Closed
Bug 1405822
Opened 8 years ago
Closed 7 years ago
/firstrun email first test for fxa
Categories
(www.mozilla.org :: General, enhancement)
Tracking
(Not tracked)
VERIFIED
FIXED
People
(Reporter: adavis, Assigned: agibson)
Details
(Whiteboard: [fxa])
Attachments
(2 files)
Testing asking for email first has been a long time planned test.
I've prepared a test recipe here:
https://docs.google.com/document/d/1seUQEXOR-JGD6ff77iAP8lzDjW-BPengXugUzo-p2UQ/edit#
The recipe is missing screenshots but below is the summary of the changes..
COPY
Since the current copy says to login and that becomes irrelevant when we just ask for email, we propose a very simple copy change for the test to:
>Take Firefox with You
>Get your bookmarks, history, passwords, and other settings on all your devices.
TRACKING
To track the full FxA funnel, please past different UTM params to our iframe:
-For the control, I propose utm_source=firstrun-login
-For the treatment, I propose utm_source=firstrun-emailfirst
iFRAME:
I'm getting the details about what needs to change so that we can display the email first flow instead.
| Reporter | ||
Comment 1•8 years ago
|
||
IIUC, in order to load email first instead of current iframe you would need to load /?action=email rather than /signin
Comment 2•8 years ago
|
||
I believe this is what Alex is asking for. Just basic copy tweak on the left, and the Email First design on the right.
See attached.
I would much prefer a more ambitious goal of moving away from the iframed-login page. It's a bad practice, and tries to split the users’ attention. My guess is that few read what's on the left before signing up (hence why so many don't know what Sync does). Ideally this page would focus on selling Sync and subsequent steps could focus on set-up (not explanation of Sync).
Here are three ideas that do not involve the iframe: https://youtu.be/rHrFCnuSJV0
| Reporter | ||
Comment 3•8 years ago
|
||
(In reply to Ryan Feeley [:rfeeley] from comment #2)
> Created attachment 8915343 [details]
> iframe.png
>
> I believe this is what Alex is asking for. Just basic copy tweak on the
> left, and the Email First design on the right.
Perfect this is exactly it! Thank you
>
> I would much prefer a more ambitious goal of moving away from the
> iframed-login page. It's a bad practice, and tries to split the users’
> attention. My guess is that few read what's on the left before signing up
> (hence why so many don't know what Sync does). Ideally this page would focus
> on selling Sync and subsequent steps could focus on set-up (not explanation
> of Sync).
>
> Here are three ideas that do not involve the iframe:
> https://youtu.be/rHrFCnuSJV0
I agree with everything you’re proposing. We can review this after 57 release and after we’ve validated that the email first flow is beneficial on firstrun.
If we change too many things, it’ll prevent us from efficiently measuring the impact of email first.
As a reminder, the hypothesis of his test is:
>If we ask users for their email address only on /firstrun and redirect them to the appropriate page (/signin or /signup), we will see a greater percentage of users complete both flows because the flow will be simpler for all users.
Comment 4•8 years ago
|
||
(In reply to Alex Davis [:adavis] [PM FxA+Sync] from comment #3)
> If we change too many things, it’ll prevent us from efficiently measuring
> the impact of email first.
Could we A/B/C test Ryan's proposal against the other two possibilities?
Firstrun is by far the highest volume integration we have, getting sufficient
numbers for more than one bucket shouldn't be a problem. I agree with Ryan
that the iframe has been a bane for us, and all the content server work is
already done to enable what Ryan proposes, the only portion that remains
is the firstrun side.
| Reporter | ||
Comment 5•8 years ago
|
||
If the Moz org team is able to do A/B/C, I would not be opposed to the last variation in rfeeley's video which asks for the email and passes it to FxA so long as we use the EXACT SAME COPY in version B and C to isolate the impact of removing the iFrame without any other variables.
This would could help us better understand the impact of asking for email first, just in a different way.
What is missing from Ryan's proposal is:
- what will it look like after the user enters his email?
- how will that experience lead users into the aha moment with the animation that leads to Activity Stream with the new onboarding tour?
