip now.
That is the whole list, and it is short deliberately. Typefaces, sizes and alignment are the things that make an email look assembled rather than written, and half of them are rendered differently by every inbox they land in.
The colour is behind the palette. Press it and seven colours appear beside it; press one and it is applied and they fold away again, which is what a colour menu does everywhere else.
The link button opens a line rather than a browser box, so it can tell you when what you typed is not an address. Type example.com/prices and it works out the rest. Select some words first and they become the link; select nothing and the address itself is what appears.
Pasting brings the words, not the wallpaper. Paste from a web page or another email and you get the text: the fonts, the layout and the tracking pixels that came with it are left behind, which is what you wanted and is a good deal safer besides.
Web addresses link themselves. Type or paste an address and it becomes a link as soon as you put a space or a new line after it, or move out of the box - no need to select it and press the link button. Drafts saved before this open with their addresses linked, and anything still left as plain words is linked on the way out, so a scheduled message goes with working links whether or not anybody opens it again. Clicking a link inside the box puts your cursor there, as it would in any editor, and a small bubble appears just under the link showing where it goes, with Open link to follow it in a new tab. Cmd-click (ctrl-click on Windows) opens it straight away. Links in the box are underlined, so you can tell them from coloured words.
The phone goes halfway through a reply, and until now that cost you the reply. Nothing you half-write is thrown away any more, and there is no Save as a draft button to remember to press: the box puts itself down and waits under Drafts in the list of addresses until you come back to it.
A reply saves itself the moment you go somewhere else. Click another conversation, another address, another tab - anything that takes the screen off the reply - and what you had written is put down as a draft on the way past, with no question asked. The only time you hear about it is when the save itself fails, in which case the screen stays where it is and says so rather than losing the words.
Closing one asks. The cross at the top right of the reply box - and of the popped-out window, and of the new-message box - offers three answers: Save it as a draft, Throw it away, or Keep writing. Nothing is thrown away without you saying the words. A box you have not typed in simply closes.
Drafts is a tab of its own, because it is not one address - it is everything you have half-written, across every address you write from. The number beside it is how many of yours are waiting. A message you have set to go out at a particular time is not one of them: it moves next door to Scheduled (see Sending it later), and comes back here if its timer is cancelled or its send is refused.
It is only there while there is something in it. Nothing half-written, no Drafts row: an empty folder is a place you can only ever go and be disappointed by, and being able to write from an address is not a reason to be shown one. It appears the moment a box puts itself down and goes again once the last one has been sent or thrown away - except while you are standing in it, so emptying the folder does not pull the list out from under you.
Your drafts are filed under you, not under the mailbox. A shared mailbox has no Drafts folder. A reply you begin to something in accounts@ lives in your Drafts, and - for colleagues who cover your post - in the Drafts folder under your name. Anybody who can open the conversation itself can also see that you have an answer on the go, and read it (see below); nobody but you can change it or send it.
The one exception is somebody's own address, and only for the people covering it. If Sam's address has been shared with you by name, Sam's unfinished replies show under Sam Blake → Drafts on the rail. That is a deliberate choice and worth being plain about: you are already reading every message that arrives at Sam's address and every message that leaves it, and the one thing you could not see was the answer Sam started on Tuesday and did not finish - which is exactly how the same customer ends up written to twice, once by Sam on Friday and once by you this morning. Nobody who has not been put on that address sees any of it.
You cannot change it, and you can send it. Somebody else's draft is theirs to finish: there is no writing box, no way to edit a word of it, and no way to throw it away. What you can do, if you are allowed to send from that address, is send it out exactly as it stands - a button at the foot marked Send it for Sam. It asks first, and says what happens: the message leaves Sam's address, signed as Sam, and there is no getting it back.
That is deliberately the only thing on offer. Editing somebody's half-written sentence and posting it under their name is putting words in their mouth; sending the paragraph they actually wrote is a decision about a message that already exists, taken by somebody who has just read every word of it. It is also the way out of the one case this privacy costs: the finished quote sitting on the address of somebody who is away for a fortnight.
Sending it for them needs the sending grant, not just the reading one. Being let into Sam's address to read it is not the same as being allowed to write as it (see Who can see what). If you have only been given the reading half, the foot of the page says so and no button appears.
Sent says who actually pressed it. The customer sees a message from Sam's address with Sam's signature, exactly as if Sam had sent it - but the entry in Sent credits whoever sent it, so there is no mystery afterwards about who let it go. If the draft had a chase on it, the chase is still Sam's: they asked for it.
Opening one shows the message rather than a writing box: who it is to, what it is about, what is attached, when it is set to go out if it is, and the words. If it is an answer to a conversation, there is a link across to that conversation. At the foot it says whose writing it is, that only they can change it, and - where you may - offers to send it for them.
And the conversation shows it too. Open the conversation any way you like and a colleague's unsent reply is there, opened out and marked Draft or Scheduled, right beside the newest message where a reply would sit.
A reply opens in its conversation. A colleague's draft or scheduled reply to an existing conversation opens that conversation, with their draft opened out where it sits and the page scrolled to it - so you read it beside the message it answers. It stays in their Drafts or Scheduled folder as well. A brand new message, with no conversation behind it, still opens on its own.
Opening a conversation shows your own draft on it, if you left one. There is one draft per conversation per person, which is the box, put down; saving again writes over it rather than leaving you to choose between two halves of the same sentence. Two people can each have one on the same conversation. You can see that a colleague has one, and read it: open the conversation and a dashed line above the messages says Sam Blake has started a reply and not sent it, with when it was last touched; press it and the words are there, along with who it is to. It is the thing most worth knowing before you write your own answer - two replies to one customer is what happens otherwise. It is still Sam's: there is nothing on it that changes or sends it.
Picking one back up depends on what it is. A saved reply lives under its conversation, so clicking it opens the conversation with your words still in the box, the customer's message above them, and set to whichever of Reply, Reply to all and Forward you had chosen. A message you had not finished starting opens the writing box again exactly as you left it - recipients, subject, attachments and all.
Walking away is instant. Clicking a tab, another conversation or anything else while you are half way through a reply takes you there straight away, and the draft is put down behind you. It used to wait for the save to finish before the screen moved, which on a slow line looked like a click the site had ignored. If putting it down somehow fails, it tries once more on its own.
Sending clears it away. The draft goes at the moment the message actually leaves, so nothing is left sitting in the list to be sent again next week. To be rid of one without sending it, close the reply with the cross and answer Throw it away. On the new-message box that is still a Throw the draft away button on the strip along the bottom as well.
What is saved is what you typed, line breaks and all, along with who it is going to, the subject and anything you attached. Attachments are still only pointed at rather than copied, so a draft carrying three quotes costs nothing until you press Send.
Some replies are written at half past eleven at night and would land better at nine in the morning. The alarm clock, sitting with the send buttons at the right-hand end of the strip along the bottom, opens the same short menu the clock at the top of a conversation opens - In three hours, Tomorrow morning, Next week, or Day & Time for a month to pick from - each with the day and time it actually lands on written beside it. It sits there rather than over with the paperclip because choosing a time changes what the buttons beside it do, and a control a foot away from the thing it changes is a control nobody connects to it.
Picking a time sends nothing on the spot. It puts the time on the line above the buttons - Set to go out tomorrow at 09:00 - and the two send buttons change to say what they would now do. Send now becomes Send later, and Send & snooze becomes Send later & snooze. Both stay where they are: choosing to send later used to take Send & snooze away and leave a chase in its place, so deciding to send in the morning quietly cost you the ability to put the conversation to bed. There is no third button either - there used to be a Save it for then beside them, which meant three ways to send on one strip with two of them a thumb apart. Cancel send later takes the time back off and everything reads as it did.
The time you picked is not lost if you wander off. Click onto another conversation, or close the box, and the time goes down with the writing - the message is set going and waits under Scheduled, rather than becoming an ordinary draft with nothing to show for the time you chose. Closing says so in as many words - Set it going? ... it goes out tomorrow at 09:00 on its own - so nothing is queued behind your back. It used to be thrown away without a word, which is how a message somebody believed was going out at nine turned out to be sitting in Drafts.
Once it is set the message waits until then and goes out on its own; you do not have to be there, or even have the site open.
Putting the conversation to sleep can be part of the same press. Send later & snooze sets the departure and puts the conversation to sleep behind it, so you see it again when you said and not when the message happens to leave. Or do the two separately: set the message for Monday morning, then use the clock at the top of the conversation to put it to sleep until Friday, and Friday is when it comes back - the message going out on Monday does not wake it up. Where both have been said, whichever of the two is later wins.
It waits under Scheduled, not under Drafts. Scheduled is its own row on the rail, next to Drafts, holding everything you have set going and nothing else, soonest first, with Goes out tomorrow at 09:00 on each row. Drafts is what you have not finished; Scheduled is what is going out without you, and a list that mixed the two made the Drafts number read as work outstanding when half of it was work already done. Like Drafts, the row is only there while there is something in it.
You can still change your mind about all of it, right up until it goes. Open it and:
- Change the writing - correct a price, add a line - and press Save changes. What it says changes; when it leaves does not.
- Move it to another time from the same alarm clock. The new time shows as Move it to Friday at 09:00 above the buttons, and Send later commits it. Keep the time it has drops the new time and leaves the old one alone.
- Send it now after all. The button is there the whole time it is waiting.
- Cancel the timer on the alarm clock menu, which hands you back an ordinary draft under Drafts with everything you wrote still in it.
The one thing it will not do is send twice. If you press Send now in the same few seconds the site is already posting it, you are told it is on its way and to look in Sent, rather than the customer getting it twice.
It goes at that time or shortly after, never before it. The site checks its own list on a schedule, so a message set for 09:30 leaves at 09:30 or a little after depending on how your site is set up. Anyone pressing the refresh button in the inbox sends whatever is due at that moment too, which in practice means a message due while somebody is at their desk goes almost at once.
Times are your site's times. Nine o'clock means nine o'clock as the site tells the time, whatever clock the computer you typed it on is keeping.
If it cannot go, it goes back to Drafts and says so. An address that has been taken off you between writing the message and its time coming, a mail server having a bad morning, a conversation that has since been deleted - whatever the reason, the writing is kept exactly as it was with Did not go out and the reason on the row, and nothing is quietly thrown away. It sits under Drafts rather than Scheduled, because it is not going anywhere on its own any more and it wants somebody to look at it. Fix it and send it, or set another time.
A message set to go out is still a draft underneath - it is simply one with a departure time - so it follows the same rule as one: it is yours, only you can see it, and only you can change it, move it, cancel it or throw it away, however many people share the address it would leave as.
Some replies end the matter for a fortnight: the answer is sent, and there is nothing more to do until the supplier gets back to you. Send & snooze sits next to Send now and does both in one press - it opens the same short menu the clock at the top of the conversation opens, sends the message, and puts the conversation to sleep until whenever you said. Pick a departure time first and it becomes Send later & snooze, which sets the message going and puts the conversation to sleep without sending anything now.
It is on the new-message box as well, where it puts the conversation you have just started to sleep. Set a departure time there too and the sleep is kept with the message: the conversation is made when the message actually goes out, days later with nobody watching, and it arrives asleep rather than sitting at the top of the list wanting attention. The notice on the message says so while it waits - This conversation stays asleep until Friday morning, unless they write back first.
Their reply wakes it either way, as a reply always does, so you only see it again at the time you picked if nobody has answered by then.
A message set for Monday morning was written without Monday's post in front of you, and sending it anyway is how you ask a question that has already been answered - or answer one a colleague already has.
So the conversation moving on cancels it. Any of these, before your message leaves, takes the timer off it, sends nothing, and keeps the writing exactly as it was, as a draft:
- They write on the conversation - from the address you were writing to, from another address of theirs, or somebody copied in on it. On a chat, WhatsApp or a text, it is them writing in the same conversation (these used to go out regardless).
- They write to you anywhere else. Mail from the person a scheduled message is addressed to cancels it on whichever conversation it was written in. Only the To line counts for this one, and it is email only - the address somebody types into a chat window is their say-so, not proof.
- A colleague answers first - from here, from their own mail program, or from a channel's own screens. Your scheduled reply is cancelled so the customer is not answered twice. Their own scheduled reply is never cancelled by their own message; they knew it was there. A forward from a colleague cancels nothing, because it told the customer nothing.
Each time, the conversation says so in a line just above the message that did it - They wrote before Chris's scheduled reply went out, so it was cancelled and saved as a draft, or Sam replied first, so Chris's scheduled reply was cancelled and saved as a draft. If it was written on a different conversation, that one gets a line of its own too. And whoever wrote it sees A message to them was waiting to go out across the top of the conversation, saying why and linking straight to it - read what was said, then send yours as it stands, change it, or throw it away.
Nothing is lost and nothing is sent twice. The message sits under Drafts with Held - somebody replied first on the row. Put a time back on it and it is queued again as normal.
A message already going out is left alone. If their mail arrives in the minute yours is being sent, yours has gone: the site will not pretend otherwise.
Only a real message counts. An out-of-office or a bounce is the mail system talking, not an answer, and cancels nothing; neither does post from a sender the site has blocked, an internal note, or old mail being read in for the first time when an account is connected.
If it cannot go when its time comes - whoever wrote it can no longer send from that address, say - it stays a draft with the reason on it, and the conversation gets a line saying so, so nobody reading it waits for a message that is not coming.
Sent is a tab beside Drafts, and it is one row per message that has left, newest first, across every address you are allowed to read. Not one row per conversation: a thread you have answered four times is four things you sent, and "did that quote actually go, and when" is a question about the message rather than about the conversation it sits in.
Each row says who it went to, what it was about, which of your addresses it left as and who here wrote it. Where the site is watching for it, what became of it is on the row too - on its way, opened, or never arrived (see Finding out what happened after you pressed Send). Clicking one opens the conversation it belongs to, which is where the answer to it will turn up.
Replies sent from here, messages written from here, and your other modules' automatic post that goes out as one of your inboxes all appear. So does anything you sent to a colleague at another of your own addresses, even though it is sitting in their inbox as post for them. Internal notes do not: nobody was sent one.
Off by default. With it on, every reply you send from here is also filed in the mailbox's own Sent folder, so the Mail app on your phone agrees with the site. With it off, your replies live here and your phone's Sent folder stays empty.
If the copy fails, your email has still gone. The message will say the copy did not file, which is a tidiness problem rather than a delivery one.
Sent means the email service took the message. It does not mean it arrived, and it certainly does not mean anybody read it. Two switches on Settings → Unified Inbox → Sent replies, both off until you turn them on, fill that gap.
Tell me when a reply is delivered, opened, clicked or bounces. Brevo tells the site what became of each message and the answer appears under the reply itself: Delivered 09:14, Opened 11:32, Followed a link 11:33, or It did not arrive with the reason behind it. Useful when you are deciding whether it is worth chasing somebody, and rather more useful than guessing.
Press any of those labels and the whole story opens. Every delivery, open, followed link and bounce the message has had, in order, each with the exact moment and the internet address it came from. The address is the useful part: an "open" three seconds after delivery from a cloud provider's network is the recipient's email security looking the message over, and the person themselves tends to turn up later, from somewhere ordinary. Brevo's own logs show the same thing, and this saves the trip. The address on a delivery is the email service's own machine handing the message over, which is worth knowing before anybody decides it looks familiar. Brevo is asked for the addresses the first time the history is opened and they are kept from then on; if it cannot be reached the list still opens, with a line saying the addresses could not be fetched this time.
Two honest limits. An open is worked out from a tiny invisible picture in the message, and some email programs - Apple Mail and Gmail among them - fetch that picture themselves before anybody has looked at a word. When that is what happened you are told their email app fetched it rather than told a fib about somebody reading it. And it only works for inboxes sending through Brevo: an inbox on its own outgoing mail server carries on saying nothing but Sent.
Followed a link sits beside the open rather than instead of it, and answers a different question: not whether they looked at your email, but whether they looked at the quote in it. A message can easily have one without the other, since plenty of email programs never load the picture an open depends on. It needs one thing switched on at the other end - link tracking in your Brevo account, which is what puts Brevo in the middle of the links in your messages so it can see one being used. If nothing ever says Followed a link, that setting is the first place to look, and switching it on costs nothing here: the site is already listening, and it reminds the email service what to tell it once a night in case anything has drifted.
One caveat, and it is the same sort of caveat as the invisible picture. Plenty of offices run every arriving email past a security system that opens every link in it to check where they go, and it does the lot within seconds of the message landing. Several clicks all at once, a minute after delivery, is more likely to be that than a person - which is why the site counts each address separately and shows you the time, rather than handing you a single number with no working.
Ask the person's own email program for a read receipt. The old-fashioned kind, from the days of office memos. Most programs ignore the request and the rest ask the reader whether to answer it, so expect a reply perhaps one time in ten - mostly from people in offices. When one does come back it lands on the message it belongs to rather than turning up in the conversation as a mysterious email titled Read: your quote, which is what happens everywhere else.
All of this amounts to keeping a note of what somebody did with an email you sent them. That is yours to decide, but if your privacy notice does not mention it, add a line before you switch them on. Nothing is recorded for any message sent before you did.
Everything above is Brevo telling the site what it saw. An inbox set to Sent through: Its own mail server - an iCloud address over SMTP, say - has no Brevo in the middle, so the site does the counting itself. It works the same way Brevo does: an invisible picture on the end of each message, and every ordinary link sent through the site's own address first (yoursite/api/email/c/...) and straight on to where it was going. The same labels appear under the reply - Opened, Followed a link, It did not arrive - and in the history each line now says who told you: Reported by Brevo, Seen by this site, or Sent back by their email program. While you are moving an inbox off Brevo you may see both on one conversation; they are never merged into one line, because they are counting different sends.
What an ordinary mail account cannot give you is Delivered. The mail server takes the message, says "accepted", and goes quiet unless something goes wrong. So the label says exactly that - Accepted by the mail server - and once two days have gone by with nothing bouncing back it changes to Accepted, nothing bounced back, which is as close to Delivered as anybody can honestly get. If it does bounce, the bounce comes back to the mailbox you send from; as long as the inbox collects that mailbox, the bounce is matched to the message it is about and the label turns to It did not arrive, with the reason in the history. The bounce itself still appears on the conversation too, as the machinery it is.
