Croissance/Mises à jour de l’équipe Croissance
Cette page contient les mises à jour sur le travail de l'équipe Croissance pour accroître l'activation de nouveaux contributeurs et les garder. La planification de ce travail a commencé en juin 2018. L'équipe travaille sur plusieurs projets à la fois, mais ils seront tous discutés sur cette page.
Comment s'impliquer ?
Il est important que notre travail soit fondé sur la réalité des communautés que nous espérons aider. Si vous avez des remarques ou des idées sur ce projet ou sur le travail de notre équipe, veuillez laisser vos commentaires sur cette page de discussion du projet.
Mises à jour du projet
Les mises à jour pour 2026 sont listées ci-dessous, à partir de cette page. Suivre les mises à jour 2026.
Vous pouvez également vous abonner à notre bulletin d'information.
Growth team updates: 2025, 2024, 2023, 2022, 2021, 2020, 2019, 2018.
Account Creation Experiments: Multiple experiments exceed targets
The project moved closer to its overall goal of identifying multiple successful interventions, and we are starting to document experiment results and scale successful interventions:
- Scale updated logged-out edit warning on mobile to all wikis (T424595): After just over one month of experimentation (since March 17, 2026), results show a statistically significant increase in account creation (p = 0.038), with a 27% relative lift among users exposed to the updated warning. This shift is accompanied by a 16% relative decrease in temporary account creation (a decrease is expected). Downstream contributor metrics were also monitored in this experiment, but there was no significant impact on constructive edit rates or other metrics monitored.
- Account Creation V1 Experiment: The first version (V1) of the account creation form experiment also showed promising results, with an estimated 32% relative increase in account creation among users exposed to the simplified form experience. Because of limitations in how account creation could be measured across Wikimedia's authentication systems, the results are considered directional rather than definitive. To validate these findings, a second version (V2) of the experiment is being launched using improved experimentation and measurement infrastructure. This follow-up work will help determine which form changes should be considered for broader deployment.
- Improving account creation visibility on mobile: To make it easier for readers to discover account creation, a dedicated account creation button was added to the mobile web header. The experiment produced a 20% relative increase in account creation, demonstrating that small improvements to account creation entry points can have a meaningful impact.
Account Creation Experiments: First significant result achieved
- The logged-out warning experiment delivered a statistically significant result, producing a 27% relative increase in account creation for mobile users exposed to the new experience. Two additional experiments, focused on account creation form improvements and mobile entry points, continued to show positive early trends while gathering more data.
- As part of data analysis to establish baselines for account creation behavior, we have learned:
- On mobile, most users drop off at the user_name field after interacting with the field. This is most likely due to a validation error.
- On desktop, the majority of drop-offs were associated with the captcha field (element_id = captcha AND action = type), accounting for 71% of drop-offs. After this, most users dropped off at user_name field.
- Password and confirm_password drop-offs account for roughly 26% of all bounces on mobile, while they only account for 6.75% on desktop. This likely highlights users’ difficulty typing characters using the smaller mobile screen and editor.
Revise Tone: Experiment results
- The Revise Tone structured task helps newcomers improve article neutrality through a guided workflow. In testing across four Wikipedias, Revise Tone task completion rates for newcomers were 38.7% higher than Copyedit task completion rates overall (p < 0.001), with no evidence of reduced edit quality. These results suggest that structured tasks can make editing easier and more successful for new contributors. Read the full Revise Tone experiment results here.
Account Creation Experiments: Three mobile-focused experiments launched
Three experiments aimed at increasing successful account creation on mobile were launched: improving the messaging shown to logged-out users, increasing the visibility of account creation entry points, and iterating on the account creation form. Early directional signals were encouraging, though measurement limitations made it difficult to fully assess completed registrations at that stage.
Account Creation Experiments: Experiments move into implementation
- Work progressed from research and design into active development. The first mobile-focused experiment, updating the warning shown to logged-out users who try to edit, was prepared for launch across multiple wikis. Prototype work also began on improvements to the account creation form, while feasibility research continued on privacy-preserving third-party login options.
- The Revise Tone structured editing task was released to 50 percent of logged-in users on English, Spanish, French, and Arabic Wikipedia as part of an A/B test. After approximately two weeks, we analyzed a set of early leading indicators to assess whether the experience was functioning as expected and to identify any issues before evaluating the feature’s full impact. Initial results suggest the feature is working well overall.
- In February 2026, across the four pilot wikis, 9,558 Revise Tone edits were completed.
- The revert rate is under 3%, which is well below typical newcomer revert rates.
- Task completion rates are broadly comparable to other medium-complexity structured tasks on desktop. Early data and instrumentation also helped identify issues with the initial experiment setup and a mobile scrolling regression. Several fixes and improvements have already been released in response. These findings highlight the importance of early leading indicator analysis and continued improvements to mobile editing stability. Overall, the results indicate that Revise Tone can support high-quality edits and meaningful skill development for newer contributors, though the task’s higher complexity may lead to lower activation when shown to brand-new editors. This reinforces the value of a progression model that begins with simpler tasks and gradually introduces more complex editing activities. A full analysis of the Revise Tone A/B test will be completed and reported as part of: T407802.