Rfeeley and I worked on a feature document linked above (i.e. "test recipe" for the marketers), if we are to propose a new variation please add details of the flow there.
If we can't easily answer the questions and outline the full flow this week, I will ask that we post-pone the extra variation. Fx 57 is released in 4-5 weeks, the onboarding team has been doing user testing before the release to make sure there is nothing obviously broken with onboarding, we don't have much wiggle room here.
Comment 6•8 years ago
|
||
(In reply to Alex Davis [:adavis] [PM FxA+Sync] from comment #5)
>
> If we can't easily answer the questions and outline the full flow this week,
> I will ask that we post-pone the extra variation. Fx 57 is released in 4-5
> weeks, the onboarding team has been doing user testing before the release to
> make sure there is nothing obviously broken with onboarding, we don't have
> much wiggle room here.
Understood.
| Reporter | ||
Comment 7•8 years ago
|
||
Per the FxA meeting this morning, we're going to keep it simple with 57 around the corner.
Let's just test what was outlined in the bug description and Comment 1. See RFeeley's attached screenshot.
To make things clear, here is the recap:
**SUMMARY**
COPY
Since the current copy says to login and that becomes irrelevant when we just ask for email, we propose a very simple copy change for the test to:
>Take Firefox with You
>Get your bookmarks, history, passwords, and other settings on all your devices.
TRACKING
To track the full FxA funnel, please past different UTM params to our iframe:
-For the control, I propose utm_source=firstrun-login
-For the treatment, I propose utm_source=firstrun-emailfirst
SCREENSHOT
https://bugzilla.mozilla.org/attachment.cgi?id=8915343
iFRAME
In order to load "email first" instead of current iframe you would need to load /?action=email rather than /signin
| Reporter | ||
Comment 8•8 years ago
|
||
Additional details that I forgot and added to test document:
Locale: EN-*
Audience: 10% of traffic
Variations: 50/50 (5% each)
Comment 9•8 years ago
|
||
(In reply to Alex Davis [:adavis] [PM FxA+Sync] from comment #7)
> Per the FxA meeting this morning, we're going to keep it simple with 57
> around the corner.
>
> Let's just test what was outlined in the bug description and Comment 1. See
> RFeeley's attached screenshot.
>
> To make things clear, here is the recap:
>
> **SUMMARY**
>
> COPY
> Since the current copy says to login and that becomes irrelevant when we
> just ask for email, we propose a very simple copy change for the test to:
>
> >Take Firefox with You
> >Get your bookmarks, history, passwords, and other settings on all your devices.
>
> TRACKING
> To track the full FxA funnel, please past different UTM params to our iframe:
> -For the control, I propose utm_source=firstrun-login
> -For the treatment, I propose utm_source=firstrun-emailfirst
>
> SCREENSHOT
> https://bugzilla.mozilla.org/attachment.cgi?id=8915343
>
> iFRAME
> In order to load "email first" instead of current iframe you would need to
> load /?action=email rather than /signin
Will the logo on this be updated for 57?
Flags: needinfo?(adavis)
Comment 10•8 years ago
|
||
> Will the logo on this be updated for 57?
The logo displayed in the FxA signin form? Yes, once [1] lands we should detect when the user is on 57+ and show the new logo.
[1] https://github.com/mozilla/fxa-content-server/pull/5603/files
| Reporter | ||
Comment 11•8 years ago
|
||
Clearing ni? since :rfkelly answered
| Reporter | ||
Updated•8 years ago
|
Flags: needinfo?(adavis)
| Assignee | ||
Updated•7 years ago
|
Assignee: nobody → agibson
Status: NEW → ASSIGNED
Comment 12•7 years ago
|
||
:agibson - I should have assigned this to myself, I've already got a WIP! Lemme push it up to get feedback and start to ask questions.