A few things are never touched: mailto: and phone links, links within the message, anything that looks like an unsubscribe or preferences link, and links already going through Brevo. Campaigns are left entirely alone - they stay on Brevo, which counts them and carries the unsubscribe link, even when the site's own mail goes by SMTP. And the site never counts its own staff: a link followed by somebody signed in to the admin is not recorded, pictures in our own sent mail and in quoted copies under a customer's reply are never fetched, and a tracked link in your own sent mail opens straight at its real address when you press it here, rather than going through the counter.
Opens are a rough guide, clicks are the signal. Apple Mail downloads every picture the moment a message arrives - recorded as Their email app fetched it, never as an open - and Outlook in an office often blocks pictures, so a message read there may never register as opened at all.
It is on by default, per inbox (Notice when replies are opened and their links followed, shown when the inbox sends through its own mail server) and for the whole site (Settings → Emails → Delivery, which has the last word). A message sent through Brevo never carries the site's own picture or links: Brevo already counts it, and two counters on one message would count every open twice.
You can attach anything from your media library. There is a ceiling of 20 megabytes for the message and everything attached to it, which is what the email service itself allows, and an attachment that will not fit is refused before you send rather than quietly dropped on the way out.
A file you drag onto the message is put on your media page once it has gone, under Inbox, the address you sent it to, and Sent. A file you picked out of the media library is left exactly where it already lives: it is one file, not two, and hauling a product photograph out of its shop folder because somebody emailed it would break the page pointing at it.
Attach a file is the paperclip at the left-hand end of the strip along the bottom, beside the price tag and the formatting buttons, with the ways to send it at the other end. On a narrow window the strip wraps and the send buttons stay hard right.
Pressing it opens a box of its own, and the box is mostly one big rectangle: drag your files onto it, or press Choose files and pick them off your computer. Take as many as you like, in as many goes as you like - the box stays open until you press Done, so attaching six things is one errand rather than six trips. Above the rectangle there is a search over your media library, for the other case: type a name and the results stand where the rectangle was, empty the box and the rectangle comes back. Anything already on the message is shown greyed out there, so nothing gets attached twice.
What you have put on is listed underneath, a file to a line, with its size beside it and a cross on the end. Attach the wrong one and you can take it straight back off without closing the box.
You can also drag a file straight onto the message without opening anything - onto the reply box under a conversation, or anywhere on the box where you are writing a new one. Drag several at once and they all go on. The box says "Drop to attach" while the file is over it, and either way the files appear as tags on their own line just above the buttons as they arrive, with a cross to take one back off. Nothing is sent until you press Send, as ever.
Files brought in this way - dropped, or chosen off your computer - can be up to 20 megabytes each. A small file travels through the site in the ordinary way; a larger one goes straight into the site's file storage first, neatly stepping around the hosting platform's four-megabyte request limit. On a storage provider that cannot take a direct browser upload, the box says so and keeps the four-megabyte limit rather than pretending otherwise.
There is no sensible limit on how many files one message may carry. The 20 megabytes above is the real ceiling and it is the one you will meet: forty delivery notes may be fine, four photographs straight off a phone may not, and either way you are told before anything is sent rather than after.
A few things are refused, and each says why: a folder (drop the files inside it instead), and anything that is a program or a script rather than a document, because the email services will not deliver one of those anyway and finding that out at Send is finding it out too late. If you need to send one, put it in a zip.
A dropped file goes into your site's storage, not into your media library: it is a customer's paperwork, and it has no more business in the picker you build pages from than an invoice that arrived in accounts@. A file you drop on and then think better of - never sent, never saved as a draft - is cleared away by the site a week later. A draft holding one keeps it for as long as the draft lives.
Half the questions a shop gets are answered with a product. The price tag beside the paperclip opens your catalogue, and what you pick is printed in your message the way the lines of an order confirmation are printed: the photograph, the name, and the price, each one linking to its own page.
The catalogue opens over the screen rather than inside the message. It used to unfold underneath what you were writing, which pushed the message itself off the bottom of the pane the moment you went looking for a chair - and a catalogue is a thing you rummage in, so it is the wrong thing to put in the middle of a sentence. Close it and the message is exactly where you left it.
Three things can go on, and the difference is worth knowing:
- A product, for something that comes one way only.
- A listing that has variations - the chair that comes in eight colours. It goes on quoted from its cheapest ("From £419.00") and links to the page where the customer picks, which is usually what you want when somebody has asked what a range costs.
- One exact variation, when they have asked about the black one with the high back. Choose one of eight under the listing opens the ones that can actually be bought, and the options that make each one are printed under its name in the message.
A long range is narrowed rather than scrolled. Open a listing that comes in six hundred combinations and you get a short row of menus above the list - one for each thing that actually varies: Width, Storage, Finish, Leg Finish. Choose the ones you have been asked about and the list comes down to what is left; Show them all puts it back. Anything that is the same across the whole range gets no menu, because choosing it would rule nothing out.
Prices carry "+ VAT" where the tax genuinely applies and your shop quotes its prices without it. Nothing is added where the price on your site already includes the tax, or where the product is not taxed at all - and if your shop calls its tax something other than VAT, the message calls it that too.
There is no heading row and no quantity: this is you showing somebody two chairs, not sending them a receipt for something they have not bought. Delivery times are left off for the same reason - what a thing costs and what it looks like is the answer to "which one did you mean", and how long it takes to arrive depends on where it is going.
It goes where your cursor is. Picking a chair drops it into the message at the point you had reached, and you carry on typing underneath it - so "this one is the cheapest", the chair, "and this one is what you actually asked for", the second chair, reads in that order when it arrives. Move the cursor, pick another, and that one lands there instead. They used to all pile up under the writing in the order you picked them, whatever the message said, which is the wrong way round for the commonest thing anybody writes.
You see it as they will see it. Each one is drawn in the message laid out the way it will arrive - photograph, name, options, price. A cross on the corner of it takes it back off, Backspace does the same, and the catalogue stays open while you add several.
You can always write above and below one. There is a line either side of every product - above the first one even when it is the very first thing in an empty reply, and between two of them that you added one after the other - so nothing you have put in ever leaves you with nowhere to type.
Prices and names are read at the moment you press Send, not when you picked - so a quote you saved on Friday and sent on Monday goes out at Monday's prices, and anything withdrawn in between is simply left off rather than quoted at a price you no longer sell it for. That is also why the block in your writing stays on a white background whatever theme you are using: it is a picture of what leaves the building, not part of the screen you are looking at.
The picker is only there if your site has a shop and you are allowed to see it.
Internal notes carry products too, and a discussion is nothing but notes. Asking a colleague "is this the one they mean?" is the shortest errand in the building, and it used to be the one thing the catalogue would not do: the price tag was hidden the moment you switched to a note, and a discussion opens on the note box and stays there, so the one conversation whose whole purpose is look at this one was the one that could not. Pick a product in a note and it is written out under your words - name, the options it comes in, the price and the address to look at it - and it joins the conversation's context line like anything else. Nothing is sent to anybody: a note is still only ever seen by your colleagues.
Text messages and calls are still the exception, notes included: the catalogue is kept off those two conversations altogether.
Answering an enquiry, a chat or a WhatsApp message? The catalogue is there too. Those channels go back out through the module that owns them and carry words rather than a laid-out table, so what arrives is the same lines the plain-text half of every email carries: the name, the options it comes in, the price, and the address to look at it - each one where you put it. Text messages are the exception, and not because of the pictures: a text is billed by the character, and three chairs quietly trebling the cost of one is not a decision to make on your behalf from a button with a price tag on it.
Your site sends plenty of email nobody types: an order confirmation, a purchase order to a supplier, a note that goods are on their way. All of it has always gone out as the one address on Settings → Emails, which is fine until somebody replies to it - and then a delivery question and a supplier's proforma are both sitting in the site's general post, waiting for whoever reads that to work out who they belong to.
Worse, none of it is anywhere you can read it afterwards. Email your site sends for you goes straight out through your sending service; it never passes through a mail folder, so it never appears in one. The order log says a despatch note went. It does not show you what the customer was told.
So two of the modules that send the most carry a box of their own, under Where this email goes, with a question apiece:
- Settings → Purchase Orders → Where this email goes - for your purchase orders, chases, cancellations and returns notes.
- Settings → Shop → Notifications → Where this email goes - for your order confirmations, despatch notes and the rest of your customer post.
Send these from. Pick one of your inboxes and that mail goes out as that address, under the name it answers on, and replies come home to it. A supplier answering a purchase order lands in the inbox the people chasing that order are already reading.
Keep a copy of these in. Pick one of your inboxes and every one of those emails turns up there as a conversation of its own, holding exactly what went out - the wording, who it went to, who was copied in, and any document that travelled with it. When the customer or the supplier answers, their reply lands underneath it, so whoever picks it up is reading both halves rather than an answer to a question nobody can see. It arrives already read, because you sent it: it does not mark the inbox unread or count as something waiting for you.
These copies carry the same Delivered, Opened and Followed a link labels a reply you typed does, so a confirmation nobody ever opened looks different from one where the customer went straight to their order - which is usually the question being asked about an order email. For a long while they carried none of that: the emails were sent by the shop or by purchase orders and the copy was filed here afterwards, and the labels only knew about mail written in the inbox itself. That is fixed, and it fixed itself - nothing to switch on, and nothing to press. It still needs Tell me when a reply is delivered, opened, clicked or bounces switched on further up this page, and Followed a link still needs link tracking on your Brevo account.
The two are separate on purpose. Keeping copies is the one most sites want, and it changes nothing a customer sees - so you can switch it on without touching the address your confirmations arrive from. Equally you can change that address and keep no copies. Either box works perfectly well on its own, and each saves the moment you choose - there is no separate Save to press.
Leave both as they come - the site's usual address, and do not keep a copy - and nothing whatsoever changes.
Four things worth knowing:
- The box only appears if you may manage this module's settings. Somebody who runs the shop but has no business repointing the site's mail does not see it.
- Only the modules you have asked for copies of are filed. Everything else carries on exactly as it did, and nothing core sends on its own account - a password reset, a sign-in code - is ever filed as a conversation. Those are not conversations, and they have no business sitting in a shared inbox.
- The document is kept only if your site has file storage set up. Without it the message is still filed, in full, with no paperclip - rather than a paperclip that opens onto an error.
- Deleting an inbox no longer takes the setting with it. A module pointed at an address you retire falls back to the site's usual one for sending, and keeps filing wherever you told it to file - rather than quietly forgetting both.
- Delete an inbox and any module pointed at it quietly goes back to the site's usual address. Nothing stops sending.
Automatically, once an hour on a paid hosting plan and once a day on the free one. That is the hosting plan's limit rather than a choice.
For when you are waiting on something, there is a refresh button in a box stuck to the foot of the rail, down the left of the inbox. The box stays where it is however far you scroll, and beside the button it says Updated and the time the post last arrived, so "has anything come in" is answered before you press anything. Press it and the site goes and fetches your post there and then, from every mail account at once, and tells you what it found in a blue line above the button. The line reads for three seconds and then fades away on its own, because it is an answer to a press rather than something to be dismissed. Something that went wrong stays up until you press again. A mailbox opened seconds ago is left alone rather than opened again, so that your mail provider does not take it personally - press anyway and the list still refreshes, and it tells you how recently your mail was looked at rather than refusing you. The same thing per account still lives on the settings page as Check now.
The refresh button only appears if you look after the inbox and there is a mail account for it to check - there is no sense offering a button whose only possible answer is no.
If you would rather not keep pressing it, Settings → Unified Inbox → Collecting → Check for new mail while the inbox is open does the pressing for you: pick a wait - a minute, two, five, ten, or half an hour - and the inbox goes and looks that often on its own while you have it open in the tab you are looking at. Anywhere else in the admin area - Settings, an order, a tab left behind another window - it keeps collecting, but only every five minutes (or at your wait, if you picked a longer one), and several admin tabs open at once take turns rather than each checking on its own. Come back to the inbox and it catches up straight away. Close the page and it stops. It says nothing at all - the time beside Updated moves on, new post simply appears in the list, in bold, where post has always appeared, and nothing else happens. A banner across the top announcing mail that is already on the screen underneath it is one thing to read too many. Only people who look after the mail accounts get it, for the same reason the button is only theirs. Off to begin with, and worth picking the longest wait you can live with: every check is a bit more work for your hosting.
The first collection does not fetch everything at once. New post is picked up straight away, and the history is walked backwards a batch at a time over the following checks until it reaches as far back as you asked for under Collecting → How far back to go (twelve months to begin with). A mailbox with years in it takes a while to fill in, and the settings page shows the progress while it does.
The first time you open the inbox, a card appears in the middle of the screen asking whether you would like a nudge when something new lands. Say yes and your browser asks you to allow it, in its own words, once. From then on, new post pops up in the corner of the screen the way a message from any other program does, and clicking it opens that conversation.
Say No thanks and you are not asked again. The bell at the foot of the rail is still there if you change your mind later, and pressing it turns nudges on and off whenever you like.
A few things worth knowing, because they are the questions everybody asks:
- It watches everything you can read. Your own address, every shared address you are on, and any colleague's address you have been let into - the Yours, Shared inboxes and Team inboxes sections of the rail. Each nudge says which one it landed in. Live chat, calls and texts bring their own alerts and are left out.
- The tab can be closed. In Chrome, Edge, Firefox and Safari on a Mac, nudges come through the browser's own delivery service, so they arrive with every tab of the site shut. The browser itself has to be allowed to show notifications, which on a Mac is under System Settings > Notifications.
- On an iPhone or iPad, add it to the Home Screen first. Apple only passes notifications on from apps on the Home Screen, not from an ordinary Safari tab. Open the inbox in Safari, tap Share, then Add to Home Screen, open the new icon, sign in, and press the bell there. In an ordinary tab, the bell explains this rather than doing nothing.
- It is per browser, per person. Saying yes on the office machine does not say yes on your laptop or your phone, and each one you say yes on is told separately. Two people sharing a computer get their own answer.
- One nudge per conversation. A second reply on the same conversation updates its nudge rather than stacking another, and several conversations arriving together come through as a count with the names underneath.
- Mail written a while ago still counts. What matters is when the site collected it, not the date on the email - so a message that sat at the mail provider for ten minutes before the inbox fetched it is still news. Anything dated more than a day before it arrived is treated as history and stays quiet.
Already had nudges switched on before they could arrive with the tab closed? You will be asked once whether to move over; say yes and your browser may ask for permission again.
If nothing happens after you have said yes, it is almost always the browser or the computer rather than the site: notifications can be switched off for the whole browser, or for this site, in its own settings, and a page cannot undo that from the inside. Windows also has a Focus assist setting, and macOS and iOS a Focus or Do Not Disturb, either of which will hold everything back quietly. When your browser has blocked us outright the bell is greyed out and says so if you hover over it.
The inbox is laid out the way a mail program is: one box the height of the window, split into columns, each scrolling on its own. Nothing on it scrolls the page.
Down the left is the rail - everywhere you can go, in a list rather than a row of tabs. Your own name and picture sit at the top of it, with two buttons beside them: a magnifier that searches everything, and the pen that starts a new message. Under that, five groups:
| Group | What is in it |
|---|---|
| Yours | Your own address if you have been given one, then All, Mentioned, Drafts, Scheduled and Sent. The handful of places one person opens all day, in whatever order you drag them into. Drafts and Scheduled are only there while there is something in them. Underneath is a More button holding the two nobody opens on purpose - Spam and Bin. |
| Shared inboxes | The addresses the business owns, each with a small arrow in front and the number of open conversations beside it. Press one and its Sent folder appears underneath - everything that has left that address, whoever wrote it. |
| Team inboxes | Colleagues whose own post you have been let in to, named after the person rather than the address. Press one and their folders open out underneath - Drafts, Scheduled, Sent, Mentioned, Spam and Bin. Only there if somebody has shared one with you. |
| Channels | The chat, form and phone conversations, where another part of the site owns them. Each wears a small picture of what it is - a handset for the phone, a handset with a screen for WhatsApp, a speech bubble for a live chat, a sheet of paper for a contact form. |
| Everything else | Contacts (the address book), Campaigns if you are allowed them, and, for whoever looks after the site, Not filed. |
The number beside an address is what is still open on it. Not what has arrived unread: a conversation you have looked at, thought about and left is still a job, and a badge that emptied itself the moment you glanced at the list said nothing about the work in front of you. It is the same number the list shows when you click the address, because clicking one lands you on Open - so the rail and the head of the list agree. Something snoozed until Thursday is not in it, having been decided about, and neither is anything marked done. Spam and Bin are the exception: those numbers are unread post, because a folder nobody empties would otherwise sit at a permanent 47 and stop meaning anything. How much is actually in a bin is written at the head of the folder, beside the button that empties it.
Spam and Bin live behind More. They are the two places you go when something has gone missing rather than as part of a morning, and two rows both meaning "gone" sitting in the middle of the handful of places you actually live in are two rows to read past every time. Press More at the foot of Yours and they appear; press Less and they fold away again. While anything is waiting unread in either of them the button says how much, so folding them away cannot hide the one thing about them worth noticing - and standing in either folder leaves it showing, so the row telling you where you are is never the row that is hidden.
The refresh button sits in a box stuck to the foot of the rail, with Updated and the time your post last arrived beside it (see How often it checks), and beside them the bell that turns browser nudges on and off (see Being told when post arrives). The box does not scroll away with the list.
Channels wear a picture rather than a dot. They used to carry the same coloured dot the addresses do, which told one row from another and nothing else - and a group of three rows that are each a genuinely different kind of thing deserves better than three identical marks. The picture is chosen from what the channel calls itself, so anything installed later gets the right one without a setting: a channel that calls itself WhatsApp gets the handset with a screen, a live chat gets the bubble, a phone gets the handset. Anything nobody has a picture for keeps its dot.
