Page MenuHomePhabricator

FY2025/26 WE 1.1: Establish baseline for constructive edits
Closed, ResolvedPublic

Description

This task involves the work of establishing a baseline of constructive edits made by new(er) volunteers, as defined by the 2025 - 2026 WE 1.1 KR, to help inform how much we will attempt to grow the number in the coming fiscal year (T389403).

Where, "constructive edits made by new(er) volunteers" is defined as:
Edits to pages in any Wikipedia main namespace by people who have published ≤100 cumulative edits (incl. people who are logged out) that are not reverted within 48 hours of being published.

Decision(s) to be made

  1. By what percentage will the WE 1.1 KR attempt to grow constructive edits made by new(er) volunteers during Q1 and Q2 of the 2025 - 2026 fiscal year?
    • Note: we'll set this improvement target in T389403.

Requirements

Using ≥3 months of recent data, broken out by platform (mobile web, desktop) and if possible, by Wikipedia project...

  • Constructive edits by month*
  • Y-o-Y changes in constructive edits*
  • [curiosity] proportion of all published monthly edits that are constructive (not reverted within 48 hours) and destructive (reverted within 48 hours)

*As defined by KR criteria above.

References

We completed a similar analysis for the 2024-2025 WE 1.2 KR in T360829.

Event Timeline

ppelberg updated the task description. (Show Details)
ppelberg moved this task from Backlog to Analytics on the Editing-team (Tracking) board.
ppelberg added a project: Product-Analytics.
Aklapper renamed this task from WE 1.1: Establish baseline for constructive edits to FY2025/26 WE 1.1: Establish baseline for constructive edits.Jun 25 2025, 4:40 PM

I've completed an initial analysis of constructive edit rates, reviewing the most recent available three months of edit data and past YoY trends.

Methodology:
Published edit and revert data was collected from mediawiki_history. Data was limited to the following edits types:

  • Completed on a main page namespace (page namespace = 0)
  • Completed on any Wikipedia Project
  • Completed on Desktop or Mobile Web
  • Completed by an editor with 100 or fewer cumulative edits on the wiki or an unregistered user

I calculated the baseline based on the 3-month rolling average for monthly measurements of the metric. This smooths out the influence of specific events in a given month.

The most recently available months for review were February, March, and April 2025. May data is pending the June mediawiki snapshot (need two days in June to determie if edits made on the last day of May were reverted)

Desktop Constructive Edits Baseline

Desktop Monthly Constructive Edit Rates

MonthNumber of constructive editsProportion of constructive edits
2025-02-0195688581.8%
2025-03-01103366282.4%
2025-04-0194305882.9%

Desktop Baseline: 82.4% (Based on rolling 3 month average)

Desktop YoY Trends:

Desktop Constructive Edit Rate YoY Changes.png (371×600 px, 15 KB)

Constructive edit rates on desktop have been stable since 2021, with only small relative YoY changes observed (between 0.5 and 1.3%). There was a 0.5% increase in constructive edits from April 2024 to April 2025.

Mobile Web Constructive Edits Baseline

Mobile Web Constructive Edit Rates

MonthNumber of constructive editsProportion of constructive edits
2025-02-0148437070.9%
2025-03-0150335270.7%
2025-04-0148002071.6%

Mobile Web YoY Trends:
Mobile Web Baseline: 71.1% (Based on rolling 3 month average)

Mobile Web Constructive Edit Rate YoY Changes (2).png (444×719 px, 17 KB)

Mobile web rates have fluctuated slightly more than desktop rates (but still fairly stable) with YoY fluctuations between a low of -0.1% in 2023 to a high of 3.4% in 2021. There was a 2.5% increase in constructive edits from April 2024 to April 2025.

Some resources for additional exploration:

  • I've compiled monthly constructive edits rate data in this google sheet
  • There's a revert rate superset dashboard that can be used to review constructive edits by interface, user edit count, Wikipedia, and platform. Constructive edits is the inverese of revert rate so constructive edits rates are indicated by the trend lines marked "false" in the chart. Reverts here are defined as all edits reverted at any time vs within 48 hours; however, this should have minimal impact especially at larger wikis where most edits are reverted within that timeframe.
  • Code repo for details on query and analysis methodology
MNeisler triaged this task as Medium priority.Jun 25 2025, 5:47 PM

Per offline discussions, @MNeisler and I are considering work on this task done.

Next steps...

  • Use baseline to set the WE 1.1 improvement target; this work is happening/happened in T389403
  • Include baseline metrics in the soon-to-be published KR project page (T398255)