@ppelberg I've calculated initial baselines of retention rates to help inform the KR target. See summary of rates and key definitions below.

Baseline Rates:
Registered Junior Editors: 10% (based on average 2025 rates); Closer to 9% if we use the most recent snapshot (Dec 2025)
Logged-Out Users: 0.2% (based on Dec 2025 rates [1])
Registered Junior Editors + Logged-Out Users: 0.5%

[1] We have limited retention data for logged-out users as we only store a unique identifier for logged-out users bucketed in an experiment and data is scrubbed after 90 days.

I've started reviewing YoY Trends and past interventions in this google doc.

OPEN QUESTIONS:

1. Do we plan to include logged-out users as part of our Junior Editor definition? If so, their inclusion significantly decreases retention rates from about 11% to 0.5%. This low retention rate is likely because the >2 seconds threshold is not quite strict enough to exclude some low intent logged-out user behavior.
   1. If we want to include logged-out users, we could try exploring the impact of increasing the 2-second threshold or limiting the initial action to only saved edits (if we limit the initial action to saved edits, logged-out user retention increases to 2.5%)
2. Target Wikis. The current retention rates looks across all Wikipedia projects; however, retention rates can vary depending on wiki. Are there specific target wikis rates you'd like to review?

Definitions
The retention rates above are based on the following definitions:

KR: "Achieve a minimum #% relative increase in 2-week retention among Junior Editors who spend at least 2 seconds in the mobile visual editor, based on controlled A/B test(s) across target wikis.[i]
i. "Retention" = people who publish  ≥1 edit (across namespaces) within 2 weeks of an edit session in which they spent at least ≥2 seconds in VE's ready state.

Initial Action: A Junior editor opened the mobile visual editor on a main namespace on a Wikipedia project and spent at least ≥2 seconds in VE's ready state.
Return Action: A Junior editor successfully published an edit on mobile visual editor on any namespace on a Wikipedia project.
Retention Length: 2 to 14 days after the user's initial action.

Per yesterday's discussion in the WE1.7 + WE1.1 steering committee meeting, we've decided to adjust the KPI for WE1.7 from constructive edit rate to edit completion rate among qualified edit sessions.

Rationale

1. WE 1.7 is about shifting people from doing "nothing" (curiously tap Edit and abandon) into doing "something" (curiously tap Edit and publish a constructive change)
2. Edit quality is important to monitor and not what we're optimizing
3. Retention is downstream of the impact this KR is intended to cause (see "2.")

@ppelberg Here are the current baseline rates for the new edit completion rate KR. These are based on mobile VE completion rates on Wikipedia main namespaces in December 2025 - February 2026, excluding English Wikipedia [i]

Edit Completion Rate Baseline: 4.7% (Logged-out users and registered users with < 100 edits)

Variation by user type
The 4.7% baseline is heavily weighted by logged-out users (2.2% completion rate). Registered newcomers actually perform 20x better, with a baseline edit completion rate of 44.2%.

This significant drop-off of logged-out users before they start typing aligns with these mobile mid-funnel analysis findings, which also found that the majority IP users abandon their edits in VE before starting to type. This may still include some automated traffic or indicate increased technical or usability barriers for logged-out users.

YoY Trends
While the volume of qualified monthly editing sessions has decreased, overall completion rates have remained stable (ranging between 4.1% and 5.2%).

However, trends in completion rates vary based on user type. Logged out users completion rates have been decreasing YoY while registered users rates have been increasing YoY. See monthly rates documented in this doc.

[i] English Wikipedia Edit Completion Rates (Non-Default Outlier)
We see a significantly higher edit completion rate on English Wikipedia. Baseline: 34.2% (14% logged-out users; 40.1% registered users)

This is likely because mobile VE is not the default editor on English Wikipedia. Since it's not the default, the only newcomers who start a Mobile VE edit on enwiki are those who explicitly switched to it. These are more high-intent editors. Most low-intent and accidental clicks would just open the default wikitext editor without switching.

I've currently excluded enwiki edits from the baseline analysis since they were inflating overall completion rate numbers and would provide an unrealistic baseline for wikis where VE is the default editor. If we decide to include English Wikipedia in an AB test, I'd recommend reviewing enwiki results separately from other wikis in the experiment if deployed while mobile VE is not the default editor.

Some other observations:

• If we don't limit to qualified edit sessions, edit completion rate decreases to about 2%, indicating that the 2-second threshold is still effective at removing some accidental click noise and automated traffic that slipped through.
• Increasing the "Ready State" threshold from 2s to 5s raises the baseline to 6.7%, so a slight but not drastic increase.

@ppelberg and I discussed narrowing the scope of the KPI to registered users.

Rationale: This focus ensures that the high volume of low-intent, logged-out traffic does not drown out the measurable impact of our interventions on registered users. We can also be more confident that users who create an account, click edit, and spend 2 seconds in the editor represent the "curious clicks" we are targeting, vs logged-out traffic, which based on volume likely still includes some automated traffic and accidental click noise.

Note: We will also monitor the impacts on logged-out users as part of secondary metrics and guardrails.

Target Recommendation for Registered User Edit Completion Rate:
Based on recent data (Dec 2025 – Feb 2026), our current baseline for this target audience is 44%.

A +10% relative increase would be high enough to distinguish product impact from monthly fluctuations and anticipated natural growth in edit completion rates.

This recommended target is based on the following:

YoY Trends

• In the last 12 months (Feb '25 to Feb '26), edit completion rate grew from 37.4% to 43.9% (+17.3% relative increase; +6.5 percentage points). If this current rate of increase continues, the baseline would naturally reach ~46% by June 2026.
• Monthly variance (noise) typically ranges from 0.5 to 1.5 absolute percentage points.
• A +10% relative target (+4.4 points) provides a result that is statistically distinct from both these seasonal fluctuations and current growth trends.

Past interventions:
Similar historical interventions (e.g., DiscussionTools on Mobile Web) achieved increases of 3 to 14 percentage points in edit completion rates. The proposed +4.4 point gain falls within this historical range while accounting for the fact that earlier interventions targeted lower baselines.

Note: In a controlled experiment, we will be able to measure the additional increase provided by the specific interventions by controlling for natural and seasonal changes.