The arrows. Every shared and team inbox wears a small arrow in front of its name, pointing right while its folders are tucked away and down once they are showing. It used to be a coloured dot that only turned into the arrow under the pointer, which hid the one thing the mark was for. The phone's bar along the top still shows the chosen inbox's colour.
The coloured dots that remain are worked out from the address itself, so a site with six inboxes gets six different colours without anybody being asked to pick them, and the same address is the same colour on every screen and for every colleague. There is nothing to set and nothing to keep in step. Your own address is the exception: it wears a small figure rather than a colour, because there is only ever one of it and saying "this one is you" is more use than saying which of five colours it happens to have landed on.
Opening the hub takes you to your own inbox. Not to All, not to wherever you were last: /hq/inbox with nothing after it is your own address, every time.
What has been handed to you is in that inbox too. A conversation in accounts@ that somebody puts on your desk shows in your own inbox beside the post that arrived there, and its number goes up to match. There is no separate "assigned to me" list any more, and there was one - it went because a list you have to remember to check is a list that goes unchecked. Nothing is moved or taken away from anybody: the conversation is still in accounts@, still in the list and the count there, for everybody who reads it.
The order of the Yours group is yours. Drag any of those entries up or down - or hold Alt and press the arrow keys - and they stay where you put them. Unlike the shared addresses and the channels below, this needs no permission and changes nothing on anybody else's screen, because it is a fact about how you work rather than about the site.
Mentioned answers the other half of the same question. Being handed a conversation and being asked about one are two different things: the first puts the whole thing on your desk, the second asks you about a bit of it. Everything a colleague has tagged you in gathers here, with its own To do, Later and Done across the top - see Asking a colleague to look at something.
The inbox borrows its finish from a Mac and an iPhone. Controls that change what the list shows are frosted-glass capsules and circles rather than framed boxes; menus and dialogs float over a gentle blur of whatever is behind them; and on a Mac the screen uses the Mac's own typeface.
- The search box is a single rounded capsule with the magnifier inside it.
- Filter and sort share one capsule of two halves beside it: three lines for the filters, and the up-and-down arrows for which end of the list comes first.
- Open, Snoozed, Done and All are a row of separate pills. Only the one you are on says its name - the rest are a picture each: an envelope for Open, an alarm clock for Snoozed, a tick for Done, a tray for All (and a figure for Unassigned on a shared inbox). Hover over one, or listen with a screen reader, and it says its name and how many are in it. The one you are on is filled in solid, in the text colour rather than any accent, so it reads the same in light and dark.
- Rows on a computer are rounded, with the one you have open picked out as a soft tint, the way Mail on a Mac draws them. On a phone they stay full-width, as an iPhone's lists do, because they slide sideways there.
The same mail program, one pane at a time. Below about 1200px there is no room for a column of places down the left, so the rail becomes a bar along the top saying where you are - the inbox's coloured dot, its name and what is still open in it - with the magnifier and the pen beside it. Press the name and the whole rail slides in from the left as a drawer: the same five groups, the same headings, the same order, your name and picture at the top of it and the refresh button and bell at the foot. Pick a place and it slides away again; the cross at the top, the dimmed page behind it or the Escape key put it away without choosing. It used to be one strip that scrolled sideways, with half the addresses off the edge of the screen and nothing to say so.
On a tablet (roughly 900px and up) the list stays beside what you have opened, exactly as on a desktop. On a phone you see the list, or the thing you opened from it, never both: a ‹ at the start of the subject line takes you back to the list, and the box fills the screen edge to edge under the admin's own bar rather than sitting in a frame. Every pane scrolls its own contents, so the conversation's subject and its buttons stay at the top while you read, and the note bar stays at the bottom. The buttons on the subject line run sideways under your thumb rather than piling up in rows.
On a phone, swipe a conversation sideways to deal with it without opening it. Swipe it left and Done and Snooze come out from the right-hand side; swipe it right and Junk and Delete come out from the left. Press one and it happens, with the same five seconds of Undo the buttons elsewhere give you - Junk still asks whether to turn the sender away, and Snooze still asks when. Let go short of halfway and the row slides back shut; tap a row that is open and it shuts rather than opening the conversation. Only one row is open at a time. Something already done offers Reopen in place of Done, and in the Bin, Delete becomes Put back and Junk is left off.
Swipe in from the left-hand edge of the screen and the drawer of mailboxes slides out, the same drawer the name on the bar opens; swipe back to the left to put it away. It starts from the very edge on purpose, so a swipe that starts on a row still opens that row. On a phone the drawer is also a third narrower than on a tablet, so the list you are choosing from stays in sight beside it. (If the site is open in Safari rather than added to the home screen, Safari may take a swipe from the very edge as "go back" before the page sees it - the button on the bar always works.)
On a phone the search, the filter and the sort sit at the bottom of the list rather than the top, where your thumb already is: the filter-and-sort capsule on the left, the search in the middle, and the pen on the right. The pen writes an email; the arrow on it opens upwards for a Discussion, a Call or an SMS instead. The pen that used to live in the bar at the top of the screen steps aside while this one is showing. The Open/Snoozed/Done tabs and the folder's name stay at the top.
Dialogs and menus come up from the bottom edge on a phone - the new-message box, the search, the file and catalogue pickers take the whole screen, and every short menu (assign, snooze, the dots on a message, the pen's arrow) is a sheet along the bottom rather than a small panel pinned to the button that opened it. Anything pressed with a finger, anywhere the screen is a touchscreen, is given a target a finger can hit.
An individual address is one colleague's own and nobody else's, until somebody is deliberately named on it - which is the whole point of it, and no use at all when that person is on a fortnight's leave and their suppliers are still writing in.
So an individual address can be shared with named colleagues. Whoever looks after the site does it on Settings → Unified Inbox → Inboxes, editing the address and ticking a name under Who else can read it. Nothing else opens one: not being an administrator, not being on some other address's list.
Once somebody is on it, that address appears in their rail under Team inboxes, named after the colleague rather than after the address - Sam Blake, not sam@, because that is how anybody covering Sam's post thinks of it. It sits flush with every other name on the rail; there is no arrow in front of it and no indent. Press the name and their folders open out underneath it - going to somebody's post is the intent to look at their post, and the folders are half of what is in it. Put the pointer on the row and the coloured dot turns into a › you can press to fold them away again while you are in there; go somewhere else and come back and they are showing once more.
| Folder | What is in it |
|---|---|
| Drafts | What they have started and not sent - from their own address, or as a reply to something in a shared mailbox you can also open. Only there while there is something in it. You cannot change one; you can send one out as it stands if you may send from the address - see below. |
| Scheduled | What they have set to go out on its own, from their own address or a shared one you can also open. Only there while there is something in it. You cannot change one or cancel its time; you can read it so you know not to answer the same customer twice. |
| Sent | Everything that has left that address, and everything they have sent from a shared mailbox you can also open. A reply Sam sends from purchasing@ is in Purchasing's Sent folder because it left that address, and in Sam's because Sam wrote it. |
| Mentioned | Conversations on that address that its owner has been tagged in. |
| Spam | What has been marked as junk out of their post, including anything you marked while covering. |
| Bin | What has been deleted out of their post, including anything you deleted while covering. Nothing in it is destroyed until somebody empties it. |
All five are theirs, and Drafts is the one worth pausing on. Covering somebody's post means knowing what they have half-answered as well as what they have answered; not knowing is how a customer gets two replies from two people. So the folder holds Sam's unfinished writing - all of it that you could otherwise read, including the reply Sam began to something in a shared mailbox. A draft belongs to whoever is writing it, not to the mailbox the conversation happens to live in: shared mailboxes have no Drafts folder, and a reply started in one used to be filed somewhere nobody else could look. Drafts and Scheduled only appear under a colleague's name while there is something in them (and while you are standing in one, so emptying it does not pull the row out from under you) - nine colleagues with nothing half-written used to mean eighteen empty folders. Opening one shows the message rather than a writing box: only Sam can change, finish or throw away one of Sam's drafts. If the draft is on Sam's own address and you may send from it, the foot of the page offers to send it for Sam exactly as it stands - see Drafts. A reply Sam began on a shared mailbox has no such button: you can simply answer that conversation yourself. Your own half-written writing, wherever you left it, stays under Drafts in Yours.
The list is headed with the colleague's name in front of the folder - Sam Blake · Drafts - the same way their Sent, Spam, Mentioned and Bin lists are.
The name itself is their inbox, exactly as a shared address's name is. Press the › again and the folders fold away. On a phone or a tablet, where there is no pointer to hover with, the arrow is simply always there.
What being on the list does and does not give you:
- You read that address, and you answer from it only if Can reply as this address was ticked beside your name. Reading and answering are two separate grants everywhere in this hub, and this is no exception.
- Their conversations join your All view and your search results, because you can genuinely open them. Nobody else's do.
- Their Mentioned folder is read-only to you. You can see what colleagues have asked them about, which is exactly the point when you are covering for them; you cannot mark it done or put it off until Thursday. Somebody else's list of jobs is theirs to work through.
- You cannot edit their drafts, and you can send one for them if Can reply as this address was ticked. The folder is there so that knowing what they have already half-answered stops you answering it a second time; the button is there so that a finished message does not sit waiting out somebody's fortnight in Spain. Changing the words, or throwing one away, stays theirs.
- It is one address, not a person. Being let into
sam@does not let you into anything else of Sam's, and it does not follow them if the address is later handed to somebody else. - Untick the name and it is gone again, immediately and completely - out of the rail, out of All, out of search.
Your own address never appears here. It is under Yours, at the top, where it always was.
Next to it is the list, with its own head: the search box first, then where a conversation stands - Open, Snoozed, Done, All - each with the number behind it and the total at the end of the row, then the filter button beside the search box. It is a switch, not a menu: press it for the things you have not read, press it again for the lot. Who something is assigned to is what the Unassigned tab and the conversation's own header are for, so there is no list of names behind it. Whatever you switch on gets a small tag under the tabs with a cross on it, so a filter you set last Tuesday is never quietly still on. All of that stays put while the conversations scroll underneath it, which it did not before.
On a shared address there is one more tab, in front of the rest: Unassigned. It is the open conversations nobody has picked up yet, with its own number beside it, and it empties as the morning is worked through - take one and it leaves the tab. It is deliberately only on addresses the team shares. Your own inbox is on your desk by definition, and across every address at once "has anybody taken this?" is a question with no one team behind it, so the tab is not offered in either place.
Newest or oldest first. The button beside the search box turns the list round. Newest first is the ordinary way to read an inbox; oldest first is how you clear a backlog, because working from the bottom means the top does not keep moving while you do it. It is remembered in the address like everything else here, so it can be sent to a colleague, and it goes back to normal on its own the next time you open the inbox fresh.
Each conversation is three lines - who it is from and whose desk it is on, what it is about, and how it begins - with the date at the top right. Unread ones are in bold with a dot beside the name. Two more things ride on a row: a turned arrow on the last line when the last word was yours, which answers "am I waiting on them, or are they waiting on me" without opening anything, and a number on the right when there is more than one message in it.
Opening a conversation, searching, writing a message and every other button on this screen now change only the part of the screen that changed. They used to fetch the whole page again, sidebar and all, which is why the inbox had a habit of blinking at you between clicks. The address in your browser still says exactly what you are looking at, so a view can still be sent to a colleague and the back button still behaves.
To the right of the list is the reading pane, where whatever you open lands: a conversation, somebody's page, a contact's card. With nothing open there is no pane at all - the list takes the full width rather than sharing it with half a screen announcing that there is nothing in it. On a wide screen a fourth column joins it, carrying what the rest of your site knows about the person you are reading - their orders, their quotes, their invoices.
On a phone you see one at a time: the list, or the thing you opened from it, with Back to the list to return.
The columns are yours to set. Point at the line between the rail and the list, or between the list and the reading pane, and drag it. How wide a list wants to be is a question about the person reading it rather than about the screen: somebody working through a hundred a day wants a wide list and a narrow rail, and somebody reading long threads wants the opposite. Double-click a line to put it back where it started. You can also do it from the keyboard - tab to a line and use the arrow keys, with Shift for bigger steps and Home to reset it.
Your widths are remembered in your own browser, so they are yours rather than the site's and nobody else's screen changes. The reading pane always keeps enough room to read in, so the other two stop rather than squeezing it out. On a narrower window, where the rail has already lain down into a strip along the top, there are only two columns and no lines to drag.
Closing one again. There is a × at the top right of anything open, level with its title, which shuts it and gives the width back. On a phone the same link reads Back to the list, because there the list is not on the screen at all. The same goes for a person's page.
Junk and deleting are part of covering somebody too: what you mark as junk or delete out of their inbox goes into their spam folder or their bin rather than yours, and both sit under their name beside Sent and Mentioned. See Junk, and turning a sender away and The bin.
Nobody appears as a picture or a circle of initials anywhere in the inbox any more - not on the list, not beside a message, not against a contact, and not at the top of the rail. Hardly anybody had a picture to put in one, so it was a column of two-letter circles taking up the width a list on a laptop needs for the name. The name is simply the first thing on the row. The old Show people's own pictures setting has gone with them, and so has the little phone, chat or form mark that sat on the corner of the circle: how somebody got in touch is no longer said on the row at all, since who they are is the question.
Working through several at once. Rows are picked the way a mail program has picked them for thirty years: a plain click opens a conversation, cmd-click (ctrl on Windows) adds or removes one on its own, and shift-click picks everything between the last one and this one. Space and shift-space do the same for anybody on the keyboard. Pick a few and a bar appears above the list offering the things you would otherwise open each of them to do: Done and Reopen in words, then the same small pictures the conversation itself uses - an open envelope for read, a sealed one for unread, a clock for snooze, a no-entry sign for junk, a basket for delete, and in the Bin an arrow lifting them back out. Hover over any of them, or reach it with the keyboard, and it says what it is. Six mailing lists on a Monday morning is now two presses rather than twenty-four. Anything that will not change says so and the rest still go through.
While anything is picked, every row grows a small circle at its front - empty on the ones not picked, ticked on the ones that are - so you can see the whole pick at a glance.
On a phone, press and hold. There is no cmd key on a phone, so hold your finger on a conversation for a moment and it is picked, with every row showing its circle. From then on a tap picks or un-picks another rather than opening it, exactly as Mail on an iPhone does. Press Cancel in the bar, or un-pick the last one, and taps open conversations again.
They are pictures rather than sentences because they used to take three lines of a narrow list column and push the conversations down the screen every time anybody picked one. The two that are still words are the two that say where a conversation stands rather than what is about to happen to it, and no drawing means "filed" to anybody. Clear, at the far end, drops the pick and changes nothing else.
Only the buttons that would actually change something are drawn. Pick six conversations you have already read and there is no read button to press; pick six that are all open and there is no Reopen. One out of six is enough to bring a button back, since that is exactly when you want it.
What is picked, the open conversation obeys too. Pick six, open one of them to check it is the right pile, and the bin, the junk sign, the clock and Done at the top of that conversation act on all six rather than on the one you happen to be reading. It used to file exactly one, with the other five still ticked behind it and nothing on the screen admitting it. The button says how many it is about to take with it when you hover over it, the question the junk sign asks names the number - and offers to block every sender in the pile, listed by address, exactly as the bar over the list does (it used to offer only the sender of the conversation you were reading, and the rest kept getting through) - and the five seconds afterwards put the lot back. Handing a conversation to a colleague is deliberately still one at a time - whose desk something lands on is a decision per conversation.
The bar stays put while you scroll. Pick the top of a run, scroll down, shift-click the bottom of it, and the buttons are still there at the top of the list rather than a screen and a half above your head. The list slides underneath it.
The list grows, it does not turn pages. Twenty-five conversations to begin with, and at the foot of them a button saying how many more there are - press it and the next twenty-five appear underneath the ones already there, so you end up with fifty in one list rather than a second page with the first twenty-five gone. Whatever you had picked stays picked, and whatever you had open beside the list stays open. It stops at five hundred, at which point the list is telling you to search for the thing rather than scroll to it.
Snooze is the exception, and it is always there. It opens the same list of times the clock on a single conversation opens - later today, tomorrow morning, next week, or a day and time of your own off the calendar - with the day and hour each one lands on written beside it. Every one of the picked conversations goes to sleep until then and comes back at the top of the list. It stays on the bar even when they are all already asleep, because choosing a new time for something is a real change rather than writing down what was written already. Bringing a pile back early is Reopen beside it, not in here.
The junk sign asks before it does anything. It tells you the lot are about to go into a spam folder - yours, or a colleague's where the address is their own rather than the team's - and asks whether you would also like to turn those senders away in future, listing them so you can see who. No, just move them does the move and leaves the front door alone. Block them all does both. Cancel, the cross in the corner and Escape all leave every one of them exactly where it is, which is what you want the moment a finger lands on the wrong button in a toolbar. Blocking means nothing further from them reaches an inbox on this site, shared or personal - it goes straight in the spam folder instead, so you can still see what they sent. Nothing already here is touched, and you can let anybody back in from the Spam folder or the inbox settings. The blocking half only appears for somebody allowed to answer messages, since shutting the front door changes what everybody on the site receives - and where there is nobody in the pile to turn away, or you are not the one who may, it asks the shorter question instead: move them, or cancel.
Answering is on the message, not on the conversation. Every message has a reply arrow at the right-hand end of its own header, beside the name and the time, and the writing box opens under it when you press it. Beside the arrow are three dots holding the rarer three: Reply all, Forward and Mark as unread. This is worth the change on a long thread: a single Reply button above nine messages could never say which of them you meant, and now the one you are looking at is the one you are answering. Press the arrow again to put the box away. If you left a draft on that conversation it opens with the box already up, so nothing half-written is out of sight. (Starting one from scratch is Write a message, above.)
