Studio FAQ and Troubleshooting

FAQ and Troubleshooting

Quick answers to the questions people ask most, and fixes for the messages you are most likely to see. Each answer links to the page with the full story. Messages are quoted exactly as Studio shows them.

---

Account and access

I never got the confirmation email. Check spam, and make sure you typed the address correctly. You can also sign in with Continue with Google for the same address. Still nothing: write to support@siamang-team.org. → Sign Up and Sign In

A confirmation or password-reset link opened the app instead of a form. You were already signed in, and the sign-in page sends signed-in people to the app. Sign out (or use a private window) and open the link again.

"Could not check your email. …" on the sign-in card. Studio looks up the address before it asks for a password, and that lookup failed. The message says why — for example "Too many sign-in attempts from your network just now — wait a minute and try again." (more than 10 attempts in a minute for the same address, or 20 from one network across all addresses; a busy office or campus network can reach that) or "The server could not be reached — check your connection and try again." Wait a moment and press Continue again. A failed lookup never sends you to Create your account: only a definite "no account" does.

"Your session expired. Please sign in again." Sign in again: Studio opens the screen you were on. Unsaved Builder and flow edits are kept as your server-side draft and come back when you reopen the editor.

Signing in or creating an account says the browser blocks cookies and site data. Studio keeps your sign-in in the browser. Allow cookies and site data for the Studio site (in a private window too), reload the sign-in page, and sign in again. If the message came after Create account, the account already exists: sign in with it. → Troubleshooting

I created my account from a colleague's invitation. Where am I? In your colleague's organization: the invitation is accepted as your account is created, and the invitation link takes you straight into the inviting organization. You also get an organization of your own, on its own trial; switch between the two with the workspace chip in the topbar. If the invitation page says "Could not load the invitation." followed by a reason, the server was busy or could not be reached — the link itself is fine; press Try again. "This invitation has expired. …" and "This invitation link is invalid or has already been used." are about the link itself: ask for a new invitation. The second one can also mean you were already added to the organization directly: sign in and open the link again, or pick the organization in the workspace chip. → If you create your account from the invitation

I landed in my own workspace, not my colleague's. After sign-in Studio opens your oldest membership. If you had an account before you were invited, that is your own workspace; switch with the workspace chip. → Organizations and Team

How do I create a second organization? Open the workspace chip in the topbar and choose Create organization (the Organizations screen has the same button). The dialog shows the address the organization will get, checked with the server so that it is free. The new organization starts on the Free plan with you as its owner, and the topbar shows no trial pill in it — the Pro trial comes once per email address, with the organization you got at sign-up. → Creating another organization

"Could not update profile. Name must be 200 characters or fewer." (or another "… must be …" message). The server refused a value you entered. The message names the field as the form labels it and says the rule it broke: a length ("must be 200 characters or fewer"), a required value ("is required"), a number range ("must be at most 730") or a format ("is not in a form Studio accepts"). Most fields stop at their limit as you type, so you rarely see these messages; correct the field it names and try again. → Limits and Quotas at a Glance

"Only owners and admins can create projects" (or rename a project, add a secret, run a connector). Your role in this organization is member. Members build, publish and analyze, but creating, renaming and deleting projects, adding or deleting secrets, running connectors, deleting responses, managing webhooks, deleting a library item another member saved and reading the organization's Activity are for owners and admins; those controls are disabled or hidden for you, and hovering a disabled one says who can use it. Ask an owner or admin, or to be made an admin. → Things members may notice

My trial ended. What changed? The organization is now on the Free plan, or on Plus if Plus or the Plus year was bought for after the trial. Nothing is deleted. On Free, surveys keep collecting within Free limits — 1,000 completed interviews per project, all its environments together, counting the ones it already has; you cannot add projects or members beyond Free's caps, schedules and live recomputation stop, and Saves with custom JavaScript or CSS can no longer be published. → Plans, Trial and Billing

"This workspace is frozen and read-only." Support has frozen the organization. Everything stays viewable and exportable; contact support to resolve it. (This is not what happens at the end of a trial.)

---

Building the questionnaire

My show-if / branch rule / piping does nothing in the published survey. Conditions, piping and quotas read a question's variable name, and the answer is stored under that name, so the question's Id does not matter. Check, in this order:

• An old build. A published survey keeps the runtime it was built with. Several things only work in a survey built with the current one: logic on a question whose Id differs from its variable (every preset, q5 / nps_5 — earlier builds stored the answer under the Id), conditions on a Matrix row, a MaxDiff or Conjoint task, or a per-choice variable of a wide Multiple choice (brands_1 = 1), {label:…} showing a label rather than a code, and piping in the title and body of Final, Screen-out and Redirect pages. Save (with nothing changed, Save re-validates with the current engine) and press Republish #N on the environment's card.
• A name inside custom JavaScript. Renaming a variable in the Builder updates the conditions, branch rules, quotas, piping, script targets and answers.<name> accesses that use it, but a string in custom code that merely holds the name is not changed — edit it yourself.
• The rule itself. Take the path in Test → Walkthrough: the side panel shows which conditions fired.

→ Question Id and variable name

"Question '…' has the id under which question '…' stores its answer". One question's Id is another question's variable name, so the survey could not tell which one a script or a Skip to means. Change that Id (Advanced → Id) or rename one of the variables, then Save. "Duplicate answer key in questionnaire: questions '…' and '…' both store their answer under '…'" is the same clash with a Matrix, MaxDiff, Conjoint or wide Multiple choice, which store their answers under their Id: change that question's Id. Studio now catches both before you save: the Id field shows "Another question already has this id." or "This is the variable … stores its answer under — the engine refuses an id that is another question’s variable.", and Validation → Structure lists the question. New questions never get such an Id; you meet this after typing one, or in an imported or older document.

