Page MenuHomePhabricator

Product Analytics: Revise Tone - Leading Indicator plan of action
Closed, ResolvedPublic5 Estimated Story Points

Description

User story & summary:

As a member of the Growth team, I want to quickly understand whether newcomers are able to complete the Revise Tone task successfully so that I can evaluate task difficulty and identify areas for early improvement.

~2 weeks after the start of the Revise Tone A/B Test, we will review the following leading indicators.

We will use these indicators to determine if any immediate adjustments or follow-up investigations are needed before continuing to full evaluation. We will use this task to document the analysis and next steps.

Leading indicators

(DRAFT: these may be adjusted)

IDIndicatorOwnerMetric(s) for EvaluationPlan of Action
1.Task completion rateProduct AnalyticsProportion of users who start a Revise Tone task and successfully publish an editIf completion is lower than the Copyedit Suggested Edit, investigate drop-off points in the workflow and adjust the UI or guidance to clarify next steps.
2.Edit Revert rateProduct AnalyticsProportion of published Revise Tone edits reverted within 48 hoursIf revert rate is significantly higher than baseline (unstructured copyedits), review reverted edits to identify recurring quality issues and adjust model thresholds or task guidance.
3.Rejection rateProduct AnalyticsProportion of cases where users reject the suggestionsIf high (>50%), it may indicate suggestions are irrelevant or too difficult. Review rejection reasons and determine if model improvements are needed, or if the task difficulty level needs to be changed.
4.Community evaluationGrowth PMProportion of completed edits considered "good" by experienced editorsSince "unreverted" doesn't mean an edit is "good" we should gather community feedback on edit quality. If Revise Tone edits are considered bad more than 75% of the time, we should consider shifting the task to "Medium" so it's less likely to be completed by brand new editors.
5.Model service availabilityMLService Availability SLO: 95% of all requests return a 200/300/400 responseIf SLO falls below 95%, prioritize infrastructure improvements or investigate failure points.

Event Timeline

Michael triaged this task as High priority.Nov 6 2025, 5:27 PM
Michael moved this task from Inbox to Product Backlog on the Growth-Team board.

The Indicator Plan of Action is now included in the Revise Tone Measurement Plan; See T397247.

KStoller-WMF renamed this task from Revise Tone: Leading Indicator plan of action to Product Analytics: Revise Tone - Leading Indicator plan of action.Nov 17 2025, 6:27 PM
KStoller-WMF lowered the priority of this task from High to Medium.
KStoller-WMF moved this task from Product Backlog to Needs Estimation on the Growth-Team board.

The Revise Tone A/B test has been delayed until January, so I've removed the Dec. deadline.

Note that the Revise Tone was released on Monday, January 26th.
The Growth team is independently reviewing tagged edits and certain KPIs via our Revise Tone Grafana dashboard, and we should have some A/B test data visible via Test Kitchen soon, but ideally we should revisit some of these higher-level leading indicators starting the week of February 9th.

That said, I know that Product Analytics is very busy, and I'm open to reducing requirements for this task if needed.

KStoller-WMF set the point value for this task to 5.Feb 5 2026, 5:14 PM
KStoller-WMF removed a subscriber: Iflorez.
NOTE: I will share some early data, but it is not yet statistically significant, and has not yet been peer-reviewed.

A full report, completed by a Data Scientist and peer-reviewed will be shared as part of: T407802: Product Analytics: Experiment Analysis - Revise Tone Structured Task (WE1.7.5, FY25-26)


IDIndicatorOwnerMetric(s) for EvaluationResult
1.Task completion rateGrowth PMProportion of users who start a Revise Tone task and successfully publish an edit✅ Overall task completion rate is ~36 percent, which is slightly lower but similar to "Add an Image" task completion rates.
2.Edit Revert rateGrowth PMProportion of published Revise Tone edits reverted within 48 hours✅ As of now, revert rates are low, ~2% on pilot wikis.
3.Rejection rateGrowth PMProportion of cases where users reject the suggestions✅ Rejections are at a healthy level. When a task is rejected, ~48% say "tone is appropriate", ~37% say they are unsure how to change it, and ~15% say they declined the suggestion for another reason. (from Revise Tone Grafana Dashboard data)
4.Community evaluationProduct AmbassadorsProportion of completed edits considered "good" by experienced editors✅ Product Ambassadors, WMF Staff, and Community members have reviewed edits, and initial reports are positive. Enwiki feedback.
5.Model service availabilityMLService Availability SLO: 95% of all requests return a 200/300/400 response✅ SLO met

More early data about the Revise Tone task can be viewed here:

This looks good to close from my end to close.

Etonkovidova subscribed.

This looks good to close from my end to close.

Looks complete to me. Looking at the Onboarding Quiz stats, the number of incorrect answers is noticeably bigger than the number of correct answers. Could it be the indication that users need more of guided learning, or, maybe to be referred to the policy/rules before answering a question?

Thanks! And that's a good insight @Etonkovidova , that I probably should have included in this leading indicators review!

The Revise Tone Onboarding Quiz stats show that many participants answer questions incorrectly, particularly Questions 2 and 5. @AAlhazwani-WMF and I discussed this earlier and concluded that this level of difficulty is acceptable for a few reasons:

  • When someone answers incorrectly, the quiz provides guidance, which creates a useful learning opportunity.
  • The vast majority of participants who start the quiz go on to complete all five questions. This suggests that, even though the onboarding is challenging, people remain engaged rather than becoming frustrated and dropping off.

Looking ahead, this may point toward a valuable direction for onboarding. Interactive and engaging formats like this quiz may be more effective than approaches that rely primarily on asking newcomers to read lengthy policies. For now, it seems reasonable that newcomers find the quiz somewhat challenging. Editing, particularly work related to maintaining a neutral point of view, is new for many people, so some difficulty is expected.