Internal notes have their own line at the bottom. One rounded box on a strip of frosted glass, always there as you scroll, edged in the same amber the notes themselves wear and saying on its face that nobody outside sees it. Type and press Enter, or the round arrow button beside it. Shift+Enter starts a new line, and the box grows with the note, up to about six lines, before it scrolls. A note is not sent to anybody and deliberately does not bump the conversation or mark it unread: us talking among ourselves should not look like the customer writing again. Type @ to ask a colleague. The names open in a short menu above the line, one under another; the arrow keys move through it and Return or Tab takes one. You can also simply write the name out - @Emma can you look at this - and pressing Enter asks Emma, whether or not you went near the menu. The only thing it will not do is guess: a half-written name, or a name two of you share, is left for the menu to settle. See Asking a colleague to look at something. For a longer note - something with a file on it - press a reply arrow and switch the box to Internal note, which still does everything it always did.
Assign sits on the subject's own line, between the alarm clock and the button saying where the conversation stands, and opens the list of colleagues to hand it to. Once it is with somebody it says so - With Marcus - and pressing it again hands it on or takes it back to nobody.
Type an @ in the note line, pick the colleague out of the names that appear under it - keep typing to narrow them, or use the arrow keys and Return - write the rest of the sentence, and leave the note. Name as many as you like. That is the whole of it, and it is what everybody was doing anyway before anything was listening.
The name has to be picked from the list rather than merely typed: a Sam Smith and a Sam Smyth are not something the site should be guessing between when the answer hands somebody a job. Delete a name back out of the sentence before you leave the note and that colleague is not told.
What it does is worth knowing, because it is more than a nudge.
It goes on their list, not yours. The conversation appears under Mentioned for everybody you named. Each of them can set it aside until a day and time that suits, mark it done, and open it again later - and that is theirs alone. It does not touch where the conversation itself stands, which is deliberate: the conversation has one status shared by everybody who can read it, so if marking your own bit done also marked the conversation done, three people asked about one order would close it from under each other.
Something set aside comes back on its own when the time you picked arrives, exactly as a snoozed conversation does.
It also lets them in. Whoever you tag can read that conversation whether or not the address it sits in has ever been shared with them - so you can ask the warehouse about an invoice in accounts@ without putting the warehouse on accounts@. Three things keep that narrow, and they are worth saying plainly:
- it is that one conversation, never the address it arrived at. Nothing else in that inbox opens;
- only somebody who could already open the conversation can hand it over, so nobody can let themselves into something by tagging a friend;
- it lets them read and work through it, not answer it. Sending as
accounts@still needs permission to send fromaccounts@.
They do need permission to use the inbox screen at all. Tagging somebody who has never been given it tells them nothing, because there is nowhere for it to land.
Asked twice is one job. If somebody tags you again about a conversation you already have - a fortnight later, chasing - the one you already have comes back to the top with the new note against it, rather than a second copy appearing beside the first. Something you had already marked done reopens, which is rather the point of asking again.
Every ask is written into the conversation's own log, naming who was asked, so "who did we hand this to" has an answer a fortnight later.
Where it stands is a box to tick. On the subject's own line, hard against the cross that shuts the conversation, there is a square. Empty means it is still open (or asleep); ticked means it is done. Press an empty one and it is marked done; press a ticked one and it opens again. Either way the Undo below appears, so a stray press costs nothing. Snoozing and waking belong to the clock beside it. Every change is recorded, so "who marked this done" has an answer.
The bin, the junk sign and the snooze clock share one glass capsule at the left of that row, with a hairline between each, the way Mail on an iPhone groups its own.
Whose it is sits beside it, as a capsule: a small figure while nobody has it, their name once somebody does, and an arrow on the right that opens the list of colleagues to hand it to.
Press an outsider's name or address to see everything they are in. On any message, the name and address it came from and every address on its To line can be pressed - but only for people outside the business: your own addresses and your colleagues' stay plain text. Pressing one opens All, every status, narrowed to each conversation that person is anywhere on - who wrote it, who it went to, or who was copied in - with a chip at the top (Everything with "sam@example.com") whose cross takes it off again. It is the exact address, so sam@example.com does not also sweep in bigsam@example.com, and the narrowing stays while you open and work through the conversations it found. The name under the subject works the same way.
The line under the subject says who it is with, by name or email address, where it used to say how they got in touch - "Phone", "Email". A conversation merged together out of emails, texts and calls is not a phone conversation, and the question anybody reading it has is who, not how.
Where the address book knows them, it is the name on their contact card whenever the message itself has only a number or an address - so a call from somebody you have saved says their name rather than 07700 900123. Press the name to open their contact record in a window over the conversation - everything the contacts page shows, with its own edit, merge and export buttons - and press the cross, Escape or the background to go back to the conversation exactly where you left it. The pencil beside the name changes who the conversation is with, choosing between everybody who appears in it, or back to whoever wrote last.
Anything that closes or reopens a conversation offers to take it back. Mark one done, open one again, snooze it, or junk it, and for five seconds afterwards a small Undo sits at the bottom of the screen. Pressing it puts the conversation back exactly where it stood - open if it was open, and still asleep until Monday if that is what it was. Leave it alone and it fades away by itself. It is there because all three of those take the conversation off the list you were looking at, and finding one again meant knowing that All exists and which of the forty rows in it was yours.
The same offer appears when you do any of it to a picked pile from the bar above the list: 6 conversations snoozed, with an Undo that puts all six back where each of them individually stood. Where you also turned the senders away, Undo brings the post back out of the spam folder and leaves the front door shut - it says so - because letting somebody back in is a decision about the whole site rather than about six messages, and it is made in the Spam folder or in the inbox settings.
And the conversation stops being open beside the list. Filing something, putting it to sleep or junking it takes it off the list you were reading, so the pane closes and the list takes the width again rather than leaving you looking at a conversation the list beside it has just dropped. Junk has always done this; done and snoozed used to sit there afterwards, which read as though the press had not worked. If the pile you acted on from the bar happened to include the conversation you had open, the same thing happens.
The alarm clock - first of the three, at the far left of them - is the reminder. Three ready-made answers - in three hours, tomorrow morning, next week - each with the day and the time it actually lands on written beside it, so a promise is not something you have to make and then undo to check. The last entry, Day & Time, opens a month you can walk back and forth through with a date and a time under it: the morning somebody said they would ring back, the day after a delivery is due. It will not accept a time that has already been, and the times it means are your site's, not your laptop's - so nine o'clock booked from a hotel in Spain still means nine o'clock at the office.
A reply cancels it, and the same goes for done. Setting a conversation to come back on Thursday is a bet that nothing will happen before Thursday. Marking one done is the same bet with no end date: nothing more will happen at all. Either way, somebody writing on it settles the bet, so the conversation goes straight back under Open, unread, where you would have seen it anyway.
That covers the customer answering, and it covers a colleague answering them from their own phone or from Outlook rather than from here. Either way somebody is dealing with it now, and it should not be hidden.
Done is the half that matters more, which is not obvious. A snoozed conversation comes back on its own on Thursday whatever happens. A finished one never does - and the number on your address tabs counts open conversations, so a finished one is not in it. So before this, a customer replying to something you had finished with landed at the top of the Done tab, with no number anywhere to tell you, and stayed there. It is now back in Open with the rest of your morning.
Answering one yourself does it too. Reply from here to something you had marked done and it goes back to Open, because writing to somebody is the plainest possible statement that you are not finished with them - and you are now waiting on an answer that will land somewhere you can see it. It used to stay under Done with your own reply sitting on it, on the reasoning that you knew what you had just done; what that actually meant was that their answer, when it came, was the first you heard of it.
A snooze you set yourself is left alone, though. Put a conversation to sleep until Thursday and then send something on it - now, or on a timer - and Thursday is still when you see it. That is the whole of Send later & snooze, and waking it up because of the message you yourself set going would undo half of what you asked for. Only done is reversed by your own reply. Somebody else writing on it still cancels both, as above.
Two things deliberately leave a conversation where it is whoever wrote them. An out-of-office and a bounce, because that is the mail system talking rather than a person - which is also what stops a mailing list nobody has unsubscribed from dragging a finished conversation back into Open every week. And an internal note, for the same reason it does not bump the conversation or mark it unread: us talking among ourselves is not the customer writing back.
When one comes back this way the conversation says so, in a line directly above the reply that did it - A reply arrived, so it is no longer snoozed, or A reply arrived, so it was opened again. Where it was your own reply that did it, the line carries your name: Marcus replied, so it was opened again. A conversation turning up in Open on Tuesday when you asked for Thursday, or one you were sure you had finished with, is explained rather than mysterious. See What happened to it, in the conversation.
Everything done to a conversation is written into the conversation itself, as a small line ruled off between the messages it happened between, with the time on the end. It used to be a folded-away list at the very foot of the conversation called What has been done to this, which nobody opened; that list is gone and every entry in it is now one of these lines.
What gets a line:
- Snoozed - Chris snoozed it until Fri, 2 October 2026 at 09:00, or moved its snooze to... when it was already asleep.
- Back from a snooze - Its snooze ran out, so it came back, stamped with the time it was due rather than whenever somebody next looked; Nobody replied, so it came back to Sam when it was a follow-up; Chris brought it back before its snooze ran out when somebody woke it by hand.
- A reply waking it or reopening it - A reply arrived, so it is no longer snoozed, ...so it was opened again, or A reply was sent from outside the inbox... when a colleague answered from their phone.
- Done and opened again - Sam marked it done, Chris opened it again.
- The bin - Chris put it in their bin, ...took it back out of their bin, and A reply arrived, so it came back out of the bin. When somebody is covering a colleague's own address, the line says whose: Chris put it in Sam's bin.
- Junk - Sam moved it to their junk, ...took it back out of their junk, and It came from a sender the site has blocked, so it went straight to junk.
- Scheduled messages - cancelled because somebody replied first (see When somebody replies first), failed when their time came, or gone out with a follow-up or a snooze set on them.
- Everything the old list had - handed to somebody, moved to another mailbox, merged (with Unmerge on the end while it can still be undone), a message deleted or moved out, a record linked, a colleague asked to look.
Where a line goes. A line caused by a message - a reply waking a snooze, a scheduled reply cancelled because of it - sits directly above that message. Everything else sits after whatever was already in the conversation when it happened. Reading newest first, the lines turn round with the messages, so each one still sits between the same two.
Who sees them. Everybody who can read the conversation - including bin and junk lines, which say whose bin or junk it was so nobody takes it to mean it went for everybody. The exception is an internal discussion, where done and the bin are each person's own: those lines are shown only to whoever pressed the button.
Mis-presses tidy themselves up. Put something in the bin and press Undo, and the line goes rather than being joined by its opposite - the same for junk, snooze and done, if it is put back as it was by the same person within a minute and a half and nothing else has happened to the conversation since. Pressing a button that changes nothing writes nothing.
Notes are not repeated as a line - the note is already there in the conversation.
There are two search boxes, and they are for two different jobs.
The one in the head of the list narrows what is in front of you. Type into it and the list you are looking at gets shorter. It is the right thing when you know roughly where the message is.
The magnifier at the top of the rail, beside the pen, is for when you do not. It opens over the whole inbox, and it starts by looking everywhere you can read, at any status - so something dealt with and filed three weeks ago in an address you never open is found rather than missed. Open it again while you are looking at the results and it keeps whatever you chose last time, so narrowing a search down takes one press rather than three. Ctrl K opens it from anywhere in the inbox, or Cmd K on a Mac.
Under the box are the narrower cuts, which are the ones a mail program has always had:
| Cut | What it does |
|---|---|
| Where to look | Everywhere you can see, one address, one channel, or the mail that landed nowhere. |
| Where it stands | Any status, or only the open, set-aside or dealt-with ones. |
| From | Part of a name or an address the message came from. |
| To | Anybody it was sent to - or copied to, which is usually the point. |
| Subject | Words in the subject line. |
| Since and Up to and including | Two dates. Both days count, in your site's own time. |
| Has something attached | Only the conversations carrying a file. |
| Nobody has read it | Only the unread ones. |
You can also search Contacts from the same box, which looks through the address book instead. That one takes the words and nothing else - who a message came from, and when, are questions about post.
Press Search and it lands on a screen of its own. The same box sits across the top of the results and whatever you open from them, with all of the cuts above laid out under it, so what you asked for is in front of you rather than behind a magnifier you have to open again to remember it by. Change any of them there: the menus and the tickboxes take effect as you pick them, and words wait for Enter. The button beside the box turns the results round, newest first or oldest first, and the cross at the end leaves the search and puts you back on the list with nothing narrowing it.
While a search is up, the list underneath is the results and nothing else - the ordinary search box and the Open, Snoozed, Done and All tabs come off, because everything they do is in the head above. Anything the search was handed to a particular colleague by is let go of as well: Assigned to me is a place in the rail rather than a cut a search can make, and one quietly still on would be narrowing your results with nothing saying so.
The address in your browser holds the lot, so a search can be sent to a colleague and the back button behaves - including back out of the search itself.
Search covers subjects, senders and the text of messages, and it only ever searches the inboxes you are allowed to read.
It is forgiving. Half a word will do - invoi finds invoice - and so will a word spelt wrong: recieve finds receive, by swapping a word your mail never contains for the few closest words it does. A word that is spelt right is taken as meant, so order does not start finding border. Numbers and references are never guessed at - PO-00025 finds PO-00025 and not its neighbour - though any part of a conversation's subject counts, so 0025 finds it too. Put a phrase in quotes, a minus before a word or OR between two, and the search goes back to being exact.
Mail turns up at the wrong address all the time - a delivery question sent to sales@, an invoice query sent to whoever answered last. Drag the conversation from the list and drop it on the mailbox it belongs in. The mailbox lights up as you hover over it; the one it is already in does not, and will not take the drop.
- Tick several first and drag any one of them, and the whole pile goes. Drag a row that is not ticked and only that row moves, whatever else is ticked - the same rule as files in a folder.
- Replies then go from the new mailbox. Its address, its name, its signature, its way of sending. That is rather the point: a customer who wrote to sales@ about a late delivery gets their answer from support@, and their next reply lands in support@ with the rest of the conversation.
- Half-written replies come too. A draft on the conversation moves with it, and a reply you had scheduled goes out from the new address rather than the one the conversation has just left.
- What was already sent stays as it was sent. The Sent folder under the old mailbox still shows what left from it. That is history, and moving a conversation does not rewrite it.
- It stays moved. The next collection does not put it back, even when more mail for it arrives at the old address.
- A five-second Undo appears, as it does for everything else on this screen, and the move is written into the conversation's history with where it came from and where it went.
You can move a conversation you can open into any mailbox you can see. Somebody else's own inbox is not on your rail unless they have given it to you, so you cannot drop mail into it. A discussion cannot be moved: it belongs to the people in it, not to an address. A conversation that has been merged into another moves with the one it was merged into.
One thing worth knowing: a brand-new email that arrives at the old address and does not say which conversation it is answering starts a new conversation there. It is, after all, new mail to that address. Anything that is genuinely a reply finds the moved conversation wherever it now lives.
Sometimes one conversation arrives as two. Somebody writes from their work address on Monday and their phone on Thursday. A supplier changes the subject line and their mail program forgets what it was replying to. A colleague forwards something in and now there are three of it. And an email between two of your own addresses is deliberately kept as two conversations - one for each of you, each with its own unread mark and its own snooze - which is right until the day you decide that this particular back and forth is one story with four people in it.
Pick the conversations and press Merge. Hold cmd (ctrl on Windows) and click to pick more than one, or click the first and shift-click the last, then press Merge in the bar that appears at the top of the list. It is there for anybody who can manage inboxes.
The conversation you already have open counts as one of them. Reading something and ctrl-clicking a second conversation gives you two picked, not one, which is what the highlight on the open row has always suggested - so the pair you are looking at is a merge without any further ceremony. Ctrl-click the open one again if you did not mean to include it.
Everything folds into the conversation that started it - the oldest of the ones you picked. It keeps that one's subject and its place in the list, and every message from the others joins it in date order, so the merged conversation reads as the one story it always was.
You are asked who it is with. Where the conversations being merged are with different-looking people - an email from Sam Jones and a phone call from 07700 900123, say - the confirmation lists them and you pick the one the merged conversation should be known by. That name is what the list, the line under the subject and new-mail notices show from then on, however the next message arrives; without it, whichever came in last decided, so a customer's name turned into a phone number every time they rang. Undoing the merge that brought that person in puts back whoever it said before.
A few things worth knowing before you press it:
- Nothing is thrown away, and you can put it back. Open the merged conversation and find the line recording the merge among the messages - Chris merged it with another, with the date - which has Unmerge on the end of it. Anything that arrived after the merge stays where it is, because it arrived on the merged conversation rather than on the old one. It used to be a panel across the top of every merged conversation for ever, saying the same thing to everybody who opened it.
- A merged conversation is still unanswered if any half of it was. Merging something you had marked done into something you had not does not quietly mark the lot done, and an unread half stays unread. Better to be shown something twice than to lose it off the list.
- Merging across two of your addresses lets both of them read it. This is the point of merging marcus@ and hi@ together, and it is also the thing to think about first: everybody who can read either address can then read the whole of the merged conversation, including the half that arrived on the other one. The confirmation says so, and names the addresses, before anything happens. See Who can see what.
- It shows in both addresses' tabs afterwards, with the second address named in the line under the subject, and it stays merged when somebody replies. Without that last part the two sides would come apart again on the very next email.
- Merging something out of your own inbox with something in a shared one moves it to the shared one. A supplier writes to chris@ first and to purchasing@ second; merge the two and the conversation now lives in Purchasing, so the team can see it and the next reply goes from purchasing@ rather than from one person's address. It used to stay wherever the older half was, which was very often somebody's own post. It is still listed in your own inbox as well, so nothing vanishes; the message that confirms the merge says where it has gone; and undoing the merge puts it back where it lived. Two shared mailboxes have no such pecking order - there the older conversation keeps its place, as before.
- You cannot merge into something that has itself been merged away. Undo that one first, or merge into the conversation it became.
Merging conversations and merging people are two different jobs. If the same human has turned up twice in the address book, that is a merge on their card - see People, and how conversations collapse.
Merging puts two conversations together. These two do the opposite, one message at a time, and they are the answer to the same problem: working out which conversation an email belongs to is guesswork, and now and then the guess is wrong. Two customers both write "Re: Quote", a mail program forgets what it was replying to, and a stray message ends up glued onto somebody else's thread.