"Question '…' stores its answer under '…', but '…' is also …". The question's Id differs from its variable and is a name the survey stores something else under: a Matrix row or another variable of a question ("…is also a variable question 'grid' stores an answer under"), another question's Other text ("…the key question 'brand' stores its “Other (please specify)” text under"), the arm an Assign to a condition writes, a codebook variable a custom script writes, or a name starting with __. Studio translates a script's answers["<id>"] to the question's variable, so a script meaning the other thing would reach the question instead. Change the question's Advanced → Id, then Save. The Id field and Validation → Structure say which name it is before you save. One exception: when the name is a codebook entry an earlier Builder left behind on renaming the question's variable, and an older script prefills the question by its Id (answers.q2 = …), delete that entry in Builder → Codebook instead — the message then says so too ("…or, if the codebook entry 'q2' is left over from renaming this question's variable, delete that entry…"). A new Id would leave the script writing the entry. → Names an Id may not take

"…the codebook still declares a variable "q2" that no question collects and nothing writes…" in Validation → Structure. An earlier version of the Builder left the old codebook entry behind when you renamed the question's variable or deleted the question. The Id is fine: delete the entry in Builder → Codebook, where it is listed unused. Deleting a question now takes its entries along, except those something still refers to or that label answers already collected; the Codebook lists those as referred to by … or answers collected, and the Save's UNUSED_VARIABLE warning for them is expected. → When a question is deleted

My variable label changed when I edited the question text. A variable's label follows the question — its text, a Matrix statement, a wide choice — for as long as it is the label Studio gave it, so a label that read "q8" becomes "How often do you drive?" as you type. One Undo restores both. Write a label of your own in the Variable section or the Codebook tab and it stays, whatever you change in the question afterward. → Labels that follow the question

I deleted a question, but its variable is still in the Codebook. Its entry stayed because something still refers to it — the Codebook's Used by reads "referred to by …" — or because the survey has been published and exports label the answers already collected under it (answers collected). The message after the deletion said which. Change what refers to it first if that is left over too, then Delete the entry; Studio asks before deleting such an entry. → When a question is deleted

A branch rule never fires. A rule with an empty condition never matches — it is not an "otherwise". + Rule no longer adds a rule until its condition is complete (Add rule stays disabled), but a rule can lose its condition (Clear) or arrive without one from an import. Studio says so under the rule ("Add a condition — an empty rule never fires."), on the Logic map ("no condition — never fires") and in Validation → Structure. Give it a condition, or use Default next for "everyone else". Also check the page for a Skip to: it is checked before the page's branch rules, so for anyone who answers that question no rule fires. → Logic and Branching

Respondents cannot get past a Matrix: "Please answer every row." Required on a Matrix asks for an answer in every row; a row answered "Not applicable" counts. Next marks the rows still empty. If only some rows matter, turn Required off, or turn on Offer “N/A” so that respondents can answer the rest with it. A survey published before this rule lets respondents on after one row until you republish it. → Matrix

Skip to jumps for every answer, not just one. That is how Skip to works (its hint: "on Next, after any answer to this question — checked before the page’s Branch rules"): when the question is answered, Next goes to the chosen page. For a jump that depends on the answer, use a Branch (next if) rule on the page.

How do I enter missing codes? In Builder → Codebook, open the variable's row and type each code followed by its label, separated by commas: -9 Refused, -8 Don't know (-7=Not asked works too). A code typed without a label borrows the value label for that code, or is labeled Missing (<code>). → Missing codes

I want one 0/1 column per option of a Multiple choice question. Options → Data layout → wide turns the question into one 0/1 variable per choice (brands_1, brands_2, …; a question whose options come from its variable's value labels gets them as its choices first) and keeps them in step with the choices; array turns them back into one variable. A published survey stores each chosen option's variable as 1 and the others as 0 once the question is answered (an option hidden by its own condition stays empty), exclusive choices such as "None of these" clear the others, and conditions and quotas on brands_1 = 1 work. A survey published before this worked needs to be published again; its earlier answers are read as 1/0 columns in Data and exports. → Multiple-choice layouts

What are the -66, -77 and <variable>_other values in my data? The codes of the answers added with a switch: "Other (please specify)" is stored as -66 in the question's column, with the typed text in <variable>_other; "None of the above" is -77; "Not applicable" is -1, declared as a missing code (where the codebook declares no such code, N/A is stored as the text na). The Inspector's hint beside each switch says what is stored ("stored as -66; the text goes to brand_other"), and the codebook labels the codes. A choice that already uses one of these codes pushes the added answer to the next free one (-67). Responses collected before these codes were written are read the same way (older exports showed __other__, __none__ or code / text columns instead). → Codes for Other, None of the above and N/A

How do I change the title of the completion screen? Theme → Respondent experience → Completion screen → Title (the default is "Thank you for participating") and Message. They are shown where the survey ends without its own ending text: on Submit of a question page, or on a Final page with no title (or body). A Final page's own Title and Body win, and the hints beside the two fields say so. Save and publish again. → Completion screen

"…references unknown variables: …" and the Save is errors. A condition names a variable that nothing in the questionnaire writes: no question collects it, no Assign to a condition script assigns it, and the codebook does not declare it. Usually it is a typo or a variable that no longer exists. Conditions may read the arm of Assign to a condition and variables declared only in the codebook (for example a value your custom JavaScript writes), but not URL parameters. → Assignment and "embedded data"

The Save badge says errors. Click it (it opens the Save in History) or open Builder → Validation. The questionnaire does not pass the engine's validation (for example a page nothing leads to, or a quota on an unknown variable), and such a Save cannot be published. Red lint findings such as EMPTY_PAGE leave the badge at warnings — the Save toast says "Saved #N — the questionnaire has 1 error (see Builder → Validation)" — and can be published after the confirmation "Publish #N with errors" → Publish anyway. A flow with errors leaves the badge at warnings too. → What blocks publishing

"Unreachable pages in navigation graph" or "Cycle detected in page navigation graph." Open Logic map → Pages: some page has no route leading to it, or routing loops back. Wire the page up (or delete it), or break the loop.

A condition says it "can never be true". It reads an answer given later in the interview. Move the question earlier or the condition later.

My page's Body text does not appear. A page's Body is shown on every kind of page except a Redirect page: on a page with questions, above them. It is HTML, not Markdown — Markdown marks such as ** appear as typed. A survey published before the Body was shown on question pages shows it there only after you publish it again: Save, then Republish #N. A text-only page (a Body and no questions) made in the Builder before this update showed respondents only its title: make any edit in the Builder — Studio then stores the page so that its Body is shown — and Save and republish.