Comment 13•7 years ago
|
||
:agibson - https://github.com/mozilla/bedrock/pull/5316 is the WIP PR I was talking about.
| Assignee | ||
Comment 14•7 years ago
|
||
Ah ok, Eric asked me to pick this up - but if you've already got a WIP by all means :)
| Assignee | ||
Updated•7 years ago
|
Assignee: agibson → stomlinson
Comment 15•7 years ago
|
||
Commit pushed to master at https://github.com/mozilla/bedrock
https://github.com/mozilla/bedrock/commit/a3ef768d9eb9294d7b20e66ae300e97d14c3b767
[fix bug 1405822] email-first experiment on firstrun
Updated•7 years ago
|
Status: ASSIGNED → RESOLVED
Closed: 7 years ago
Resolution: --- → FIXED
Comment 16•7 years ago
|
||
Commit pushed to master at https://github.com/mozilla/bedrock
https://github.com/mozilla/bedrock/commit/5411a3045a65b3b075f775ab3edfda931dfee6ff
[bug 1405822] Fix firstrun experiment switch name
Comment 17•7 years ago
|
||
Commits pushed to master at https://github.com/mozmeao/www-config
https://github.com/mozmeao/www-config/commit/42e874d1635765da37cfe4326979d7fbaf479f46
[bug 1405822] Enable /firstrun email experiment
https://github.com/mozmeao/www-config/commit/a13fc1e0506bc614dfaa2f48c0ae85ba302fb42a
Merge pull request #63 from alexgibson/enable-firstrun-email-experiment
[bug 1405822] Enable /firstrun email experiment
| Assignee | ||
Comment 18•7 years ago
|
||
Just checking in here - how are we doing in terms of data? Is there an expected date when this experiment can be shut off?
Flags: needinfo?(adavis)
Comment 19•7 years ago
|
||
FYI :adavis is out this week, returning on the 15th.
I tried to have a look at the metrics so far, but unfortunately there seems to be a problem with our redshift cluster and I haven't yet been able to execute the necessary queries. I expect we should have more than enough data though, so if there's a pressing need to turn off the experiment please feel free to do so. Otherwise we can double-check with Alex when he returns.
| Assignee | ||
Comment 20•7 years ago
|
||
(In reply to Ryan Kelly [:rfkelly] from comment #19)
> FYI :adavis is out this week, returning on the 15th.
>
> I tried to have a look at the metrics so far, but unfortunately there seems
> to be a problem with our redshift cluster and I haven't yet been able to
> execute the necessary queries. I expect we should have more than enough
> data though, so if there's a pressing need to turn off the experiment please
> feel free to do so. Otherwise we can double-check with Alex when he returns.
We do have some other experiments coming on the horizon, but I think we're fine to wait until the 15th. Thanks!
Comment 21•7 years ago
|
||
I've started gathering the results of this experiment here:
https://sql.telemetry.mozilla.org/dashboard/fxa-experiment-firstrun-emailfirst
| Reporter | ||
Comment 22•7 years ago
|
||
Small update: Leif and I are still doing the full analysis.
We were digging in more in Amplitude for a couple of hours yesterday but there were some metric anomalies (which :pbooth is digging into).
Leif will fork :rfkelly 's query and change the cohorting to use the experiment cohorts rather than the utm params being used in it since there may be problems with the utm params too according to :pbooth
Leaving ni? on me
| Reporter | ||
Comment 23•7 years ago
|
||
:agibson , it is our recommendation to roll-out the treatment (email-first version).
We've logged our analysis here and are making sure to get it to everyone who is concerned:
https://docs.google.com/document/d/1seUQEXOR-JGD6ff77iAP8lzDjW-BPengXugUzo-p2UQ/edit#
In summary, more users were syncing on the new version and overall more users were retained in the long run. There were just some tradeoffs as always between logins and registrations.
Flags: needinfo?(adavis)
| Assignee | ||
Comment 24•7 years ago
|
||
(In reply to Alex Davis [:adavis] [PM FxA+Sync] from comment #23)
> :agibson , it is our recommendation to roll-out the treatment (email-first
> version).
>
> We've logged our analysis here and are making sure to get it to everyone who
> is concerned:
> https://docs.google.com/document/d/1seUQEXOR-JGD6ff77iAP8lzDjW-BPengXugUzo-
> p2UQ/edit#
>
> In summary, more users were syncing on the new version and overall more
> users were retained in the long run. There were just some tradeoffs as
> always between logins and registrations.
