Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closeout for sign-in confirmation #193

Merged
merged 5 commits into from Oct 3, 2016
Merged

Conversation

@vbudhram
Copy link
Contributor

@vbudhram vbudhram commented Sep 22, 2016

This PR adds result and dashboards from sign-in confirmation.

Ref: mozilla/fxa-features#3

@vbudhram vbudhram self-assigned this Sep 22, 2016
@vbudhram vbudhram added this to the FxA-83: signin confirmation milestone Sep 22, 2016

<img src="sync-sign-in-success-rate.png" height="300">

This meets our original goal of not effecting this metric.

This comment has been minimized.

@rfk

rfk Sep 23, 2016
Member

nit: "affecting"

@vbudhram
Copy link
Contributor Author

@vbudhram vbudhram commented Sep 23, 2016

@davismtl @rfk Took a shot at the results from sign-in confirmation. Was there anything else you think could be added?

Copy link
Contributor

@davismtl davismtl left a comment

What would be your overall conclusion?
Are there any next steps that you would like to call out at the end based on the conclusion?

Pre sign-in confirmation metrics show that we had between a 42% to 52%
sync sign-in success rate. From [first dashboard](https://app.datadoghq.com/dash/163668/fxa-content---signin-confirmation), with sign-in confirmation, we have maintained this sign-in rate.
Pre sign-in confirmation metrics show that we had
between a 42% to 52% sync sign-in success rate.

This comment has been minimized.

@davismtl

davismtl Sep 23, 2016
Contributor

I have trouble with this metric when I look at the details in datadog.
(fxa.content.signin.signin.success + fxa.content.force_auth.signin.success + fxa.content.complete_signin.verification.success) / (fxa.content.signin.submit + fxa.content.force_auth.submit) * 100
Why are there 3 different login success events?
Why are there 2 different form submit events?

I've tried to piece things together (by reading this) but it's a bit unclear to me and some events are missing I think.

This comment has been minimized.

@shane-tomlinson

shane-tomlinson Sep 26, 2016
Member

Why are there 3 different login success events?
Why are there 2 different form submit events?

This is the most common question we have about the metrics and why I proposed mozilla/fxa-content-server#4163.

Most reported event names contain a built in screen name. This is so we know on which screen an event originates. Logging success without any additional context is a pointless metric - "what kind of success?" is the obvious question that follows. Way back when, there was only 1 way to sign in, from the /signin screen. At that time it was sufficient to have fxa.content.signin.success, we knew what signin.success meant and where it was coming from. Users can now sign in from 6 different endpoints: /signin, /force_auth, /signup, /oauth/signin, /oauth/force_auth, and /oauth/signup. When we started to add endpoints, we still wanted to know from which screen the user was successfully signing in, so we "extended" the event name - fxa.content.<screen_name>.signin.success. It has become unruly because there is no single fxa.content.signin.success anymore.

This comment has been minimized.

@shane-tomlinson

shane-tomlinson Sep 26, 2016
Member

I forgot, the signin event is further complicated by signin confirmation and the coming signin unblock, and even OAuth permissions which we don't currently include in our metrics.

Figuring out what signin.success means is not an easy task. Does it mean the user has fully completed the signin flow, post email verification? With signin confirmation, we can't measure that using only client side metrics because the client doesn't always know when the user has completed the flow since users sometimes close FxA before verifying their email.

This comment has been minimized.

@rfk

rfk Sep 26, 2016
Member

+1 to what @shane-tomlinson said; my high-level thinking is basically that we want an explicit "you completed what you were trying to do" event that's emitting at the very last stage of the flow. For sync signins that's when the browser actually succeeds in connecting to sync. For oauth signins that's when the relier successfully obtains their access token. Neither of those can be measured using only client-side metrics.

IOW, bring on the flow.completed flow event for future explorations of this type :-)


<img src="sign-in-success-time.png" height="300">

This chart shows that most users complete sign-in

This comment has been minimized.

@davismtl

davismtl Sep 23, 2016
Contributor

How does it help us to know that confirmation happens usually within an hour?

This comment has been minimized.

@vbudhram

vbudhram Sep 27, 2016
Author Contributor

I think this helps us eliminate some reasons why the server-side metrics and client-side metrics differed on sign-in confirmation.

in these metrics is that one method might not be emitting
or emitting twice, for different scenarios. The true
sign-in confirmation success rate could possibly be a
mix of these metrics.

This comment has been minimized.

@davismtl

davismtl Sep 23, 2016
Contributor

Do we have telemetry that would help us answer this? Were we tracking all the way back then?

@rfk
Copy link
Member

@rfk rfk commented Sep 27, 2016

What would be your overall conclusion?

+1, I think this is worth trying to summarize explicitly. From my perspective the feature seems worthwhile on balance, but we definitely pay a tax for the increased security, so follow-up work is to try different ways of avoiding sending the email when it's safe to do so.

@vbudhram
Copy link
Contributor Author

@vbudhram vbudhram commented Sep 29, 2016

@davismtl @rfk Added an explicit conclusion with some more supporting data. Mind another look over?

Copy link
Contributor

@davismtl davismtl left a comment

Looks good to me. Only 2 little changes but nothing serious

email. Sign-in confirmation helps to provide these users
with some protection on their accounts.

<img src="sign-in-change-password.png" height="300">

This comment has been minimized.

@davismtl

davismtl Sep 30, 2016
Contributor

This image isn't loading for me. :-/


<img src="sign-in-change-password.png" height="300">

Overall, the sign-in confirmation feature has provided

This comment has been minimized.

@davismtl

davismtl Sep 30, 2016
Contributor

Prob worth making everything past this it's own section so that it pops. Conclusion and Next Steps

@vbudhram vbudhram merged commit f23707f into master Oct 3, 2016
2 checks passed
2 checks passed
continuous-integration/travis-ci/pr The Travis CI build passed
Details
continuous-integration/travis-ci/push The Travis CI build passed
Details
@vbudhram vbudhram deleted the sign-in-confirmation-results branch Oct 3, 2016
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Linked issues

Successfully merging this pull request may close these issues.

None yet

4 participants