Piping shows {answer:x} literally. The variable name is wrong (piping uses the variable name, not the question's Id), or the question has not been answered yet at that point. In a survey published before piping worked in the title and body of Final, Screen-out and Redirect pages, those pages show the placeholder until you publish again.

I renamed a variable and a flow broke. Renaming a variable in the Builder updates the questionnaire — conditions, branch rules, quotas, piping and scripts — but not your flows: a node parameter that names the old variable now names one the codebook does not have. Update the flow, or rename back. Once the survey is in the field, answers are stored under the variable name, so renaming a variable and republishing leaves you with two columns (see Two columns where I expect one below); on a published questionnaire the Inspector's Variable section says so: "This questionnaire has been published. Renaming a variable renames its column in the data: answers already collected keep the old name, answers collected after you publish again get the new one."

I cannot edit — "Name is editing". A colleague holds the edit lock; you are following their draft live. Take over claims the lock, or leave a comment. → Working Together

"Apply or revert your source edits before saving." You changed the JSON in More ▾ → Source. Press Apply (or Revert) there, then Save.

---

Publishing and fieldwork

The Publish button is disabled. The panel says why: the project has never been saved, the current Save has errors (its questionnaire does not validate), or the environment already runs that Save ("…nothing to publish"). A flow with errors does not stop publishing, and a Save with lint errors or warnings is published after a confirmation.

My survey was published before the latest Studio update. Do I need to do anything? Publish it again: Save (a Save with nothing changed will do), then Republish #N on the environment's card. A deployment keeps the survey runtime it was built with, so until then it lacks, among other things: quota cells that stop respondents, the closed / paused / full notice as the page opens, One per browser, partial responses and the drop-off funnel, the page Body above the questions, "About N minutes", the completion Title, the newer Wording fields, auto-height in the script embed, option shuffles that keep "None of the above" in place, a required Matrix that asks for every row, answering a Matrix row by row from the keyboard, answers that say they are keyed by variable name, so Studio stores them exactly as sent, fonts served from the survey host instead of Google Fonts, a captcha that loads only at submit and sends its token (an older build marks every response captcha: unavailable), and browser storage that waits for the respondent to start and lasts a week instead of 24 hours. Responses it already collected need nothing: Data, exports and flows read them in today's layout. The one exception is a quota cell on a Matrix row or on a per-choice variable of a wide Multiple choice: it counts only responses collected after the republish. → Older surveys and responses

"Custom JavaScript in the questionnaire is included from Plus…" / "Custom CSS in the theme is included from Plus…" Your plan does not include custom code in published surveys. Remove the script or the custom CSS, or upgrade. → Plans, Trial and Billing

I published but the link shows the old questions. Check that the card's #N is the Save you expect — press Republish #N if not — and hard-refresh the page (browsers cache survey files).

The published survey's progress indicator differs from the preview. Published surveys follow Theme → Progress (bar, dots, both or hidden; the bar when you never chose), like the Builder's previews: bar shows only the bar and its text, dots only the page dots, hidden nothing at all. Page labels and Progress text switch off the words above the page title and beside the bar. A survey published with an earlier version of Studio can show a different indicator until you publish it again — Save, then Republish #N. → Question style and progress

A quota cell is full but respondents keep coming. A full cell stops only the respondents whose answer falls into it, when they leave the page with that answer; everyone else goes on. If people with that answer still get through:

• the survey was published before quota cells closed — publish it again;
• you are looking at a preview: previews never check quotas;
• the count includes only completed interviews — screen-outs and unfinished interviews do not fill a cell, so a cell can show fewer than you expect;
• they left that page before the cell filled: each answer is checked once, when its page is left, and the submission does not check quota cells again, so a cell can end a little over its limit;
• the server did not answer the check within 4 seconds — the respondent is let through and checked again on a later page.

→ When a cell is full

The survey closed on its date. How do I extend it? Once the closing date passes, the card reads ○ Closed ("Closed — deadline passed date") and anyone who opens the link sees "This survey is closed". Press Extend on the card, pick a later date and Extend to this date: the survey collects again at once, without a new Save or a rebuild. The date can come from the questionnaire's deadline, the environment's closes_at in studio/settings.json (the earlier of the two wins), or the Closing date chip. → Set or extend a closing date

Respondents answered everything and then saw "This survey is paused" / "This survey is closed" / "Thank you for your interest…". The survey page checks as it opens whether the environment is paused, closed (by Close or its closing date) or full (a response cap reached), and shows the notice at once. Someone who already had the survey open when that happened meets it when they submit, and their interview is not stored as completed. Resume, extend the closing date, or raise the cap (new projects cap main at 1,200 and pilot at 50 completed interviews; on Free the project stops at 1,000). A survey published before the check as the page opens shows these notices only at submit until you publish it again. → Paused, closed and full surveys

The count is not moving. Check you are looking at the right environment, that it is not paused or closed, and that interviews are being completed. With a cap, the Responses tile and bar count completed interviews only — partials and screen-outs show in Data but not there (the tooltip gives the total of all rows).

The build failed. Open Build log on the card. If a republish failed, the previous version is still serving the link. Fix the cause, Save, and publish again. A failed card with no earlier version still live behind it offers Deploy current Save #N (Retry Save #N when the failed build was already the current Save).

Access codes are not asked for. Generating codes creates a new Save; republish the environment.

The header "Preview" survey shows "Preview — answers are not stored". That is what a preview build is: it never stores answers, says so in a banner at the bottom, and ends on the survey's normal completion page. A preview built before this banner existed ends with "Submission failed" instead; Preview again to rebuild it. To test real submissions, use the pilot environment. → Testing Your Survey

Where do I get the panel ids to reconcile with Prolific / Cint? From Outcomes · reconcile with the provider at the end of the Panel chip on a live environment's card: its completed, screened-out and partial CSVs list each respondent's provider id, for every response of that outcome. The id parameter is matched however it is capitalized (PROLIFIC_PID, RID). → Reconciling outcomes

---

Data

The numbers in Data and in my flow disagree. Data counts every row — all environments, partials, screen-outs. Flows usually filter (environment, Only completed responses, dedup, speeders, filters). Run to each node and watch the row count.