I'll flag this with Eric and make sure it gets on our board. Thanks
Flags: needinfo?(erenaud)
Comment 25•7 years ago
|
||
Alex - I confirmed with the Retention team that we're good to go with taking ?v=b live to 100% of audience, as soon as possible.
Flags: needinfo?(erenaud)
| Assignee | ||
Updated•7 years ago
|
Assignee: stomlinson → agibson
Status: RESOLVED → REOPENED
Resolution: FIXED → ---
| Assignee | ||
Updated•7 years ago
|
Status: REOPENED → ASSIGNED
Comment 26•7 years ago
|
||
| Assignee | ||
Comment 27•7 years ago
|
||
This is now up on demo3:
https://www-demo3.allizom.org/en-US/firefox/58.0/firstrun/
Note: to test on demo you'll need to both enable UITour [1] and set up a profile to test the iframe on demo [2].
When testing a signup for a new account, I'm not receiving any verification emails. I guess this is likely due to the staging server config, but I'd like to verify that the full flow works as expected before pushing this forward to production. Can someone please investigate? (I remember when we tested the experiment, Shane had this running against his own instance and not the regular staging server).
[1] http://bedrock.readthedocs.io/en/latest/uitour.html#local-development
[2] http://bedrock.readthedocs.io/en/latest/firefox-accounts.html#demo-server-testing
Flags: needinfo?(adavis)
| Assignee | ||
Updated•7 years ago
|
Flags: needinfo?(rfkelly)
Comment 28•7 years ago
|
||
> I'd like to verify that the full flow works as expected before pushing this forward to production
I just tried it out and this worked end-to-end for me, using the following steps:
* Create a new profile
* Add pref "browser.uitour.testingOrigins" with value "https://www-demo3.allizom.org"
* Add pref "identity.fxaccounts.autoconfig.uri" with value "https://accounts.stage.mozaws.net"
* Go to about:preferences#sync and click "sign in"
* Check that it correctly loaded the signin page from https://accounts.stage.mozaws.net, but don't actually sign in
* Restart the browser
* Visit https://www-demo3.allizom.org/en-US/firefox/58.0/firstrun/
* Go through signup process and get connected to sync
So AFAICT this is looking :thumbsup:!
Flags: needinfo?(rfkelly)
| Assignee | ||
Comment 29•7 years ago
|
||
(In reply to Ryan Kelly [:rfkelly] from comment #28)
> > I'd like to verify that the full flow works as expected before pushing this forward to production
>
> I just tried it out and this worked end-to-end for me, using the following
> steps:
>
> * Create a new profile
> * Add pref "browser.uitour.testingOrigins" with value
> "https://www-demo3.allizom.org"
> * Add pref "identity.fxaccounts.autoconfig.uri" with value
> "https://accounts.stage.mozaws.net"
> * Go to about:preferences#sync and click "sign in"
> * Check that it correctly loaded the signin page from
> https://accounts.stage.mozaws.net, but don't actually sign in
> * Restart the browser
> * Visit https://www-demo3.allizom.org/en-US/firefox/58.0/firstrun/
> * Go through signup process and get connected to sync
>
> So AFAICT this is looking :thumbsup:!
Ok, I verified this works using a different email address via the steps above. Thanks for testing!
Flags: needinfo?(adavis)
Comment 30•7 years ago
|
||
Commit pushed to master at https://github.com/mozilla/bedrock
https://github.com/mozilla/bedrock/commit/3cd528bd9f17592c7a71aa9a2a5362840e5cb84e
[fix bug 1405822] Set Firefox Accounts email-first form as default on /firstrun
Updated•7 years ago
|
Status: ASSIGNED → RESOLVED
Closed: 7 years ago → 7 years ago
Resolution: --- → FIXED
| Assignee | ||
Comment 31•7 years ago
|
||
Verified in production https://www.mozilla.org/en-US/firefox/58.0/firstrun/
Status: RESOLVED → VERIFIED
You need to log in
before you can comment on or make changes to this bug.
Description
•