- On January 26, 2026, the Revise Tone task launched to 50 percent of logged-in users on English, Spanish, French, and Arabic Wikipedias (T413065).
- In the first five days on English Wikipedia, more than 1,500 edits were completed, with a revert rate under 2 percent, well below typical newcomer rates. Revise Tone edits are visible by narrowing to the Suggested: revise tone task in Recent Changes (on pilot wikis).
- Early community spot checks and Product Ambassador reviews found most edits to be constructive. Early feedback from newer editors reinforce the value of offering tasks with different levels of complexity as editors gain confidence.
- Initial Revise Tone metrics show most users who start the onboarding quiz complete all five questions, suggesting this is a more engaging way to introduce editing norms than previous less-interactive onboarding.
- We are starting engineering work to A/B test an improvement to the "Logged out warning message" that appears when a logged out user on mobile navigates to edit (T415160).
- We are also conducting early research to document the existing validation logic used in the Account Creation form, with a focus on username validation, password validation, and related error messages (T414235). We will also complete a brief investigation to figure out how we can modify these forms in a clean way, gated by a feature flag, and so that we can toggle it with a Test Kitchen A/B test (T415689).
The Revise Tone Structured Task is scheduled to launch on January 22, 2026, to 50 percent of users on English, Arabic, French, and Portuguese Wikipedias (T413065).
Growth Product Ambassadors have communicated the plan across all participating wikis, and no objections have been raised. An opt-in mechanism for experienced editors has been implemented (T414499), using a reusable URL parameter override that simplifies internal and volunteer testing.
A high-priority Revise Tone bug has been resolved (T413358). Work is also wrapping up on onboarding quiz improvements to ensure availability via the Help Panel while preventing repeated exposure to the same users (T414245, T413368).
The ML team is improving filtering for quotes and certain article sections, which must be completed before the A/B test launch (T412210). Additional ML performance optimizations are under consideration but are not blocking the experiment (T411758).
The Growth team has shared initial ideas for its next focus area. These will be discussed with pilot wikis in the coming months, with designs shared as the strategy develops. Current plans include four potential experiments, which may evolve based on community feedback and team capacity:
- 1. Improving the Account Creation Experience: Simplify and modernize the core account creation flow, with a focus on mobile usability and clarity. Reduce cognitive load by streamlining the interface, consolidating legacy warnings, and clearly communicating the benefits of account creation in newcomer friendly language.
- 2. Encouraging Temporary Accounts to Register: Explore when and how temporary account holders are most motivated to create a permanent account. Test targeted prompts at high intent moments, supported by lightweight research such as surveys, and emphasize how registration unlocks features like mentorship, Suggested Edits, and community recognition.
- 3. Making Account Creation Easier to Find on Mobile: Address the limited visibility of account creation for logged out mobile readers by testing clearer, more intuitive entry points and better differentiation between login and registration.
- 4. Surfacing Reading Lists to Logged Out Readers: Explore whether exposing Reading Lists to logged out readers can drive deeper engagement and encourage account creation by demonstrating how an account enhances the reading experience through a low pressure path to registration.
The Wikimedia Research team has published findings from two research projects that examine how new editors on English Wikipedia learn, ask questions, and integrate into the community, combining in-depth analysis of mentorship systems with a broad survey of successful newcomers’ tools, experiences, and sentiments. By linking how newcomers seek help and receive guidance with what enables them to persist and succeed, the work provides a more holistic foundation for improving early contributor support and shaping Growth team interventions.
- This work explores how mentorship for new editors functions on English Wikipedia, with the goal of making it more effective and sustainable for both newcomers and experienced volunteers. It combines qualitative analysis of newcomer questions and trajectories, quantitative measurement of mentorship activity across core help spaces, and early prototyping of tools to address observed gaps.
- The research finds that mentorship serves both technical and social learning needs, that many newcomers disengage quickly if responses are slow or unclear, and that mentors face high cognitive and time burdens despite growing demand, especially in article creation support. Large-scale analysis of hundreds of thousands of questions highlights recurring themes and duplication, suggesting opportunities to better route, surface, and answer common questions. This work helps lay the groundwork for future interventions that could reduce mentor load while helping newcomers get timely, relevant guidance.
- This project surveys successful newcomer editors on English Wikipedia to understand which tools, features, and community spaces they discover and use early in their editing journeys, as well as their interactions, sentiments, and emerging interests. Using a representative survey of editors who registered within the past six months and reached at least 25 edits, the research provides insight into tool awareness and usage, mentorship experiences, mobile editing challenges, help-seeking behaviors, and perceptions of community feedback such as reversion and thanks.
- The findings highlight strong intrinsic motivation and intent to continue editing, alongside significant gaps in awareness of community spaces, uneven experiences with mentorship and reversion, and a strong preference for independent ways of finding answers, including search engines and generative AI tools. Results also show meaningful differences by edit experience and editing platform, especially for mobile-first editors.
- This work informs Growth team product decisions, including the Newcomer Homepage, and contributes foundational evidence for the Wikimedia Foundation’s 2025 Annual Plan for Contributor Experience and future newcomer research across languages.
 | Growth team assure la maintenance de cette page ou ce projet.
Obtenir de l'aide :
|