I cannot find a response in the grid. The grid loads the newest 100 rows of a table (25 per page), and typing in Filter loaded rows… narrows those rows only. Press Enter (or Search all rows) to search every row of the table on the server: a row matches when any of its values contains the text — an answer, the response id, the respondent_id, or a panel id or invitation token in meta. The note then reads "N rows of the whole table match “…”", with the newest 100 matches loaded; Clear search goes back. To delete the row you find, see Handle a data erasure request.

Where are the URL parameters and durations in my export? In columns of their own, after the answers: url_<name> for each link parameter (up to 8), duration_s, started_at, captcha, tab_switches, hidden_seconds and pastes — in every format of Export ▾, and in flows under the same names. The Data grid itself still shows them together in the meta column. An export you downloaded before this change has no such columns; export again. → Data Exports

My matrix answers changed from 1–11 to 0–10. A Matrix answer is its column's code from the codebook: a 0–10 scale stores 0–10. Surveys built before this stored the column's position (1–11); Data, exports and flows now read those older responses as the column's code too, so old and new responses sit on one scale. An export or a flow result you made before therefore differs from one you make now. Surveys published before the change store positions until you publish them again, and are read correctly either way. → Older surveys and responses

SPSS labels are missing or look wrong. Exports are labeled with the current Save's codebook. Fill in labels in Builder → Codebook, Save, and export again.

Two columns where I expect one. Usually a variable was renamed mid-fieldwork. Another cause: a column named like a question's Id (q5) next to its variable (nps_5). Earlier versions of Studio stored answers under the question's Id; Studio has since moved those answers to the variable name, except where the move could have mixed two questions' answers — those stay under the Id. Either way, harmonize the two in a flow with Recode or Derive rather than editing the raw table.

Can the same person answer twice? Why is there no "respondents" count? Studio does not identify people, so every interview is a response of its own. The totals on Data → Insights and Live therefore read responses (every row: completed, screened-out and partial) and completed (submitted interviews that did not end on a screen-out page — what quota cells and response caps count). To refuse a second interview from the same browser, switch on One per browser on the environment's card; for one answer per person, use email invitations or a panel's own checks. → Accept one response per browser

There is no Delete button on the rows. Only owners and admins can delete responses, so only they see Delete on the rows of the responses table. Ask one of them to handle the request.

I deleted a response. What else changed? The response counts and any quota cell it had filled go down: the survey's cells are recounted from the responses that remain. The deletion is recorded in the Activity log, without the response's content.

---

Flows and reports

"Could not start the preview. The preview limit for this hour is reached on your plan." Run to here is limited per person per project per hour (Free 30, Plus 120, Pro 600). Wait, or use Run. "…A preview is already running" — wait for it to finish.

A flow shows a red "errors" pill and its Run button is disabled. The flow did not pass the engine check at the current Save, so it was saved without code and cannot run ("Fix this flow's errors and save first"). The pill's tooltip lists the errors node by node, and opening the flow shows them in a banner. Only that flow is affected: the questionnaire can be published, and the other flows, Run all and previews work — in Run all it fails with "flow 'name' did not pass the engine check at Save #N, so it has no script to run: open it, fix its errors and save". Press Check, fix what it lists and Save. → A flow with errors

"edges/N/from: Additional properties are not allowed ('title', 'type' were unexpected)". An earlier version of Studio saved an output added with + Add output in the Report view in a form the engine rejects. Open the flow: Studio repairs its connections and says "Studio repaired this flow's connections." Click Save changes, and the flow runs again. → A flow with errors

In what order does Run all run my flows? In dependency order: a flow that reads a table another flow writes (through a Project table node, or a Responses node set to that table) runs after the flow with the Write table node. Flows that do not depend on each other run in alphabetical order of their names; you do not need to name flows so that a writer sorts first. The flows table, the pipeline strip and the Run flow dialog list the flows in that order. Flows that read each other's tables get an amber cycle pill and run one after another in alphabetical order. → Run all

Run all failed, but some flows ran. One failed flow does not stop the run. The others still run, except those that read a table the failed flow did not write — their log line reads "skipped: needs flow, which failed". The run ends as failed (its last log line, "failed: names", lists them), and View logs lists every flow as ok or failed, with each error. The flows that succeeded keep their reports under Reports, and the combined report is written anyway, titled "Combined report (incomplete)": its first section, "Missing from this report", names each missing flow and why. Fix the flows it names and run all flows again for the complete report.

The Schedules section shows "Requires Plus" instead of "Schedule a run". Schedules are available from the Plus plan ("Schedules are available from the Plus plan"); on Free the button opens the plans. Run flows by hand, or upgrade. → Schedules and Webhooks

How do I delete, rename or duplicate a flow? With the ⋮ at the end of its row on Flows (Rename…, Duplicate…, Delete…) or the editor's More ▾ (Rename flow…, Duplicate flow…, Delete flow…). Each is a Save of its own, so restoring an earlier Save undoes it. A rename moves the flow's schedules and comments; a delete pauses its schedules; past runs and reports keep the old name. → Rename, duplicate or delete a flow

"flow has unsaved changes. Open it and save them (or undo them) first …" A rename takes the flow as it was last saved, so Studio refuses it while you have unsaved changes to that flow. Save (or undo) them, then rename.