Both live behind the dots beside a message, under a rule, below the everyday entries. Both ask before they do anything. Neither appears at all unless you can manage inboxes - the same permission merging takes, and for the same reason: these change what everybody on the address sees, rather than what your own screen shows.
Move to its own conversation. The message leaves this conversation and starts a fresh one of its own, with anything attached to it. The rest of the conversation stays exactly as it was, and nothing is sent to anybody. You land on the new conversation as soon as it is made.
- It keeps the address it was in, so it is still in the tab you found it in.
- It arrives without a name on it. Whoever it is with gets worked out from the message itself on the next check, rather than inheriting the name it was wrongly filed under - which is usually the reason for moving it in the first place.
- The way back is Merge. If it turns out you were wrong, pick the two conversations and merge them together again. There is no separate undo, because the one that exists already does exactly this job.
- The only message in a conversation cannot be moved: it is already a conversation of its own.
- A message from one of the other channels - a chat, an enquiry, a call, a text - cannot be moved. The part of the site that owns it decides what is in its conversations, and it would be handed straight back on the next collection.
Delete this message. One message goes, along with anything attached to it, and there is no getting it back. The rest of the conversation stays.
- Nothing is deleted from your mail account. The email is still sitting in whatever mailbox it arrived in; this is a fact about what the site holds. Same for the other channels: the part of the site that owns a chat still has every word of it.
- It does not come back. The site writes down that you threw it away, so the next check walks past it rather than collecting it again - which is what would otherwise happen the first time somebody filed that email into another folder from their phone.
- It goes for everybody who can see the address, which is what makes it different from the waste basket at the top of a conversation. That one is yours alone and puts back with the same press - see The bin.
- Both are written into the conversation as a line where it happened, so there is a record that something left. What was in the deleted message is deliberately not kept there: a deletion that quietly filed away the subject and the sender would not be much of a deletion.
- On a voicemail from one of the calling channels you will see a Delete button in the foot of the message instead. That one is a bigger thing - it deletes the recording at the phone company as well - so the two are never offered side by side.
- Any recording on a call or voicemail - where the channel supports it - has its own buttons under the player: Delete from Twilio (a copy is saved on the site first, so it still plays here) and then Delete saved copy. They act on that one recording and leave the message where it is. A line under the player says where the recording is kept, and once it has gone from everywhere the player is taken away. See the Twilio page, Taking one recording off Twilio.
Every conversation has a no-entry sign at the top of it, between the waste basket and the little alarm clock. It is deliberately not a basket: a basket means Delete everywhere else in the world, and the basket beside it means exactly that (see The bin). This one means refused, which is what pressing it does - press it and you are asked before anything moves; answer, and the conversation goes into a Spam folder, whose depending on where it came from.
On a shared address, it goes in yours. Nobody else's screen changes at all. What one person files as junk is the supplier newsletter a colleague reads every Tuesday, and on an address like sales@ there are several people with several opinions - so the button records yours and leaves everybody else's list exactly as it was.
On a colleague's own address, it goes in theirs. If you have been let in to somebody's own inbox to cover their post while they are away, junk you clear out of it lands in their spam folder rather than yours. That is the honest answer twice over: it is their post, and your own spam folder has no business filling up with a fortnight of somebody else's rubbish. The button says so before you press it, and the question that follows says it again - "it will go into Sam's spam folder".
Marking something as junk marks it read. Same reason as the bin: it has been looked at and dealt with, so it stops counting as new, and "Not junk" brings it back read. On a shared address that shows as read for your colleagues as well, just as opening it would. Post the site turns away by itself because the sender is blocked is the one exception - that arrives unread in the spam folder, because nobody has looked at it yet and the folder's number is how you find out it happened.
It also leaves your view of their inbox, which is rather the point of covering somebody: you are trying to see what they would see, and post they have already binned is post you should not be working through.
You can look in their spam folder, and take something back out of it, under their name on the rail - Spam sits there beside their Sent and Mentioned. So a mis-click while covering is a mis-click you can fix.
Nothing is deleted. The conversation keeps every message in it. Open the spam folder, press the same button again, and it goes straight back where it was.
The press asks before it moves anything. Where there is a sender who could be turned away, pressing the sign puts up a second question - would you like to block them as well? - and nothing has happened yet when it appears. There are three answers:
- No, just move it is the ordinary one. The conversation goes into the spam folder and the front door stays open.
-
Block them does the move and shuts the site's front door on that address. Nothing further from them reaches any inbox on this site - the shared ones, every colleague's own, and the ones nobody has opened in months. It is one list for the whole site, because a block that only covered the address you happened to be standing in is no block at all: they would simply write to
hello@instead and turn up in somebody else's list an hour later. - Cancel, the cross in the top corner, Escape and a click on the background all mean the same thing: you did not mean to press it. Nothing moves and nobody is blocked.
The two decisions are deliberately kept apart. Throwing one message away is a small decision about your own screen; refusing somebody in future is a decision about the whole business, and it should not happen by accident on the way past. The cross is there for the same reason in the other direction - the junk button sits next to the buttons you press all day, and a mis-press should cost you one click rather than a hunt through the spam folder.
On a channel that can refuse people itself - the phone, say - the same question is asked about them rather than about an address. Would you also like to stop them getting through on Phone in future? Block them does the junk and the refusing; No, just move it does the junk alone. It is the same decision in the same place, which is the point: refusing a caller happens for the same reason refusing an address does, and it should be asked after the junk has been dealt with rather than offered as a button somebody can lean on.
Where there is nobody to block - a conversation between colleagues, a caller who withheld their number, a sender already turned away, or you not being allowed to shut the front door - the question is a shorter one: Move it to the spam folder?, with Move it to spam and Cancel. The way out is there either way, because that is the half of it a mis-press needs.
Straight into the Spam folder, and nowhere else. Post from a blocked address is still collected - it is just dropped in the spam folder the moment it arrives, marked as dealt with so it is out of everybody's way, and left unread so the folder can tell you how much of it there is. It never appears in an inbox, never adds to an address's number, never nudges anybody's browser and is never handed to a colleague as a job.
It is in the spam folder for everybody, which is the one thing on this screen that is not a matter of opinion. Junk you mark yourself is your own view of one conversation; a block is a decision about the site, so the post it turns away is out of everybody's lists and in everybody's Spam folder.
Why collect it at all? Because "did they ever actually write?" is a question somebody asks eventually, usually the week a customer rings up cross about being ignored - and until now the only way to answer it was to go and log into the mailbox yourself. A nuisance is a nuisance either way and you never have to look; somebody blocked in a temper turns out to have written four times, and now you can see that.
Blocking never deletes a conversation. Their old conversations stay exactly where they are - often that history is the whole reason you want them stopped.
It does take them out of your contacts. Every address that writes in becomes a contact, and a spammer you have turned away has no business sitting in the contacts list, the people search or a campaign audience looking exactly like a customer. So blocking an address removes the contact that is nothing but that address (and anything merged into it), and no new one is made while they stay blocked. A contact who can still be reached some other way - a phone number, or a second address nobody has blocked - is kept, and only loses the blocked address. Blocking ada+offers@example.com touches that address only, never plain ada@example.com. Unblock somebody and their next email makes them a contact again, from scratch.
If one of them turns out to be real, open it in the Spam folder and press the same junk button at the top of it. That takes the conversation out of the spam folder for everybody and puts it back in the ordinary lists. The sender stays blocked: letting one conversation through and opening the front door again are two different decisions, and they are two different buttons.
There are two ways to the list, and it is the same list either way:
- The Spam folder itself. In the head of the list, between the search box and the filters, there is a button that opens Blocked addresses. Newest block first, since the one you are hunting for is nearly always the one somebody made this week.
- Settings → Unified Inbox → Collecting, under Blocked senders - the way to it that does not need the inbox open at all.
Both list every address the site refuses, who blocked it and when. Press Let them through and their post arrives normally again from the next round onwards. What was collected while the door was shut stays in the Spam folder until somebody takes it out, one conversation at a time.
You need permission to reply in order to block somebody or to let them back in, since it changes what everybody receives. Seeing the list needs the same permission as the rest of the inbox settings, so the button in the Spam folder only appears for somebody who has that. Marking something as junk yourself only needs permission to read, because it only changes your own screen.
Beside the junk sign at the top of every conversation is a waste basket, and it means what a waste basket means everywhere else: get rid of this one. Pick several in the list and there is a Delete button on the bar above them that does the same to the lot.
They are two different jobs, which is why there are two buttons. Junk says "do not show me this sort of post" and is really about the sender; the bin says "I have finished with this particular conversation" and is about the conversation. With only one of the two on the screen people reach for whichever is there, and a spam folder slowly fills with post that was never spam - at which point it is no use for the one job it does have.
A deletion is yours, exactly as junk is. What you delete leaves your lists and changes nothing at all on a colleague's screen. On a shared address like sales@ it goes into your bin, because the address belongs to the team and the decision is yours alone.
On a colleague's own address it goes into theirs. Covering somebody's post while they are away means clearing their inbox on their behalf, so what you delete out of it lands in their bin - not yours, which has no business filling up with a fortnight of somebody else's post. The button says so before you press it. Their bin sits under their name on the rail, beside their Sent, Mentioned and Spam, so a mis-click while covering is a mis-click you can fix.
Deleting something marks it read. Something you have binned is something you have dealt with, so it stops counting as new: it leaves the address's unread number, and if you put it back later it comes back read rather than bold. Read is one mark on the conversation rather than one each, so on a shared address it reads as read for your colleagues too - the same as it would had you simply opened it.
Shared addresses have no bin of their own, for the same reason they have no spam folder: a bin belongs to a person, and a shared address is nobody in particular.
Nothing is destroyed by being deleted. The conversation keeps every message in it and sits in the bin until somebody empties it. Open the Bin, press the same basket again, and it goes straight back where it was - or pick a pile and press Put back, which is what the Delete button turns into inside the folder. You also get five seconds of Undo on the press itself, in the corner of the screen.
Nothing empties itself. There is no thirty-day timer and no nightly sweep. A conversation that vanishes on a date nobody chose is exactly what makes people stop trusting a bin, so this one holds what you put in it until somebody presses the button.
A reply brings it back out. Deleting a conversation is a decision about what had been said when you pressed the basket, not about whatever comes next. When somebody writes on a conversation that is sitting in a bin, it comes back out of every bin it was in and lands on the lists unread, the same way a reply reopens something marked done, and its history says why. Otherwise the new message would sit unread in a folder nobody opens, waiting for the next Empty bin to throw it away before anybody had read it. Out-of-office replies, bounces and post from a sender you have turned away leave it where it is.
In the head of the Bin, between the search box and the filters, there is a basket with its lid off. That is the one press on this screen that genuinely throws something away, and it asks first - it tells you how many conversations are about to go, that they go for everybody who could see them rather than only off your own screen, and that there is no way back. The keyboard starts on Cancel.
It takes the permission that looks after the site rather than the one that reads it, so the button only appears for somebody who has that. Putting something in a bin hides it from one screen and undoes itself with the same press; emptying one takes the conversation away from every colleague who could read it, and those are not the same act.
Nothing is ever deleted from your mail account. Whatever arrived by email is still sitting in the mailbox it arrived in, untouched - this site has never deleted a message from anybody's mailbox and does not start here. Emptying the bin clears what is held on this site: the conversations, and any files attached to them.
And an emptied bin stays empty. A live chat, an enquiry, a call or a text is a copy of something the part of the site that owns it still holds, and it hands that over again whenever it is asked - so what you threw away is written down as thrown away, and the next check walks past it instead of fetching it back. If somebody carries on typing in a chat you had deleted, the conversation comes back with what they have just said and none of the history you got rid of.
Newest at the top, as you would expect. What counts as newest is the later of when the message was written and when it reached you - and those are not always the same day.
Move an email into one of the folders the site watches and it is dated whenever the sender wrote it, which may have been last Tuesday. It is collected on the next check, lands in the right conversation, and would otherwise sit exactly where last Tuesday puts it - halfway down the list, under a subject line about something else, with nothing on the screen moving. Conversations that get post this way come to the top instead, where you can see them.
Nothing else moves. A conversation that has never had anything filed into it by hand sits where the date on its newest message puts it, exactly as it always did, and once a genuinely newer message arrives the conversation goes back to being ordered on that.
By default a conversation reads the way it happened: the first message at the top, the latest at the bottom, and the writing box - once you have asked for it - under the lot.
Reading that way round, opening a conversation takes you straight to the part worth reading rather than to the beginning of it: the message that came in while it was still unread, or - if you had already read the lot - whatever was said last, an internal note included. A thread with forty messages in it therefore opens on the one that has just arrived, not four thousand pixels above it. Scroll and it leaves you where you put it.
Tick Show the newest message at the top of a conversation on Settings → Unified Inbox → Collecting and it turns round - the latest message is the first thing you see when you open one, and the writing box moves up with it so the reply sits beside the thing you are replying to. Handy on a long back and forth where the only part anybody needs is the end of it.
It is a site setting rather than a personal one, so everybody reads the same way round.
Most sites end up with one inbox they live in and two or three they glance at, and alphabetical order has no opinion about which is which. Drag a row up or down and it stays there. It saves as you drop it - there is nothing to press afterwards.
All four groups can be rearranged: Yours, Shared inboxes, Team inboxes and Channels. Each keeps to itself, because an order only means anything against the other things in the same list - a live chat sitting in the middle of your email addresses would be nobody's idea of tidy.
Nothing announces any of this while you are simply running your eye down the rail. It is a list of places to go and it behaves like one; take hold of a row and the pointer closes into a hand, which is the moment it starts mattering.
If you would rather not use a mouse for it, put the keyboard focus on a row, hold Alt and press the up or down arrow keys.
Yours is yours, and the other three are the site's. The order of Yours - your own address, All, Mentioned, Drafts, Scheduled, Sent, and Spam and Bin behind More - is saved against you and seen by nobody else, so it needs no permission at all: somebody who lives in Sent can put it second, and everybody else's rail carries on exactly as it was. The other three are one arrangement that everybody who opens the inbox sees, which is why only people who can manage inboxes can change them. If you can read the inbox but not manage it, those three sit where whoever looks after the site has put them and do not drag; Yours still does.
One quiet detail about the channels: you can only rearrange the ones you can see, and what somebody may see is decided by the module each channel belongs to. Move the two you have and the ones you do not stay exactly where they were for the colleagues who do.
Most people on a site live in one address. Somebody who does purchasing wants purchasing@ in front of them the moment they open the inbox, not a list of everything the company has ever been sent.
This is about a shared address you want somebody to open on. An individual inbox is already theirs and does all of the below on its own - see "Shared and individual inboxes" above.
Settings → Unified Inbox → Inboxes → Edit an inbox → Who can read it. Beside each name there is Their own inbox. Tick it and, for that person only:
- the address sits at the top of the rail under Yours, ahead of All, and comes out of Team inboxes;
- it is what they land on when they open the inbox, instead of All;
- everything handed to them shows in it, wherever it was filed - see "The screen" above;
- its signature goes at the foot of their replies, whichever address they are answering from.
All does not go anywhere - it moves down one, to second - and the rest of the addresses follow in the usual order. Everybody else's rail is exactly as it was: this is one person's arrangement, not the site's, so it does not shuffle the rail for anybody else and it does not affect the order you have dragged the addresses into.
One address each. Ticking the box on a second inbox moves the person there rather than giving them two; the screen says whose address it currently is before you do it. Untick it and they go back to opening on All, which is where everybody starts.
You can only give somebody an address they can actually read, so if the inbox has a guest list, they need to be on it. Take them off it later and they quietly go back to All rather than landing on a tab that will not open.
A message someone else wrote is shown inside its own sealed frame, on a light background in both light and dark mode. It takes up as much room as the message actually needs - a two-line reply is two lines, a long newsletter is as long as it is, and neither has a scrollbar of its own inside the page's. Messages that kept one anyway were losing a fraction of a pixel in the measuring: a message a hair over five hundred pixels tall was given exactly five hundred, and the hair left over was a scrollbar over the whole thing. The frame is now measured to the fraction, rounded up rather than down, and given a couple of pixels of room to spare. A message written wider than the column you are reading it in is shrunk until the whole of it fits, the way a phone shows a desktop-width email, rather than being given a scrollbar along the bottom that hides half of it. The whole message is on the page and the page scrolls it, which is rather the point. A message that will not say how big it is gets a generous height rather than the letterbox it used to get. The sender chose their colours assuming a white page, and repainting the background dark while leaving their text alone is how a message ends up black on black. The rest of the screen follows your theme as usual. Marketing email is nearly always built out of tables, and a table gives each column the width its contents insist on. The frame used to tell every message that a word could break anywhere it liked, which quietly told those columns that nothing insisted on anything - so in Safari a button could come out at a quarter of its proper width with its own label running off the side in white on a white page, reading "View" where it should have said "View in WhatsApp Manager". Nothing looked broken, it just said the wrong thing. That is sorted, and a long unbroken web address still wraps rather than sticking out. Sites sat behind a content network need one extra thing from it: Cloudflare will happily rewrite the frame on its way out, swapping every address in a message for the words "[email protected]" and shuffling the frame's own script into something no browser will run, which left messages stuck at their opening size with the layout cut about. The frame now tells Cloudflare to leave it alone, and it clips nothing at all until it has actually scaled a message to fit - so if anything ever does stop that script, the message spills over and can still be read rather than quietly losing its right-hand side.
Pictures hosted elsewhere are not loaded until you ask. Press Show pictures and they are fetched by the site rather than by your browser, so a marketing email learns nothing about you, your location or when you opened it. Links open in a new tab.
Pictures that came inside the message are simply shown. A signature logo written in Outlook, a screenshot pasted into the middle of a sentence, the artwork on a quote: none of those are kept on a website anywhere, they arrive as part of the message itself. They now appear as the sender meant them to, with no button to press - nothing is fetched from anybody's server to show one, so there is nothing to be careful about. They used to come out as empty boxes, under a note cheerfully explaining that anything carried inside the message was already there. It was not.
