Page MenuHomePhabricator

Set up A/B test for “Revise Tone” Structured Task
Closed, ResolvedPublic

Description

Implement a experiment that splits eligible editors into a 50 percent Control and 50 percent Treatment on pilot wikis.

  • Control receives the current Suggested Edits experience (Status quo Easy task set and ordering as currently configured per wiki.)
  • Treatment receives “Add a Link” and “Revise Tone” as the Easy tasks shown in Suggested Edits.
Details
  • Use the GEImproveToneSuggestedEditEnabled feature flag (T401192) to implement a 50/50 split, assigning half of participants to the Control experience and half to the Treatment experience.
  • Which Pilot wikis will be included in this A/B test?
    • TBD: We will need to limit to some or all of the wikis where Tone Check has been tested.
  • Are we testing on Mobile or Desktop?
    • TBD: Ideally both, but only mobile metrics will be part of the final WE1.1 measurement for success, so the Mobile UX and release should be prioritized.
  • Will be set up this experiment via the Experimentation Lab?
    • We should instrument and set up this experiment using the Experimentation Lab, unless there are major blockers the will significantly delay or derail experiment findings.
Acceptance Criteria:
  • Add experiment bucketing and assignment
    • bucketing is needed on the Homepage for who has access to Revise Tone
    • in Leveling Up, Revise Tone should never be suggested, regardless of buckets
  • Allow for QA testing time
  • Agree on Release Date with Product and Movement Communications.
    • Release date TBD

Note:

Event Timeline

Michael renamed this task from Set up A/B test for “Improve Tone” Structured Task to Set up A/B test for “Revise Tone” Structured Task.Oct 10 2025, 7:08 PM
Michael removed a project: Growth-Structured-Tasks.
Michael updated the task description. (Show Details)

Change #1198309 had a related patch set uploaded (by Sergio Gimeno; author: Sergio Gimeno):

[mediawiki/extensions/GrowthExperiments@master] ReviseTone: A/B test experiment setup

https://gerrit.wikimedia.org/r/1198309

Sgs moved this task from Needs Estimation to FY2025-26 Q2 Sprint 2 on the Growth-Team board.
Sgs edited projects, added Growth-Team (FY2025-26 Q2 Sprint 2); removed Growth-Team.
Sgs moved this task from Incoming to Doing on the Growth-Team (FY2025-26 Q2 Sprint 2) board.

Change #1201576 had a related patch set uploaded (by Sergio Gimeno; author: Sergio Gimeno):

[mediawiki/extensions/GrowthExperiments@master] [WIP] cypress(ReviseTone): override experiment variant

https://gerrit.wikimedia.org/r/1201576

Change #1202721 had a related patch set uploaded (by Sergio Gimeno; author: Sergio Gimeno):

[operations/mediawiki-config@master] [beta] GrowthExperiments: add revise-tone experiment setup

https://gerrit.wikimedia.org/r/1202721

Change #1198309 merged by jenkins-bot:

[mediawiki/extensions/GrowthExperiments@master] ReviseTone: A/B test experiment setup

https://gerrit.wikimedia.org/r/1198309

Change #1202721 merged by jenkins-bot:

[operations/mediawiki-config@master] [beta] GrowthExperiments: add revise-tone experiment setup

https://gerrit.wikimedia.org/r/1202721

Would it be right to say that since accounts are being opted into the control and treatment group for Revise Tone, the experiment is setup? I've confirmed on my side that a newly account can be opted into the treatment group and the Revise Tone isn't being suggested in Leveling up.

I'm okay waiting for a second opinion on this before stating the acceptance criteria as done.

This sounds good given we're re-testing bucketing as-well while testing the instrumentation T405177.