My Data file node cannot find the file I uploaded. Choose it by name in the node's File list, or upload it from there with Upload…. Files may store a file under another name than it had on your computer (Опрос.csv becomes Opros.csv); the Upload dialog says the stored name before you upload. A value saved earlier as a path, such as ./assets/panel_wave2.csv, shows under "Other location, kept as it was written" — "A run only brings in files uploaded under Files, so it won't find this one." — with Use the uploaded panel_wave2.csv when Files has it. An upload deleted since shows with "(not in Files)" after its name. If the run log says "note: assets/name is not among this project's Files", the name is wrong or the file was deleted. → Where files go · Files

The topbar and the project tabs are gone in the flow editor. Focus mode is on: it gives the Canvas view the whole window. Press F (outside a text field or a list field of the inspector, where a letter picks a value), or Esc once nothing is selected and no list is open, or click Focus mode under the canvas's zoom buttons ("Exit focus mode (F or Esc)"). List and Report always show the topbar and tabs, and so does every screen outside the editor; the Save badge sits beside the flow's name while the mode is on. Studio remembers the choice in this browser. → Focus mode

The analyses below my Data file do not offer the file's columns. The file is read when it is uploaded, and again when the node's reading options change; meanwhile the node's Columns panel says "Reading the file's columns… Nodes below it aren't checked for names until it is read." Then the pickers below list the file's columns under "From name". A File under "Other location" is not read at all: its Columns panel says not read — "Only a file uploaded under Files is read for its columns. Choose the upload under File above (upload it first if it isn't there), and its columns show here." — and has no Read again. Click the field's Use the uploaded … (for an older path) or Use name (for a typed name), or choose the upload from the list. If the panel says cannot read, fix what it names (the encoding, the delimiter, the sheet, the header rows) under Reading options and click Read again. The questionnaire's variables are offered below a file only when the file is this survey's data; if it is and the panel says the file keeps its own labels, set Codebook to questionnaire. → The file's columns in the flow

My CSV opens as one column, or its letters come out garbled. Excel with Russian or most European settings saves "CSV" with ; between the fields, a decimal comma and, for Cyrillic, Windows-1251; "Unicode text" is UTF-16 with tabs. A Data file node finds all of that by itself with Delimiter, Encoding and Decimal mark on auto, and its Columns panel says what it found ("CSV · semicolon-separated · CP1251 · decimal comma"). The code page is told by how the file's letters read as words, and a note says so when it was a guess: "The file is not UTF-8: it is read as …, the code page its letters read best in. If names or answers look garbled, set Encoding (or save the file as CSV UTF-8)." If it still reads wrong, set the option under the node's Reading options. → Data file

Numbers with a decimal comma (4,5) come in as text. Decimal mark on auto reads a comma where the file's numbers are written so — with ; or tabs between the fields, also in a column whose first decimals come late. In an Excel sheet, a column of nothing but values like 1,500 (one and a half, or fifteen hundred?) stays text, with a note in the Columns panel, until you set Decimal mark to , or .. In a text file, auto always settles on one mark. In a file with , between the fields, a comma is never a decimal mark.

How do I analyze the data I collected in Qualtrics? Import the survey's .qsf in the Builder and Save; export the data from Qualtrics with Use numeric values; upload it and read it with a Data file node. The node finds the export's header rows by itself, the questionnaire labels the answers, and the Columns panel offers to leave out Qualtrics' personal-data columns (IPAddress, RecipientEmail …). → Bringing in Qualtrics data

Files says "possible missing codes", or the Columns panel lists codes such as -7, -8, -9. Many surveys store "don't know", "refused" or "not asked" as negative codes (-7, -8, -9, -99) or as 97–99 far above the real answers. Until they are declared, they count as answers: they show up in frequencies and pull a mean down. Click Mark them as missing in these columns in the node's Columns panel — it writes each code only for the columns it was found in, one column a line (q5: -9, q6_1: -7, -8), so a -7 among a column's amounts or an age of 99 stays a value — then add a Missing values node below to turn them into blanks. A list for every column (-7, -8, -9) also works, but makes a code missing in columns where it is a real value (a negative code is left alone in a column of other negative amounts). The panel only suggests; check the codes against the file's documentation. → The Columns panel

I chose another file, and Missing codes still names the old file's columns. Codes for columns a file does not have would do nothing, so the field says "Missing codes name columns this file doesn't have: …" — click Remove them. If the new file has codes of its own that look like missing codes, the panel offers Add this file's missing codes (n columns).

"column name … is longer than 63 bytes, which a table column cannot be" from a Write table node. A project table keeps a file's column names as they are, up to 63 bytes each — a Cyrillic letter takes two, so about 31 Russian letters. The node stops before writing anything, in a preview as in a run. Shorten the name in the file and upload it again, or leave the column out with a Select columns node before the Write table.

Can Run to here change a project table? No. A preview never writes project tables: a Write table node it reaches says "Not written: a preview never writes project tables. A run writes N rows to table 'name' …", and the Live tiles are not touched. Only a run of the flow — Run, Run all, a schedule — writes the table.

Is there a t-test? Which test does Group means run? Yes. The t-test node compares two groups (Welch's by default, or Student's), two answers of the same respondents (Design paired) or one mean against a value (one_sample), with t, df, p, the mean difference and its CI, and Cohen's d. Group means chooses a test for you with Test auto — Student's t-test or a one-way ANOVA for interval data, Mann-Whitney or Kruskal-Wallis for ordinal — or runs the one you name: student, welch, anova, welch_anova, mannwhitney, kruskal — and after anova, welch_anova or kruskal, a post-hoc comparison of every pair. Rank tests are also in Compare groups (with Dunn's test), paired ones in Paired tests (Wilcoxon, McNemar, Friedman, Cochran's Q), Fisher's exact test in Crosstab, and Pearson, Spearman or Kendall in Correlation and Correlation matrix. For an ordered outcome there is Regression with Model ordinal. → Which test

"Tukey's HSD follows a one-way ANOVA — set Test to anova, or Post-hoc to none." A post-hoc test belongs to one test: Tukey's HSD to anova, Games-Howell to welch_anova, Dunn's test to kruskal (in Compare groups, to kruskal or auto). Change Test or Post-hoc as the message says. It is an error: the flow cannot run until you do. A warning such as "Post-hoc is not run while Significance test is off." only names a setting the node would ignore. → Rules between parameters

A field I filled in disappeared from the inspector. The node does not read it with its current choices — a t-test's Groups once Design is paired, TURF's Portfolio unless Search is fixed. It is kept, and named under the other fields: "Not used with these choices, and kept for when they apply: …", with Clear it. It is not checked while it is not read, so a stale value there never stops the flow. Switch the choice back and it is used again. → Parameters and variable pickers

Group means with Test auto and with kruskal give different Kruskal-Wallis results. The codebook's missing codes. A test you choose by hand leaves them out ("Missing codes left out = …"); auto counts them as answers, as it always has, so that stored flows keep their numbers — and says so: "Missing codes counted as answers = Trust: Acme: 44 (9 = Refused); run Missing values first to leave them out". Put Missing values before the node and both agree. The same holds for Compare groups with and without Dunn's test, and for Correlation with spearman and with the other methods.