Post that arrived before this was fixed is included as far as it can be. Older messages were filed without the name the picture is referred to by, so those are matched on the filename instead - which is what Outlook and Apple Mail use anyway, so most of them come back. A few will keep their empty space, and nothing short of collecting the message again would fill it.
Your own post is never held back. Anything you sent, an internal note, and anything arriving from one of your own addresses, a colleague, or your own domain opens with its pictures already there. The warning is about what a stranger learns when your browser fetches their tracking pixel, and there is no stranger in your own signature - being asked about it on every reply you ever sent only taught people to press the button without reading it.
Reading your own Sent post no longer counts as the customer reading it. Brevo adds an invisible counter to every message it sends for you, and the copy that comes back into your Sent folder still has it in. Because your own post opens with its pictures already showing, glancing at what you sent on Tuesday used to fetch that counter - and Opened 11:32 would appear under your own message, with the open count in Brevo's reports going up to match. The customer had not touched it. You had. That one picture is now left alone, wherever it turns up and whoever asks: it is a transparent dot nobody can see, so nothing on the screen changes and the figures are finally about the person you sent it to.
Once you have pressed it for a message, that message keeps its pictures. Go off to something else, come back to it a fortnight later, and it opens with them already there rather than asking again - the sender found out the first time, and there is nothing left to protect by making you click twice. It is remembered by the browser you pressed it in, so a different computer starts the message the careful way round. The five hundred most recent are held; older ones quietly drop off the end.
Attachments are fetched when somebody opens one, not while collecting. Once fetched they go into your media library like any other file, filed under whoever the message was with: Inbox, then their email address, then Received or Sent. So one supplier's whole paper trail sits in one folder, and you can find last March's invoice from the media page without going anywhere near the conversation it came on. The one exception: when another part of the site is listening for post and has asked for a particular kind of file - purchasing asking for PDFs, say (see Other parts of the site hearing about the post) - files of that kind on post arriving from outside go into the library as the message is collected, so it has something to read. Only PDFs and ordinary pictures can ever be stored this way, each checked by what is actually inside it rather than what the sender called it, and never more than 30MB per email. Web pages, SVG drawings and programs are never stored without somebody opening them, whatever anybody asks for. Set Attachments to never fetch them and nothing is stored this way either.
Worth knowing before you rely on it: anything filed there can be seen by everybody on your site who can use the media library. Attachments used to be kept where only the inbox could reach them, precisely so that an invoice out of accounts@ did not turn up in front of everybody who can edit a page. Putting them on the media page gives that up, on purpose. If that is not what you want for a particular address, the people who can open that inbox is still the setting that matters, but the files themselves are now in the open.
Two sorts of file stay where they were, because both would bury everything else: pictures that are part of the message itself (a signature logo, a screenshot pasted into the middle of a sentence) and anything on an internal note, which has nobody to file it under. Those are still readable on the conversation, which is where anybody looking for them would look first.
Older files sort themselves out. Anything collected before this was how it worked was tucked away where the media page could not see it. The nightly tidy moves it across a few at a time, so a busy account catches up over a handful of nights and a quiet one is done by the morning. There is nothing to press and nothing to switch on, which is rather the point: "your files are in the wrong folder, shall I move them?" is not a question anybody should have to answer about their own site.
Deleting a message deletes its files. Throwing away a message, emptying the bin, erasing somebody's data, and the keeping window expiring all take the attachment out of storage and off the media page with it. The one exception is a file you attached from your media library: that one is yours, it belongs to the library rather than to the email, and sending it somewhere never puts it at risk.
There is a setting called Fetch everything as it arrives. It currently behaves the same as fetching one when somebody opens it. Pulling every attachment on a busy account through the hourly check needs a budget and a storage conversation of its own, so it is honestly a setting that does not do anything yet rather than one that does it badly.
Install the live chat (or Chatwoot live chat), contact form or Twilio modules alongside this one and their conversations appear here too, under Other channels.
One module can bring more than one channel. Twilio brings two: Phone for calls, voicemail and texts, and WhatsApp on its own. They are separate because they are separate things to answer - WhatsApp will only carry an ordinary message for 24 hours after somebody last wrote, and the phone has no such rule - and each has its own switch under Channels.
A conversation that changes after it arrives. Most messages are said once and never altered. A few are not: a voicemail, or a recorded call on a Twilio number set to type up its calls, reaches the hub as a message with a Listen button and no words, because the writing out takes a few minutes and the check runs on the hour. When the words do arrive they are added to that same message rather than filed as a new one, and the conversation stays where it is in your list - nothing new was said, it merely became readable. So a call you looked at this morning may have words under it this afternoon. If Twilio cannot type it up, the audio remains and the Twilio call log shows the failed attempt.
When Unified Inbox is installed, this is where those messages are answered. The contact form's own inbox tab and the live chat's own inbox tab stand down, because having two places to answer the same enquiry is how one of them stops being read. Each module keeps everything else it does: the chat widget, the form block, the phone numbers, its own settings. Uninstall Unified Inbox and their tabs come straight back, with nothing lost.
A colleague who is allowed to see the contact form but not this hub keeps their own tab. Nobody is locked out of their own messages.
A contact form can name the inbox its enquiries belong in. Edit the page, open the Contact Form block, and pick one under Deliver enquiries to.
From then on an enquiry from that form is ordinary post in that inbox. It sits in the list with the emails, it counts towards that inbox's number, and whoever is allowed to read that inbox is allowed to read it - which is worth reading twice, because it is the whole rule. Point a form at an inbox only two people can open, and only those two see what comes in on it.
It also stops appearing under Other channels, because it is in an inbox now and one thing should be in one place. Enquiries from forms that name no inbox stay where they always were.
Whichever inbox it lands in, the conversation says which form it came from - the form's own title, on the line under the buttons where attached orders and quotes sit. It is not something anybody attached, so it does not open anything and it cannot be taken off: it is what the conversation is.
Once you have read one, it stays read. Reading an enquiry here now tells the contact form so, which is what stops the next collection finding an enquiry the form still thinks nobody has looked at and turning it bold again. If a channel cannot be told - or will not listen - it no longer matters: nothing goes back to unread unless a genuinely newer message has arrived on it. Somebody writing again still does exactly what it always did.
Moving one afterwards is allowed and it stays moved. Nothing drags it back to where the form says it should be.
Settings → Unified Inbox → Channels lists the channels this site has, with a switch each. Turn one off and it stops appearing down the left, stops being counted and stops turning up in All.
It is worth having once your forms are pointed at real inboxes: a Contact form heading listing the same enquiries a second time is a second place to look for the same thing. Nothing is thrown away - the conversations are still collected, so switching it back on brings back everything that arrived in the meantime.
The one catch, and it is on the screen too: an enquiry from a form that names no inbox has nowhere else to be seen, so it will not be shown at all while its channel is off.
An enquiry that does name an inbox carries on as normal in that inbox, and you answer it there exactly as you would with the channel switched on - the reply still goes back out through the form, with your signature on it. Switching a channel off tidies the left-hand side; it never takes the Reply button off a conversation you can still see.
Three things worth knowing:
- A reply on one of these goes back the way it came, so the writing box asks nothing about where it is going: instead of the To, Cc, Bcc and Subject lines an email opens with, one line above the words says who is about to receive it and on which channel. There is nothing to type there because there is nothing to choose - a WhatsApp message goes back to the number it came from, an enquiry to the address on the enquiry. The box also drops what the channel cannot carry: no paperclip, and no Forward, because there is nowhere to forward one to. Email is untouched and keeps the whole box.
- A contact form enquiry is the exception, because it is answered by email. Its reply box is a full email box apart from the address: bold, links and lists, the paperclip, the catalogue as a proper table, and a Subject link so you can change what the reply is titled instead of always getting "Re:" and the enquiry's own subject. Forward is there too, and opens the same box with To, Cc and Bcc; it sends from the address the enquiry is filed in, or your own default address when it is not filed anywhere. The reply itself still goes out through the contact form, so the site's signature and email design still apply.
- You still get the formatting the channel actually has. WhatsApp has bold, italic and strikethrough - it writes them with a marker either side of the words rather than with a typeface - so a WhatsApp reply gets those three of the seven, and what you type in bold arrives in bold. What it does not get is the four that mean nothing at the far end: colour, a link with its own wording, and the two kinds of list. A channel with no formatting of its own, like a text message, gets no buttons at all rather than a row of them whose work would be thrown away between here and the customer.
- Replying to a chat uses your own chat account. If you have not connected yours yet, the reply is refused with a line telling you so and where to fix it, rather than going out under somebody else's name.
- A phone conversation is one outside number, not one call. Every call, voicemail and text with that number is one story, which is the whole point. Calls that were forwarded to somebody's mobile, and calls placed with Make a call, count as one call apiece and belong to the customer - your own mobile never turns up in the list as though it were a customer of yours.
- Text messages appear when the hourly check runs, not the instant they arrive. There is no live feed of incoming texts in this version.
- A voicemail message can be thrown away from the conversation it sits in: a Delete button under the message, which asks first and then removes it from the phone system as well as from here. Only the messages people leave - the log of a call, and a text, are kept by the phone system itself and this will say so rather than pretending.
- A WhatsApp reply has a clock on it. WhatsApp only lets a business write freely for 24 hours after the customer's last message; after that Meta takes nothing but wording it has approved in advance. Reply here after the window has closed and you are told so plainly, rather than being shown a sent message the customer never gets. The approved wording is sent from Settings → Twilio → WhatsApp, which is also where you add it. See WhatsApp.
- A caller is blocked by marking their conversation as junk, if the channel they came in on can do it. Press the no-entry sign at the top of the conversation and the question that comes up asks whether you would also like to stop them getting through in future; No, just move it does the junk and leaves them alone. There is no longer a standing Block them button in the header - it was the one control up there that changed what a stranger gets when they next try to reach you, it sat in front of everybody all day, and it was pressed by mistake. On the phone, blocking means their next call is dropped the moment it arrives: nobody's phone rings, no message is taken and no alert goes out. It deletes nothing they have already said, and once somebody is blocked an Unblock them button appears at the top of that conversation, which is the only time it is on the screen at all. See Blocking a caller.
An address, a phone number or a chat account belongs to a person, and a person can have several of each. Somebody who emailed in March and rang in April is one person with one history rather than two strangers.
Organisations are guessed from the email domain, with the usual free providers left well alone: nobody's gmail.com address turns Gmail into one of your customers.
Some things worth saying plainly:
-
A role address is one person.
accounts@supplier.commight be four people in real life, and here it is one. That is by design; splitting it would mean guessing. - Your own people never become customers. Your inboxes, your staff's own addresses and anything at your own domains are excluded. If it guesses your domains wrongly, set them yourself under People → Your own domains.
- There is a directory now, under the Contacts tab, and it was left out of earlier versions on purpose - an address book is where a conversation hub turns into a CRM by accident. What it holds is names, numbers and where to post something. There is no stage, no value, no next action and no forecast, and there is not going to be. See Contacts below.
- People can be merged when the same human turns up twice, and a merge can be put back afterwards. A person can be split apart again if two people got folded into one.
- Conversations can be merged too, which is a different job from merging people: one is "these two threads are the same story", the other is "these two cards are the same human". See Merging conversations into one.
The Contacts tab is the address book. It is the same people the section above describes, listed instead of waited for.
Everybody who writes in is already in it. What is new is that you can now write things down about them, and add somebody who has never written at all.
A first name and a last name, separately, so the list sorts by surname the way an address book does. Then a job title, the organisation they work for, any number of categories, as many email addresses and phone numbers as they actually use, a website, a postal address in the usual British parts - address, town, county, postcode, country - and a note.
Somebody the post introduced us to may only ever have had one name, read off the From line. Opening their card splits it into the two boxes for you, so Jane Smith arrives as Jane and Smith and you can correct it if the guess was wrong.
Phone numbers are not just text. A number typed onto a card is a number the site recognises: when that person telephones, the call lands on their record rather than starting a stranger.
One address belongs to one person. If you type in an address that somebody else here already has, it stays with them and the card tells you so, rather than quietly moving it. Two people holding one mailbox is a merge for you to decide on - see the section above.
A contact can have as many phone numbers and postal addresses as they need, and each can say what it is for: Order and Delivery for a customer's two numbers, Billing and Delivery for their two addresses, or anything else you type (Head office, Site, Mobile). On the card, Phone numbers is a list with a label box beside each number and Add another number under it; under Where they are, the ordinary address boxes take a What for label and Add another address adds a further one, with its own label, below.
Saving the card means "this is the whole list": a number or further address you take off the card is taken off the contact. (An import, or a paid order, only ever adds - they never remove anything.) A number somebody else already holds stays with them and the card says so, the same as for email addresses.
When a customer's order is paid and they are added to the address book, the card is filled in the same way: the order's number is labelled Order, the delivery address's number Delivery, the billing address Billing and a different delivery address Delivery too. A card you have already typed is never rewritten - an order only adds what the card does not hold.
Their page shows each address with its label above it and each number with its label beside it, and Download everything we hold includes both. A contact found by searching a postcode is found by any of their addresses.
The sort button in the contacts toolbar orders by date added, newest first to begin with, and flips to oldest first. Searching is still how you find a particular name. The import button is the circle to the left of the filter and sort buttons.
The second half of the tab. Most appear on their own, one per company that writes in from its own domain. Each has a card with a name, a mail domain, an address, a phone number, a website and a note, and says how many contacts are in it - which is what makes Acme Ltd with fourteen contacts and Acme Limited with one obvious enough to tidy up.
You can add one yourself, which is the only way the haulier who only ever telephones is going to be in there. Leave the mail domain blank for a company that never emails; fill it in and the next person who writes in from that domain joins them automatically.
Removing an organisation keeps everybody in it. They keep their own records and simply stop showing the name. A contact is not the company they work for, and no conversation, order or invoice is touched.
The labels you file contacts under: Supplier, Trade customer, Haulier, Do not ring - whatever suits how you actually work. The filter button in the contacts toolbar (the three lines, beside the sort button) opens a list of them with ticks: tick one or several and the list shows everybody who wears any of the ticked labels. Everybody at the top of the list clears it. The choice is held in the address, so a filtered view can be sent to a colleague.
A contact can be in several at once, on purpose. Somebody can genuinely be a supplier and a customer, and a field that only allows one would have you picking which of two true things to record.
You make a category by using it. Open a contact, press Edit their details, and the Categories line offers everything the site already has as ticks, plus New category for one that does not exist yet. Nothing is saved until you save the card, so a label typed onto a contact you then abandon does not leave a category behind.
Renaming and removing them is under Settings → Unified Inbox → People, along with the order they appear in and how many contacts are in each - which is what makes an empty one worth removing and a typo worth renaming. Removing a category keeps everybody who was in it. They keep their records and simply stop showing the label.
One thing said plainly, because it is the line this feature must not cross: a category is not a stage. Nothing moves between them on its own, nothing else on the site reads them, and there is no order to them beyond the one you drag them into. It is the label on the drawer, not a sales process. The single label anything puts on for you is Customers, on somebody who pays for a shop order - see Shop customers - and even that only ever adds.
Contact, in the menu beside the new email button (the little arrow, under Discussion, Call and SMS). It opens the new contact form in a dialog over whatever you were looking at. Fill in what you know - a phone number on its own is a perfectly good contact - and save. Anybody who can reply to the post can do this; you do not have to look after the site.
Import a file, beside it. This one is for whoever looks after the site: two thousand contacts in one press is a wider thing than correcting one of them.
Three steps.
- Choose the file. A CSV, saved out of a spreadsheet or exported from wherever your contacts are now. There is a blank one to download if you would rather start from ours.
- Match the columns. Every column in your file is listed with its best guess at what it is - and it has seen what Outlook, Google Contacts and most accounts packages call things, so an ordinary export usually needs nothing changed. Change anything that looks wrong, and set anything you do not want to Leave this column out. A single Name column is understood and split into first and last, and a Category, Group, Type or Tags column is understood as categories - several to a cell if they are separated by commas, semicolons or pipes.
- Check and bring them in. The first five rows are shown as they would actually be saved, so a column in the wrong place is visible before rather than after. There is also a Category box here for putting everybody in the file into one, for when what four hundred contacts have in common is the file rather than anything written in it - it goes on top of whatever a category column says, never instead of it. Then press the button.
Three things it does that are worth knowing:
-
Somebody already here is left exactly as they are, unless you tick Fill in contacts already here. Off is the safe answer: a file with half its columns blank would otherwise blank those fields for everybody in it. Running the same import twice with it off changes nothing the second time.
-
One company named a thousand times is one organisation. Names are matched however they are typed, so
Acme Ltdandacme ltddo not become two. -
An import only ever adds a category, never takes one off. A sheet saying "Supplier" is telling you one thing about somebody, not the whole of what they are, so anybody already labelled keeps what they had. And "Supplier" and "supplier" are one category rather than two.
-
A row that will not go in is reported by its row number, the same number your spreadsheet shows down the left, along with what was wrong with it. The rest still go in.
-
A big file goes up in pieces, a few hundred rows at a time, and the button counts them off as it goes. That is why a long import no longer stops halfway with a message about the file being unreadable: it never was, there was simply too much of it to send in one go. If something does stop it partway, the contacts already brought in stay in - bring the rest in on their own rather than running the whole file again.
-
If a row is refused, it says which row and which column - "row 414, column H" - rather than telling you the file could not be read and leaving you to find it.
Five thousand rows at a time is the limit. Split a bigger file and bring it in in two goes.
The file never leaves your computer. It is read in your browser so the matching step has something to show you, and only the rows you can see are sent to the site. There is no copy of your address book sitting in storage afterwards, which is rather the point.
Where the shop is installed as well, somebody who places an order becomes a contact without anybody typing them in. Their card gets what they gave at the checkout:
- Name - the one given with their email address. When the billing address names the same person, its first and last name are used as they were typed, so Mary Ann Jones arrives as Mary Ann and Jones rather than Mary and Ann Jones. An invoice addressed to somebody else in their office does not change who the contact is.
- Email address, which is what everything else is matched on.
- Phone number - the contact number they gave, or the one on the billing address. Never the delivery address's, which is very often whoever is on site to sign for the parcel.