The test line reads "Test = not run: …". The data cannot carry the test, and the rest of the line says why and what to choose instead — for example "Welch's t-test (unequal variances) compares two groups and Region has 3 — choose anova or welch_anova". A group of one, or answers without any spread, give such a line too, rather than a number.

"Gender has 3 answers (…); a t-test compares two — name them in Group A and Group B, unless the data this node reads holds only two of them." / "Gender has 3 groups (…); a t-test compares two — name them in Group A and Group B." The first is a warning of the Checks block, before any run: Groups has more than two answers in the codebook and neither Group A nor Group B is named — "Gender has 3 answers (1 = Male, 2 = Female, 3 = Other); a t-test compares two — name them in Group A and Group B, unless the data this node reads holds only two of them." It does not stop the flow, since a filter upstream may leave two. If none does, the run then stops with the second, "Gender has 3 groups (1 = Male, 2 = Female, 3 = Other); a t-test compares two — name them in Group A and Group B." A t-test compares two groups. Pick them in Group A and Group B (the lists offer the answers of Groups), or use Group means with anova or welch_anova for all three. A multiple-choice Groups is refused because its groups overlap: run Explode multiple choice and compare by one option's 0/1 column.

A Result chart says "Kind 'scree' does not suit the table output of …". A Result chart draws what the connected output holds: a Group means table has means, not eigenvalues. The message names what that output draws, and when another output of the same analysis draws the kind you chose it says so — "Kind 'scree' does not suit the loadings output of Principal components (p), which draws 'loadings'; its variance output draws 'scree'." — so connect that one. Leave Kind at auto and the chart draws the first kind the result suits; the inspector lists them under the field and marks the others "(not for this result)". → Result chart

"A Result chart cannot draw the table output of Frequencies (…)". A Result chart draws the results of analyses — means, tests, TURF, MaxDiff, factor analyses, regressions, key drivers, maps, prices (the message lists them all). A frequency table is a distribution: draw it with a Bar chart (Show percent), from the data. "The stat output of Regression (r) only tells a chart its weight and base; connect the output it draws, table, too." means the same: connect the analysis's table as well.

A Likert chart says "The items of a Likert chart must share one scale, and these do not: …". A diverging bar needs one middle, so every item needs the same answers — the same value labels in the codebook, missing codes aside. The message names both scales. Draw the items of each scale in their own chart, or recode them onto one scale first; recode a scale that runs high to low (1 = Strongly agree) so that it runs low to high. A multiple-choice item, or items with no value labels, no whole-number valid range and no Likert scale question behind them, are refused the same way. → Likert chart

"Split by needs one answer per respondent, and … allows several: …". A Bar chart split into groups needs each respondent in one group. Draw the multiple-choice question as the Variable and split it by a single-answer question instead, or split by one of its options after Explode multiple choice. A multiple-choice Variable can be split, but its options overlap, so they are drawn side by side (Layout grouped), never stacked. → Bar chart

My Bar chart looks different since I changed Sort (or Show). At the defaults — Show count, no Split by, Sort code, no Top N, Confidence intervals off, a bar layout — the Bar chart is the one it has always been. Any other setting (a histogram and a donut too) draws the newer chart: its own colors, the base and notes under the plot, and the codebook's missing codes left out (the classic chart draws a "Refused" as a bar). Set them back to their defaults for the classic chart.

My Bar chart shows no significance letters. Letters are drawn on a chart split into groups, in percentages, side by side: Split by set, Show percent, Layout grouped — the rules on the node say which is missing ("Significance letters compare the groups of Split by — set Split by.", "… compare percentages, as the Banner table's do — set Show to percent.", "… are drawn on bars side by side (Layout grouped), not on stacks."). With all three, the note under the chart says why a bar has none: "No group's share of any answer is significantly higher than another's.", or "Not tested, fewer than 30 respondents who answered: …" for a small group. → Bar chart

The letters on my Bar chart differ from the tab book's. The comparisons are the same; the letters are not. The chart letters the groups of its Split by A, B, C, …; a tab book letters every column of its banner in one run — A–C for gender, then D–F for region — so its letters match the chart's only when Split by is the banner's first variable. A Banner table can differ in its numbers too: it counts a missing code as an answer and tests on everyone in a column, while the chart and the tab book leave missing codes out and test on those who answered.

"bins: The bins' edges must increase from one to the next, …". A histogram's Bins is auto, a number of bins from 1 to 100, or the edges in increasing order, separated by commas: 18, 25, 35, 50, 65, 100. The field suggests all three as you type. A histogram of a nominal or ordinal question is refused ("A histogram draws the distribution of a number, and Region is nominal: draw its answers as bars (Layout = grouped).") — draw it as bars. → Bar chart

"Age is a number with 84 different values given, and this chart draws a bar for each: …". Percent bars, a split or a donut draw a bar or a slice per value, and refuse a number of more than 30 values. Set Layout histogram, cut it into ranges with Bands first and draw the bands, or set Top N to 30 or fewer to draw only the values given most (a Top N over 30 says "… and top=40 draws a bar for each of the 40 given most: give top=30 or fewer, …"). The check warns before the run when the codebook's valid range holds more than 30 whole numbers ("Age is a number of up to 84 values, and bars draw each value given: …"); it does not warn of the chart of the defaults (Show count, no Split by, Sort code), which draws every value as it always did.

My Trend has no created_at to pick. It is at the end of the Time list, under the heading Beside the answers, with updated_at and started_at — the timestamps the survey's responses carry, which the codebook does not list. (There is no submitted_at: Studio's responses do not have one, and the check calls it an unknown variable.) A flow that reads an uploaded file or simulated data has them only when the data has those columns: the node warns ("time: "created_at" is a timestamp of the survey's responses, and this node reads Data file: the run stops unless that data has a column created_at. …"), and the run stops with "The data has no column 'created_at' to read Time from." when it does not. → Trend

"created_at spans 905 days: too many points for one chart. Choose a longer Period." A Trend draws at most 500 points. Two and a half years by day is more; by week or month it fits. A Time of more than 500 different codes ("… has 731 different values: not wave codes. …") is not a wave variable — pick the wave's variable or a date.

My Trend's points are hollow, or the bands are gone. A point with fewer respondents than Minimum base (30) is drawn hollow and without its band, and its table row says "base below 30"; the table still gives its interval. With more than four lines the bands would hide one another, so none is drawn ("No bands: the 95% intervals of 5 lines would hide one another; the table gives each point's.") and each line's points take a shape of their own. Use a longer Period, fewer groups, or the table.

"9 (Refused) is a missing code of Trust: Acme, not an answer: …". A Trend's Answer codes are answers; a missing code is left out of every base. Check the answers you track — 4 and 5 for a top-2 box. The checklist does not offer missing codes; this appears for a code typed or stored before.

Where is my tab book? A run of the flow writes it under the node's File name and keeps it as outputs/<flow>/<name>.xlsx — a Tab book added from the palette is named <flow>_tabbook — on the Reports screen (with an Excel button), in Files and on the run's card. The field says where: "After a run: Files → outputs/flow/name.xlsx". A preview runs the node but keeps no file ("Not kept: a preview never keeps the files nodes write. …"). A path saved earlier outside outputs/ is not kept at all; the field shows it under "Other location" with "This file isn't in outputs/, so a run doesn't keep it under Files." and a one-click fix. → Tab books

My tab book left a question out. The Contents and Notes sheets say why, as does the node's statistic (Skipped): with Questions empty, open answers and rankings are left out, and so are interval and ratio questions (name them for their means); a question nobody answered, or one of more than 30 different answers without answer labels that is not a number, is left out whatever you name. Check the question in Questions, code an open answer first (Code open answers), or derive a ranking's first choice (Derive). → Tab book (Excel)

"… holds multiple-choice answers, and a banner column is a group of respondents that no one else is in. …". A tab book's banner columns must not overlap. Run Explode multiple choice and use the 0/1 column of each option you want as a banner variable, or choose another variable.

My charts do not take the report's colors. Only a chart whose Palette is theme (a Heatmap's Color map theme) takes the Chart colors of its Save report's Look; a named palette keeps its own. A chart in two reports takes each report's colors in that report. On its node's preview and on a Live tile it is drawn in the flow's Save report look — in the default colors when the flow has no Save report, or several with different looks. The house style reaches a flow only when it is stamped into its Save report node (Use the house style, Apply to every flow). → Chart colors

"theme: chart_text_color: '#cccccc' on the charts' white background has a contrast of 1.6:1; …". The Look's chart colors must read on white: text at 4.5:1 or more, a bar or a line at 1.3:1 or more. Pick a darker shade; a color checker shows the ratio. The node names each problem — a color that is not hex (#2a78d6), fewer than 2 or more than 12 series colors, a color given twice, two equal diverging ends. → Chart colors

Regression says "Region is nominal: its answers (…) have no order, …". Model ordinal is for ordered answers — dissatisfied to satisfied. A nominal outcome has no order, and the model would take one from its codes. Use logit for an outcome of two answers, or make an ordered variable with Recode (Scale ordinal) first. The check says so before the run; for a variable a node of the flow makes nominal, it is a warning and the run refuses it. → Regression

Where are my report's tables in Excel? Check Also Excel in the Report view (the Save report node's Also save tables to Excel) and run the flow (or Run all): <report>.xlsx is written beside the report — on the Reports screen under Excel, in Files, and among the run's outputs. A run without the box does not delete an earlier one, so the Excel button then offers the workbook of the flow's last run that wrote it: check the box and run again to bring it up to date. → Tables in Excel

My research bundle stops at a t-test, Export file, Result chart, Trend or Tab book node. The bundle's engine pin may lag behind Studio: at the last merged upstream commit, the new statistics nodes, Key drivers, Perceptual map, Price sensitivity, the Likert chart, the Result chart, the Trend and the Tab book (Excel), every Export file, Code open answers and TURF node, a Bar chart with Show percent, Split by, Sort value, Top N, Confidence intervals or Layout histogram or donut, a Heatmap with Pearson or Kendall, a chart whose Palette (a Heatmap's Color map) is theme, a Save report whose Look names chart colors, Regression's ordinal model, Save report's tables in Excel and a test chosen by hand stop with an error. The bundle's README says so; install the engine revision it names. → Installing the engine

The report is empty or missing. A report needs a Report section connected to a Save report node, and the flow must have run. The node's File name field keeps the report in outputs/; a path saved earlier outside it shows as "Other location", with a fix. Run all keeps only the report that is the flow's Report path; another Save report is kept by a run of that flow on its own, and its field says so. → Reports

"The table clean_responses does not exist yet: it is written by 1. Clean raw responses. Run that flow (or Run all) first, then this one." The flow reads a project table (a Project table node) that no run has written yet — a preview never writes one. Run the flow the message names, or Run all, which runs it first; then preview or run this flow again. A new project started from the example study is in this state: its flows after cleaning read clean_responses, and segments reads scored_responses, which wellbeing writes. See Tables between flows.

A flow is missing from the combined report, or Run all says "report … was not written". The combined report takes each flow's Report path (flow settings), a list of the flow's Save report nodes. A flow with — none — is left out. A path the flow does not write — saved earlier, or set through the API — fails that flow, and the field shows it with "(no node saves this)" after it: "report outputs/x.md was not written: the flow's Report path names a file none of its nodes saves — choose one of its Save report nodes in Report path". The flow's tables were still written, so the flows that read them run, and the combined report is marked incomplete. The Save warns about this beforehand (REPORT_PATH_UNWRITTEN: "The flow's Report path is “…”, but no Save report node saves there: Run all will fail this flow. Choose one of the flow's Save report nodes in Report path, or clear it."). Choose one of the flow's reports in Report path, or — none —. → The combined report

Live tiles are stale. Tiles show the flow's last completed run, or a newer Run all that ran it; a failed run leaves the previous tiles in place (check "updated …"), and previews never change them. On Free, press Recompute now; automatic recomputation is Plus and above and needs Live: recompute on new responses checked and saved. A failed automatic recompute does not email the owners. → Live Monitoring

The public live link shows "page not found". It was revoked or rotated, or the organization's plan no longer includes Live.