- Company - the organisation they said they were buying for, found by name among your organisations or made if it is new.
-
Billing address - or the delivery address, on an order that has no separate billing address, which is how the shop's own invoice reads it too. The country is written out in full:
GBbecomes United Kingdom. - The Customers category, made the first time it is needed. One you already have called Customers, however it is capitalised, is used rather than a second one made beside it. Renaming it stops it being found, and the next paid order makes a fresh one.
Nobody is copied. Somebody already in your contacts - because they emailed before they bought, or have bought before - is found by their email address and filled in rather than replaced: a box that is empty on their card gets the order's answer, and anything already written there stays exactly as it is. The name and the postal address go in whole or not at all, so a card is never left with a postcode from one order under a street from somewhere else. They gain the Customers label and keep every other. A number they gave the shop is added to the ones they have, unless it already belongs to somebody else here, in which case it stays with them.
When it happens: the moment the order is paid for, however that happened - card, a provider, a bank transfer marked as received by hand, or a free order - and, for bank transfer and any other method paid by hand, the moment the order is placed as well. Somebody who emailed in and then ordered on bank transfer has their number, address and the Customers label straight away rather than whenever the money lands; clearing the payment later runs it again and changes nothing, because it only ever fills blanks and adds. A delivery number that is the order number in international form (+44 7311 351565 against 07311 351565) is recognised as the same phone and not added twice. Every number the address book stores is written one way: a number in the site's own country (Settings > General, dialling code) nationally with no spaces - 07311351565 - and any other with its country code and no spaces. A text or call from +44 7311 351565 therefore finds the card that says 07311351565 rather than starting a nameless contact of its own. Calls and texts read the same way: the conversation's title, the caller's name where the channel only gave a number, and the number on each message are all written 01217329511 rather than +441217329511. Replies still go out on the number the channel knows them by.
Who is never added: anybody at one of your own domains, one of your staff, or one of your own inboxes - the same rule the post follows, so a colleague testing the checkout from a work address does not become a customer.
It is on unless you switch it off, under Settings → Unified Inbox → People → Customers from the shop; the switch only appears where the shop is installed. The first time a site takes this update, the customers it already has are not added in one go - they arrive as they next order. Nothing about the shop changes and it needs no update of its own: the shop already says when an order has been paid for, and this module is simply listening.
Every contact card says whether that person is happy to be sent marketing email - Happy to receive them, Asked not to receive them or Not asked - and, under it, where the answer came from and when: Order DW000218 · 5 October 2026, 19:25, Unsubscribed by email link, or Changed by hand by whoever changed it. Consent nobody can trace back is consent nobody can show, so the trail is kept, not just the tick.
- From the shop. Where the checkout asks the question, each order's answer goes onto the customer's card when the order comes through, alongside everything else above. The newest answer wins: somebody who said yes in March and no in October is a no, and an older order arriving late never overrides a newer answer. An order whose checkout never asked leaves the card as it was.
- By hand. Somebody tells you on the phone? Change it on their card under Marketing emails. It is recorded as said now, with your name, so it outranks every order they placed before.
- Unsubscribing from a campaign marks their card Asked not to receive them as well as taking them off the list.
- Campaigns leave them out. Somebody whose card says no is not put on a campaign, and the Who step says why: They asked not to be sent marketing email. "Not asked" is treated as before.
- Merging two cards keeps whichever of the two answers was given last; undoing the merge puts the other back. Their answer is in the data export too.
Pressing a link in somebody's email does not follow it. It opens a small panel first, showing where that link actually goes: the site it lands on, the whole address written out, and - when the two are not the same thing - what the link said versus where it points.
That gap is the entire mechanism of every phishing email ever sent. "View your invoice" over an address in another country looks exactly like "View your invoice", and a mail program that follows the press without showing you where you are going is helping. From the panel you can open it in a new tab, copy the address to paste somewhere that checks it, or close and go no further. Nothing loads until you choose.
Addresses in plain-text messages are links now too - they used to be text you could not press at all. Anything that is not an ordinary web or email address is shown and refused: the panel tells you what it is, and will not open it.
The panel also says when an address is not the real one. With link tracking switched on, Brevo replaces every link in what you send with one that goes to Brevo first, writes down that it was followed, and then passes you on - so opening one out of your own Sent folder is recorded as the person you wrote to having clicked it. The panel now says so before you press anything. It still opens if you want it to; you simply know what it costs.
Holding cmd (or ctrl) while you press still opens a link straight away in a new tab, the way it does everywhere else.
The same email to a great many people, one at a time, slowly, from an address that is already a real mailbox on this site.
That last part is the whole idea. This is not a newsletter service and it is not trying to be one: no template gallery, no rented sending domain, no branded footer with a picture in it. It is the same composer you already use, pointed at a list from your address book, with a clock in front of it - one message every ninety seconds inside working hours, so two thousand emails leave looking like two thousand emails a person sent rather than one burst a filtering service is paid to notice.
You will find it as Campaigns in the rail down the left of the hub, under Everything else, next to Contacts. It has its own permission, so somebody who can answer the post is not automatically somebody who can write to five thousand customers.
It is laid out exactly like your post: every campaign down the middle column with its own progress stripe, and the one you have opened in the pane beside it. Picking another is one press and you keep your place, so glancing at what a second campaign is doing no longer costs you the one you were writing. The edges between the three columns drag, and they are the same edges you have already dragged in the inbox.
On a phone there is only room for one at a time, so you get the list, and then the campaign you opened from it with All campaigns to get back - the same as opening a conversation.
One page, top to bottom, in the order you would think about it, with one Save button at the bottom of the pane. It stays put as you scroll and it saves the whole thing, so there is never a question about which button saved which half. The name is at the top, typed where you read it.
Nothing you type is thrown away by something else happening. Topping the list up, pausing it, stopping it, pressing Send now beside somebody, or a colleague saving the same campaign from another tab used to empty every box back to the last saved version - half-written follow-ups included. Now what you have typed stays where it is, a line appears saying the campaign has changed underneath you, and Save keeps your version. Two things do snap back to the saved copy, and only if the campaign STARTS while you are typing: who it goes to, and the wording of the first message. Those are locked from that moment, so keeping your edit would leave you with a form that could not be saved and a greyed-out box you could not put back. If you try to leave a campaign with something unsaved in it - clicking another campaign, the Do-not-email list, or closing the tab - you are asked first.
Save is there on every campaign, finished ones included. Two things lock once the first message has gone - who it goes to, and the wording of that first message, because some people have already had it and two versions of one mailshot with no way to tell who got which helps nobody. Everything else stays yours to change: the name, the follow-ups, the clock. (Previously a finished campaign had no Save button at all, while its boxes still let you type - so the changes went nowhere.)
Who it goes to - the address it comes from, and which of your contacts get it. Pick one or more of the labels from your address book, or none of them for everybody. It counts them for you as you go, and it tells you who is being left out and why: people who have unsubscribed, addresses that have bounced, colleagues at your own domain, anybody another campaign wrote to in the last week, two contacts sharing one address. Saving is what writes the list down - there is no second button to press - and once it is written it is fixed, because a list that changes every time somebody imports a spreadsheet is a list whose finish date moves. Top up later adds anybody who has appeared since and leaves everybody already on it alone.
What it says - the subject and the message. Write it as you would write an email, because it is one. Somebody's name goes in with {{first_name|there}} - the bit after the bar is what anybody with no first name on their record gets, so nobody receives "Hello ,". There are five of these: first name, surname, whole name, company and their email address, and there is a row of buttons that types them for you. It will show you how the message reads for three real people off your own list, chosen so that the one with no first name is first. You can turn the signature off for this campaign, and you can send yourself a test - which you have to, before it will start.
When it goes - the clock, and it is two things rather than fifteen. First, when not to begin before, if you care. Then one sentence telling you what the pace actually comes to: "About 320 a day, at any hour, any day of the week, 90 seconds apart", with roughly when it will finish underneath. That sentence is the answer, so most people read it and move on.
Everything behind it is under Change the pace, the hours or the days: sending hours, weekdays only, days to sit out entirely - weekdays-only still sends on Christmas Day - the gap between messages, how much to vary it, a daily ceiling, and a warm-up that starts small and doubles, worth switching on if this mailbox normally sends a handful a day and is about to send three hundred. Every box in there can be left empty, and each one tells you what empty means. No hours set means any hour of the day and night. Nothing is filled in for you, because a time you did not type is a decision somebody else made on your behalf.
Before it can go - the checklist, underneath the form. Anything with a cross by it stops the campaign; anything with an exclamation mark is a warning you can read and press past. Start sending appears once there is nothing left to fix, and if you have typed something since your last save it saves that first - one press, not two. It used to disappear the moment you touched anything, which is exactly when you had just finished writing. Leaving the sending hours empty is one of those warnings: perfectly reasonable for a reminder, less so for a mailshot that lands at three in the morning on a Sunday.
Where everybody has got to - at the foot of the same page, as soon as there is a list to look at: everybody on the campaign and where they have got to, filterable. Who has had it, who replied, who bounced, who unsubscribed, and who was left out with the reason written beside them. Anybody still waiting has a Send now beside them, for when you do not want to wait for the clock - it skips the hours and the gap, not the checks, and it will not go out of a campaign that has never been started. (This used to be a separate screen behind a Progress button, which meant reading what went out and reading what it said were two places with a toggle between them. It is one page now, so the button and its Back to the wording twin are both gone.)
A campaign can carry up to three follow-ups, each with its own wait and its own wording. They only go to people who have said nothing: a reply from any address you hold for that person stands the chase down, and so does a bounce or an unsubscribe. An out-of-office does not - a fortnight in Spain is not an answer.
A follow-up with no subject of its own goes out as a reply to the first message, so it lands in the same conversation in their mail program rather than arriving as a stranger asking whether they saw the last one. That is usually what you want.
Taking a follow-up out of a campaign that has already started leaves a gap in the numbering, on purpose. Everybody still waiting is waiting for a follow-up by number: shuffle the rest up and the people queued for the second one quietly receive the third instead. So the ones after it keep their numbers, anybody who was waiting for the one you removed is simply finished, and nobody gets the wrong email. On a campaign that has not started yet they are renumbered as you would expect, because nobody is waiting for anything.
Three things stop a campaign without being asked, and all three exist because nobody watches a screen for a fortnight.
- Too many bad addresses. If more than five in a hundred come straight back, it stops and says so. A list full of dead addresses is how a domain ends up in a filter, and stopping at forty is a great deal cheaper than stopping at four hundred.
- The mail service saying no. An unverified sending address or a rejected key would fail the same way for every message behind it, so it stops rather than failing nine hundred times.
- The address disappearing. Delete the inbox it sends from and it stops, with everything it has already sent kept exactly as it was.
Pause and Resume are yours to press, on the bar at the foot of the campaign - they used to be on its row in the list as well, which put three buttons under the thumb of anybody scrolling past. Pausing takes effect at the next gap: whatever was going out at the time finishes. Stop for good ends it, follow-ups included.
A finished campaign carries Send it all again on the bar at the bottom. It is mostly for the campaign you fire at a test mailbox twenty times while you get the wording right, and for the reminder that goes to the same list every September.
It is behind a warning, and the warning is the honest bit rather than the polite bit:
- Everybody on the list gets it a second time, the people who have already had it included. A customer who receives the same email twice unsubscribes, and they are right to.
- The record of the last run goes. Who opened it, what bounced, who replied - the progress table starts again from nothing, and there is no getting it back.
- Unsubscribes and dead addresses are kept. Anybody who opted out stays out, and so does every address that bounced. That much survives whatever else this does.
- It comes back as a draft. Nothing leaves until you press Start sending, and the checks run again first.
A campaign marks itself finished the moment the last person on it has been dealt with. Adding people to a finished campaign with Top up sets it going again, because adding somebody to a finished campaign is somebody saying it is not finished. Paused and stopped are left where they are: both of those were a decision, and a top-up is not a way to overturn one.
Every campaign message carries a footer with your business name and address and a link to stop receiving them. Somebody who uses it is off the list within the second - not just that campaign but every campaign this site will ever send, because "unsubscribe" means from you rather than from one mailshot. It works without a login, it keeps working long after that campaign has been deleted, and mail programs that offer their own unsubscribe button will show one.
You can switch the footer off per campaign. Before you do: marketing email in the UK is expected to carry a way to opt out, and without one people press the spam button instead - which does far more damage to whether your ordinary email arrives than an unsubscribe ever could. The screen says so, once, and then it is your call.
The Do-not-email list - the button above the campaign list - holds everybody who is off the list, and why. You can add an address by hand - somebody who asked you at a trade show - and take one off, for the address that bounced during an outage and works again on Thursday. It stops campaigns only: replies, order confirmations and anything somebody has specifically asked you for still go out, which is right and is what the law expects.
Your business name and address for that footer go in Settings → Unified Inbox → Campaigns, along with how many days must pass before any campaign may write to the same person again. Seven is the default and it is a guard rather than a preference: two campaigns should not both land on the same customer on the same morning.
This is worth two minutes of your attention, because it decides whether a campaign keeps to the pace you set.
While any inbox screen is open in front of you, it sends on time. Reading your post is enough - it does not have to be the Campaigns screen, and it never did make much sense that it did. The page quietly asks every half minute whether anything is due, which costs nothing when nothing is: the gap between messages is kept on our side, so a request that arrives early sends nothing at all.
When nobody is looking, it relies on your site's own scheduled round, which on most hosting comes past every few minutes at best and once an hour at worst - so an unattended campaign creeps along rather than keeping to your ninety seconds. A tab left open behind another window counts as nobody looking: browsers stop the clock on a background tab, which is exactly what you would want a page left open for a fortnight to do.
If you want it to keep proper time with nobody watching, the When section shows an address you can point any free website-pinger at, once a minute. Treat it like a password: anybody who has it can nudge your campaigns along. Asking it too often is harmless - the gap is kept on our side, and a request that arrives early sends nothing at all.
- No attachments. A mailshot with a PDF on it goes to the junk folder, and a few thousand copies through your own mailbox is a bad afternoon for everybody. Link to it instead.
- No click tracking. Rewriting every link in an email is what marketing mail does, and it is one of the things filters look for.
- No conversation per send, unless you ask for one. Five thousand of them would bury your actual correspondence, so nothing a campaign sends appears in your post by itself. A reply makes a conversation, on the next check, exactly as any other email does - which is the only part you wanted in the inbox anyway. If you do switch on Also file every one in the mailbox's Sent folder, those copies come back on the next check and each one opens its own conversation, with the person it was sent to - not one conversation with two hundred messages in it. (It used to be exactly that. Two mailshots with the same subject looked like one conversation because they had your own address in common, which is true of everything you have ever sent and tells nobody anything.)
- No A/B subject lines, no scores, no funnels. It is still an inbox.
One line sits directly under the buttons at the top of a conversation and stays there while you scroll: the conversation's context - "Purchase order PO-0023, Order DW0234".
It used to be called attachments, which was asking for trouble: a message carries a file, and a conversation is about an order, and one word for both meant nobody could tell from it which was meant. Files on a message are still attachments. The records a conversation is about are its context.
Who you are talking to is not repeated there. Every message in the conversation carries the sender's name and their company at the head of it, and the row in the list beside it says the same again.
Every name on that line is a link, and it opens in a new tab. Following an order out of a conversation is almost always reading the order while answering the message, so the conversation - and the half-written reply under it - stays where it was.
If that line has more on it than will fit, it stops with an ellipsis rather than pushing the message down the screen. The arrow at the end of it opens the full list, which is also where anything is added or taken off.
Taking something off works from that list. The cross beside a record asks first, and answering Take it off used to close the list without doing anything - the question was drawn over the top of the page rather than inside the list, so answering it read as a click somewhere else and the list shut before the answer landed. The record itself is never touched either way, and you can add it again afterwards.
A conversation with nothing on it yet has no line at all. It used to get one anyway, reading "No context yet" - a whole row of a pinned header spent saying there was nothing to say. The arrow moves up instead, onto the end of the line that already reads Email - General Enquiries - 1 message - last message 20:32, where it costs nothing and is still exactly where you would go to attach the first one.
Anything you put on a message out of your catalogue joins that line once the message has gone, beside the orders and the purchase orders, and clicking it opens the product in a new tab. So a conversation that started "have you got that in oak" says what it was about a fortnight later, without anybody reading back through it. Quote the same chair twice and it appears once. This happens whichever channel the reply went out on - an enquiry answered with two chairs says so on the line as plainly as an email does.
If whoever is writing has bought from the shop before, Existing customer sits on that same context line. It is not a record and nobody put it there - it is simply the one fact about a message from an unfamiliar name that changes how it gets answered, and otherwise you have to go and look for it.
Click it and the orders list opens in a new tab, already narrowed to that person. Which order they mean is usually the reason for the trip, so it opens the list rather than picking one for you.
Attach the actual order and it goes away. At that point the line says which order, which is the better answer to the same question, and having both on it was the vague version elbowing the precise one. Take the order off again and it comes back.
You only see it if you are allowed into the shop in the first place. On a site with no shop it never appears.
When a person's address matches something else on your site - their orders, their quotes, their unpaid invoices, their member account - it appears on their own page, reached by clicking their name anywhere in the inbox. Each one is a link through to the real record. Nothing is ever written to those records from here, only read, and a module you have not installed simply shows nothing.
It is no longer stacked under the conversation. It used to be a fourth column beside the messages on a very wide window - and on anything narrower, a run of headings underneath the note line at the bottom of the conversation, which is to say below however many thousand pixels of quoted email the thread happened to hold. Nobody ever scrolled to it, and putting it there cost the site a question to every record-keeping module every time anybody opened a conversation. What is worth knowing beside a conversation is what the conversation is about, and that is the one line in the header above. The rest is one click away, on the page it belongs on.
When somebody quotes an order number, a purchase order number or a quote reference in a message, it becomes context on the conversation automatically. Nothing is added until the number is checked and found to exist, anything added that way says it was found automatically, and it comes off in one click. If your reference numbers are an unusual shape, the three boxes under People → Spotting references take your own pattern.