---

Coding open answers

Where is Code open answers… — coding with the AI assistant? Switched off on this platform. It sent the text your respondents wrote to the AI provider, so its buttons (Code open answers…, Code more answers…) and its dialog are gone. Build the codeframe yourself: on the Code open answers node, New codeframe… (or Files → Codeframes → New codeframe…) opens the codeframe editor, where you code answers by hand and write word rules that code the rest. Codeframes the assistant built earlier keep working. → Coding Open Answers

"AI coding of open answers is switched off on this platform." The API's answer to a request to start a coding job, and the log of a coding job queued before the switch (in Run history). Nothing was read or charged. Use the codeframe editor instead.

An answer I expected to get a theme is uncoded. Type it into Test a phrase: it shows each rule that fired, was vetoed or met a negation. The usual reasons: the word is negated (wasn't late — late finds only mentions that are not; not_late finds the negated ones); a Must also contain word is in another clause (a comma or but ends one — set Reads to the whole answer); the term misses a word form (delay does not find delayed; write delay*); a But not word is there; or the answer is in another language than the term. Add a term, or code the answer by hand. → Writing rules

"2 errors — the codeframe cannot be applied or saved until they are fixed." The list under it (and the mark beside each term or setting at fault) says what: "the codeframe has no themes" for a new one — Add theme; a term beginning re: — "regular expressions are not supported: …"; a theme without a label, the same words replaced twice. Save changes stays disabled ("Fix the codeframe's errors first") until they are gone. → Mistakes the editor catches

"This codeframe is version 1: a theme for each answer it was built from, and no rules. …" An older codeframe — built by the assistant, or the example study's. The node applies it as before. Editing it makes it version 2: its decisions are kept, its example answers are left out (version 2 keeps only fingerprints). → Older codeframes

The themes add up to more than 100 %. The codeframe gives an answer several themes, and every % is of the respondents who answered; the table's Percentages says so. A net's row counts each respondent once. → The table

The theme table and a Frequencies of the theme variable give different numbers. The theme table counts everyone who answered, with No theme and Uncoded rows. In the theme variable, an answer coded as no theme and an uncoded one are empty, so a Frequencies of it counts only the respondents with a theme. Quote the theme table for shares of those who answered.

A Split by, a donut, a banner or a Likert chart refuses my theme variable. The codeframe gives several themes an answer, so its theme variable is multiple-choice, and these refuse it as they refuse a multiple-choice question. Use it where a multiple-choice question goes (a Frequencies, a Crosstab's rows, a Bar chart's variable), or turn off An answer may have several themes.

The nodes after Code open answers do not offer its theme variable. The codeframe is not saved yet: a codeframe started from the node is kept only in its browser tab until you save it in its editor, and the node says so ("path is not saved yet: it is kept in this tab until you save it in its editor."). Save it, and save the flow.

"analysis/… is not saved in this project: it was started and never saved. Start it again, or choose another codeframe." The node names a codeframe that was started from it in a tab that is now closed, and never saved. New codeframe… starts it again under the same name, or choose another codeframe.

Answers that just arrived are not in the codeframe editor. The editor reads the answers again once its last reading is a minute old. Check Answers from and Only completed responses too: they choose which responses it lists.

---

Email invitations

"Import contacts" or "New mailing" is disabled. Invitations need a paid plan (Plus and above) — during an unpaid trial they stay locked — plus a published, live environment and at least one subscribed contact. An environment that has closed by its closing date is not offered. If the button's tooltip reads "Mailings are paused for this organization", see the next answer. → Email Invitations

"Mailings are paused for this organization — …" Mailings were paused automatically after too many bounces or spam complaints; the panel names the reason and adds "Contact support to resume them." A mailing that was sending when this happened reads "Paused — reason; the rest stays queued until support resumes it." Contact support.

Who gets a reminder? Remind goes to the invitees of that mailing who have not completed — through the invitation or any earlier reminder — so someone who answered through a reminder's link is not reminded again, and a second reminder is safe. Its tooltip and the reminder dialog give the number, and the preview shows the message as its first recipient will get it ("Preview · as email"). The mailing's Completed column adds "+N via reminders".

---

Integrations

My webhook does not fire for some events. A webhook receives only the events selected under Events when it was added — Deploys: live, failed, stopped; Runs: completed, failed — and every event when none is selected. Webhooks added with an earlier version of the chips (deploy, run) have been switched to the matching events and now receive them. One that subscribed only to terminal shows an amber pill "never fires: terminal" — nothing emits that event, so delete the webhook and add it again with the events you want. An unsigned pill means its deliveries carry no X-Siamang-Signature header; add it again with a secret to sign them. Check Recent deliveries for errors from your endpoint. Webhooks are (Plus), and only owners and admins can see and manage them: for members the Webhooks card holds only that notice. → Schedules and Webhooks

Can a webhook post to Slack? Yes. Add the Slack incoming-webhook URL as the endpoint: every payload carries a one-line text (for example deploy.live · acme/pulse (main) · <link>), which Slack posts as the message. → Slack and other chat tools

Saving a connector fails with "…String should match pattern…". Connector names may contain only lower-case letters, digits and _, and must start with a letter. → Connectors

Can I use a project secret in a flow? No — as the Secrets tab says, connectors and repository deposits read them by key; flow runs cannot, and they have no network access anyway.

The Zenodo DOI starts with 10.5072 / points to sandbox.zenodo.org. Use sandbox.zenodo.org is checked by default in the Deposit dialog. Deposit again with it unchecked and a production Zenodo token. → History and Versions

---

Still stuck?

• Profile → Support links to documentation, examples, the issue tracker and feature requests, and Contact team writes to support.
• Email support@siamang-team.org with the organization slug, the project slug and the Save number — that is usually enough to reproduce anything.

See also

• Recipes
• Limits and Quotas at a Glance
• Glossary
• Key Concepts

---

← Keyboard Shortcuts · Studio contents · Glossary →