Most people do not quote their order number, and a supplier answering a purchase order almost never does. Add context, at the foot of the list behind that arrow, opens a list of your records to pick from rather than a box demanding a number.
It opens on whichever kind suits the address you are reading. If purchasing sends its emails as this address, it opens on purchase orders; if the shop sends as it, it opens on orders. Which address each of them sends as is the Sending address panel on that module's own settings - Settings → Purchase Orders, and the Notifications section of Settings → Shop. If you have not chosen one, the list simply starts on the first kind of record you have.
The records belonging to whoever is on the conversation come first, newest first among them, so a supplier's own open orders are usually the first thing in the list and adding one is a single click. Typing narrows it, by number, by the name on the record or by the address it was placed with. You can still add something by its number alone: type the number and press Add, and it goes on if a record with that number exists.
"Whoever is on the conversation" means everybody on it, not only the name at the top: whoever wrote, whoever it was written to, and anybody copied in - your own addresses, your colleagues and your own domains taken out, since none of those is the customer. Orders are matched on the address itself. Purchase orders are matched on the domain, because a supplier writes from whichever desk happens to be answering: an email from anybody at dynamicos.co.uk opens on Dynamic Office Solutions' own purchase orders, newest first, whether or not that particular person is the address the order was sent to.
On a discussion between colleagues there is nobody outside the building to rank by, so the list simply opens on your newest records. It used to open on nothing at all - the list came back empty for orders, quotes and purchase orders alike, however long you left it. That is fixed.
You only see records from parts of the site you are allowed into. Somebody who reads the inbox but has no business in the shop is not offered a list of your customers' orders, and if the site keeps no records of that sort at all, there is no button to press.
Order confirmations, purchase order emails and the like are sent by your email service and never touch your mailbox, so no mail client can be asked about them. They appear on a person's timeline anyway, from the site's own record of what it sent: what went, when, and whether it failed. There is no copy of what it said, because that record is a delivery ledger rather than an archive, and the page says so where it shows them.
One honest limitation: your site's own notification emails, such as "somebody has filled in your contact form", no longer clutter the inbox as unread conversations. They are recognised and kept quiet. They are not yet folded onto the enquiry they are about, so a form submission can still show up twice: once as the enquiry, and once as the site's own note to you about it.
Four permissions, handed out through Managing users as usual:
| Permission | What it allows |
|---|---|
unifiedinbox.view |
Read conversations, mark them as junk in your own spam folder, delete them into your own bin, and read the address book |
unifiedinbox.reply |
Reply, forward, note, assign, snooze, mark done, block a sender, add and correct a contact, make a category |
unifiedinbox.manage |
Settings, mail accounts, inboxes, the list of blocked senders and letting one back in, merging conversations, merging people, moving one message out to its own conversation, deleting one message, emptying a bin, export and erase, importing an address book, removing an organisation, renaming or removing a category |
unifiedinbox.campaigns |
Write, start, pause and stop campaigns, and manage the do-not-email list |
Campaigns is its own permission on purpose. Renaming a folder and emailing five thousand customers are not the same act, and a site that has given somebody the first has not thereby given them the second. Sending a campaign from an address still needs permission to reply from that address as well - a campaign is a great many replies, and it is not a way round the guest list.
Giving somebody an inbox of their own is a convenience, not a permission: it decides what they open on and what they sign off as, and nothing about what they are allowed to read.
On top of that, each shared inbox has its own list of who may read it. Leave the list empty and anybody with the view permission can read that inbox. Name anybody at all and it is those people and nobody else. A shop assistant does not need accounts@.
That applies to searching and to the All view as well, not only to opening a conversation: a snippet from an inbox you cannot open never appears in your results at all. Administrators are the exception, on the grounds that whoever edits the guest lists could add themselves to one in two clicks anyway.
Being tagged in a note is the one way past a guest list. A colleague who could already open a conversation can ask somebody else to look at it, and that person can then read that conversation and work through their own copy of the ask - but nothing else in the inbox opens to them, and answering still needs permission to send from the address. See Asking a colleague to look at something.
Merging two conversations across two addresses opens both halves to both address's readers. That is what merging them means - a conversation half the people on it cannot open is not one conversation - but it does widen who can read something, and it is worth a thought before you do it rather than afterwards. The confirmation names every address involved. It can be put back: separating the conversations out again takes the reading back with them. See Merging conversations into one.
An individual inbox is the one exception to that exception. It has no guest list to add yourself to, and being an administrator does not open it - see "Shared and individual inboxes" above. Administrators keep the configuration and lose the reading, which is the honest half of the bargain.
Collecting → Delete conversations older than is blank to begin with, which means nothing is ever removed. Set a number of months and a daily tidy-up removes conversations whose last message is older than that, along with any files attached to them. There is no way to get them back afterwards.
The whole conversation goes, never half of it. Deleting the old half of a thread and keeping the recent half leaves something that reads as though your customer opened with a reply.
Keep a conversation for ever if it has an order, a purchase order or a quote attached is on by default, and is what stops a window aimed at old mailing lists quietly taking the correspondence behind a disputed invoice with it. The settings page tells you both numbers before anything happens: how many the next tidy-up would remove, and how many are old enough but are being kept because something is attached to them.
On a person's page, an administrator can download everything held about them as a file, or erase them.
Erase covers this hub and nothing else. The dialog counts up what will go before you press anything, and says in the same breath what will not:
- their conversations, messages, attached files, addresses and phone numbers, everything worked out about them, and every campaign row holding their name: removed;
- their orders, invoices, quotes, purchase orders and member account: untouched, and named by module so you know where to go next;
- your site's record that automated emails were sent to them, which holds their address and the subject lines: untouched.
That last one is deliberate rather than forgotten. It is a core record rather than one of this module's, and under-deleting where you can see it is safer than over-deleting where you cannot.
The export file carries the same three sentences at the top, because whoever opens it next may be a solicitor rather than the person who exported it.
Under Settings → Unified Inbox → Other apps you can have an inbox notify a web address every time a message arrives. Nothing about the inbox itself changes, and switching it all off breaks nothing.
Each one has:
-
A web address to tell. Has to start with
https, and has to be somewhere on the open internet - an address pointing back at the site's own server, or at something on its private network, is refused when you save it and again before every note goes out. - Which inbox. One of them, or every inbox including any you add later.
-
What to tell it about. Tick any of four:
- Post arriving from outside - an email or message lands in the inbox.
- A colleague writing in a discussion - a note in an internal discussion put to whoever owns the inbox, or filed in it.
- Being asked to look at something - a colleague tags whoever owns the inbox in a note, on an email conversation or a discussion.
- Being handed a conversation - a colleague assigns a conversation to whoever owns the inbox.
- What to send. Either the details of the message that arrived - who it was from, the subject, which inbox, which conversation - or the same fixed message every time, which is what an address expecting its own wording needs.
- Whether to include what the message actually said. Off by default, and worth leaving off unless the other end genuinely needs it. Switching it on sends a copy of your post to that address every time one arrives.
- A signing password. If the other end expects one, every note is stamped with it so it can tell the message really came from your site.
-
Extra headers, written one per line as
Name: value. This is where a key goes if the address you are telling asks for one.
An inbox is not a person, but an inbox that belongs to one person is as good as. Point a subscription at Marcus's own inbox and it hears about Marcus: post landing in it, discussions put to him, the times he is tagged, and conversations handed to him - wherever those conversations happen to live. Point it at every inbox and it hears about all of it, for everybody. Point it at a shared inbox and it hears about discussions filed there and its conversations being handed to somebody - but never about anybody being tagged, because nobody owns a shared address, and a tag goes to a person.
Nobody is told about their own doing. A note you wrote, or a conversation you took yourself, sends nothing - your automation already knows. The automatic hand-overs (a conversation put on your desk because your own post started it, or because you asked to be reminded to chase it) are the same thing and send nothing either.
Each message is told about once. An email lands in Marcus's inbox and his address is told. A few seconds later somebody assigns it to him - and nothing more is sent, because it is the same email and his address already knows about it. The hand-over is only passed on when there is post the address has not heard about yet: a subscription that only listens for hand-overs, or a newer reply that arrived since. A discussion put to Marcus is also him being asked to look at it; that is one note too, not two.
For the details of the message style, these three add two things to what an
arrival carries: by, the colleague who wrote the note or handed it over,
and for, the colleagues the note is about, each saying whether they were
addressed (on the discussion), mentioned (tagged) or assigned. The
event field says which of the four it was: message.received,
discussion.received, mention.received or conversation.assigned.
Then it keeps quiet for a while. An automation told about new post usually goes and deals with the whole inbox, not just the one message. So after each note a subscription sends, it says nothing more for ten minutes - whatever happens in that time is not sent at all, because whatever the first note set going will find it anyway. Each subscription has its own Then keep quiet for setting in minutes; 0 sends every one, and a day is the most it will take.
Login codes are never passed on. An automation told about Marcus's post often signs in to deal with it, and signing in emails Marcus a code - which lands in Marcus's inbox. Passing that on would set the automation off again, for ever, and would post a login code to somewhere outside the site besides. So the site's own account emails - login codes, recovery links, email-change codes - are filed like any other post but never sent to a web address.
At the top of the page sit a signing password and a set of extra headers shared by every subscription. Most sites tell one thing about their post - a workflow tool, an ops channel - from several inboxes, and typing the same key into five places means changing it in five places on the day it is rotated. Set it once here and every subscription set to "shared" follows on its next send.
Each subscription then chooses, for the password and for the headers separately: use the shared one, give this one its own, or neither. Anything set up before this carries on exactly as it was - a subscription that had its own keeps using it, and one that had none stays unsigned until you say otherwise.
Neither the shared password nor the shared headers are ever shown again once saved: the boxes replace them rather than edit them, and there is a Remove it beside each when one is set. The list says which of the two each subscription is actually using, so "shared, but nothing shared is set" cannot quietly look like "signed".
Notes go out on the mail check's schedule rather than the moment a message lands, so an address that is slow or switched off can never hold up your mail. One that does not answer is tried again after a minute, then five, fifteen, an hour, three hours and twelve - and after twenty failures in a row it switches itself off and says so on the screen. Editing it starts it again.
Send a test fires a made-up message at the address there and then, shaped as the first thing on the list it is ticked for, and tells you what came back. Nobody's real post is used to prove a web address works. History shows the last twenty notes, what each one was about, and what happened to it.
Some of your other modules may want to know when an email arrives - purchasing filing a supplier's proforma on the right purchase order, say. The inbox tells them, and they decide whether it is any of their business. On a site where nothing is listening, none of this happens and nothing extra is done.
What they are told about: post that has arrived from outside, once it and its attachments are safely stored. Never your own replies, never a colleague's note, never an out-of-office, bounce or newsletter, never post from a sender you have blocked, and never a conversation somebody has marked as junk. Nor the history a new mail account is read in with - a proforma from last March is not news.
What they can do about it: put a link on the conversation to the record the email is about, exactly like the ones the inbox finds from a reference number. It says it was found automatically and comes off in one click - and once you have taken one off, it stays off: offering the post again does not put it back. Adding it yourself later works as normal. They can also leave one line on the message itself, such as "Filed on PO-01234 as the proforma", shown just above the message. Hover over the line to see which part of the site wrote it.
If something goes wrong: a module that takes too long or falls over is given up on and the next one is told; collecting your post is never held up. Anything that was not offered at the time - the site was being updated, the check ran out of time - is offered on the next hourly check, for up to three days.
Offer the last 14 days again, on each address under Settings → Unified Inbox → Inboxes, hands that address's last fortnight of post to the listening modules again, whether or not they have seen it. It is for switching a module on after the paperwork has already arrived. Nothing is done twice - a module that has already filed something recognises it. The button only appears when something is listening, and only on an address post actually arrives at.
A module listens by registering a handler in its own cactus.module.json. Colleague-to-colleague email between two of the site's own addresses is offered too, on the side that received it. Mail the inbox classes as automated is never offered: anything with an Auto-Submitted header other than no, a Precedence of bulk, list or junk, any List-Id, a delivery report or other bounce, or mail from the site's own sending address. Some carriers' and suppliers' order systems set those headers on perfectly useful mail, so a module that needs it cannot rely on this point for it today.
{
"point": "unified-inbox.message-received",
"id": "my-module-inbound",
"serverOnly": true,
"import": "./lib/inbound-handler",
"component": "myInboundHandler"
}serverOnly keeps it out of the map public pages load; the inbox reads the server map. Do not import anything from @/modules/unified-inbox - restate the types below in your own file. They are structural, so a later inbox that sends more fields does not break you.
component names either the handler function itself, or - when you need files in hand - an object:
export const myInboundHandler = {
attachmentTypes: ['application/pdf'],
handle: async (event: InboundMessageEvent, context: { signal: AbortSignal }) => { /* ... */ },
}attachmentTypes asks the inbox to store files of those kinds in the media library as mail arrives, so mediaId is set. It lives on the export, not on the manifest entry, because core stores a parsed manifest and drops fields its schema does not know. Only application/pdf, image/png, image/jpeg, image/gif and image/webp are honoured; anything else you name is ignored. Each file is judged by its own first bytes, not the sender's label, and stored as the kind its bytes prove. At most 30MB per message. Without attachmentTypes, every file comes with mediaId: null unless somebody has already opened it.
export type InboundMessageEvent = {
messageId: string // uin_messages.id - your idempotency key
threadId: string
fromAddress: string // lower case; '' when the channel had no address (a text, a call)
toAddresses: string[]
ccAddresses: string[] // who was copied in, as toAddresses; [] when nobody, or a channel with no Cc
subject: string // '' when there was none
bodyText: string // capped at 20,000 characters
sentAt: string // ISO 8601, the date on the message
attachments: Array<{
attachmentId: string
filename: string
mimeType: string
sizeBytes: number
mediaId: string | null // core Media row; null for inline parts (ignore them) and unstored files
}>
}
export type InboundMessageOutcome = {
links?: Array<{ moduleName: string; recordType: string; recordId: string; label: string }>
note?: string // one line on the message
}
export async function myInboundHandler(
event: InboundMessageEvent,
context: { signal: AbortSignal },
): Promise<InboundMessageOutcome | void> { /* ... */ }The rules:
-
Be idempotent on
messageId. The same message is offered again by the hourly catch-up when a run was cut short, and on purpose by the "offer again" button, which ignores what was offered before. Doing something twice on a second offer is your bug. The inbox claims each message before offering it, so offers of one message do not normally overlap, but a claim can expire under a run that died slowly: make the guarantee with a unique constraint andON CONFLICT, never by looking first and then inserting. - Five seconds. The signal is aborted when they run out and whatever you return afterwards is ignored. Do the cheap check inline (is this sender any of mine?) and queue anything slow - reading a PDF - for your own cron.
- Throwing is allowed and logged, and the next handler still runs. The message is still marked as offered: a handler that fails every time is not retried every hour.
- Return nothing for mail that is not yours, which is nearly all of it.
-
linksare stored as automatic links on the conversation (at most ten per handler). Only name records that exist. A link somebody has taken off that conversation is not put back. -
noteis folded to one line and cut at 200 characters. A later offer replaces your earlier line rather than adding a second; returning no note leaves the old one alone. -
Read the file by
mediaIdwith core'sdownloadMedia. Files of the kinds you asked for are stored as the message is filed (unless Attachments under Collecting is set to never fetch them). For older messages, the catch-up and the button fetch them from the mail server first, allowing eight seconds per file. A file that cannot be had in time is left out, no further files are fetched for that message (the slow fetch is still running and a second would open another connection to the same mailbox), and the message is offered anyway with those files'mediaIdnull. A message with no sender address gets nothing stored, having no library folder to file it under. A slow mail server is not the same as running out of time: when a run simply has no time left to fetch a file you asked for, the message is not offered yet - it waits for the next run, with any file already fetched kept, so you are not handed it without the file. (The one exception is a message whose files keep failing: it is offered with what there is rather than waiting for ever.) - When a message is offered: email, straight after it is filed, during the collecting pass that found it. Messages from other channels (forms, chat, texts) as they are copied in, if dated within the last three days. The hourly catch-up re-offers anything from the last three days never marked as offered; a module installed later sees those, and older post through the button.
Reply Catcher does a small part of what this does: it catches replies to contact form messages in your real mailbox and threads them back onto the enquiry. Unified Inbox does the whole job, for every address rather than for the contact form alone.
If you have both, do not point them at the same mailbox. Two things reading one mailbox files everything twice, in two places, and unpicking that is an afternoon nobody enjoys.
Cactus will not let it happen quietly. If Reply Catcher is configured against the same mailbox as one of your mail accounts here, that account is not collected and the settings page says so, naming the account and what to do about it. Reply Catcher was there first and is filing into a screen somebody is already using, so it wins until you decide otherwise. Mail left sitting in a mailbox is not lost; mail filed twice is a mess.
The tidy ending is to turn Reply Catcher's mailbox connection off, or uninstall it, once you are happy here.
Replies Reply Catcher has already caught are not imported. They stay where they are, on the contact form enquiries they belong to, and are perfectly readable there. Anything still sitting in the mailbox gets collected here in the ordinary way.
Choosing "remove code and data" takes every conversation, message, person and campaign with it - the do-not-email list included, so be sure before you do it on a site that has ever sent one. The attachment files themselves are not taken. They stay in your media library, in the Inbox folders they were filed into, exactly as they were - you keep every invoice and every signed order, which is usually the point. What goes is the record of which message each one came on, so from then on they are simply files in folders named after email addresses. Nothing is left claiming them either, so they will show up in the media library's Unused count, ready to be tidied away if you decide you no longer want them.
Said out loud rather than left to be discovered:
- Two people replying at the same time is not handled. Nothing warns you that a colleague is typing an answer to the same conversation.
- Labels of your own. Status, assignee and the inbox itself are what there is.
- Rules for filing by sender or subject. Filing is by the address post was delivered to, plus the whole-folder option and a catch-all.
- Live incoming text messages. Texts appear on the hourly check.
- Suggested replies. Nothing here writes anything for you.
Wiki: Home · Modules · Contact form · Live Chat · Twilio · Reply Catcher · Managing users · Managing media · Configuration reference