closing summary

over the past weeks we explored designs for the post-account-creation setup and validated them in two rounds of unmoderated mobile usability testing. the direction holds, so we are closing this task and moving to a design spec for the production experiment.

what we designed

a 3-step dialog shown immediately after account creation:

1. welcome plus "Personalize your Home",
2. "What brings you to Wikipedia?" with 3 options,
3. "What are 3 of your interests?" seeded from real wikipedia articles.

interests are expressed by searching real articles or tapping related-article chips, rather than picking from a fixed list of topics. the dialog is dismissible at any point. closing it lands the newcomer on the Newcomer Homepage in its default state. the step structure is deliberately convertible: the same steps can be stacked into a single dialog or page for a one-step variant, with no change to content or logic.

how we tested

2 rounds, unmoderated, mobile, same protocol and tasks:

1. the flow built with bespoke components (protocol, report)
   
2. the same flow built with stock codex components (protocol, report)
   

rationale for running two: codex components tend to be optimized for desktop, so some touch targets get small on mobile and screens can take on a form-like look. we wanted to see whether a more mobile-native, touch-friendly build made a material difference.

what we learned (findings common to both rounds)

• all participants went through account setup. nobody skipped. keeping setup as short as possible works well in conjunction with the revised account creation.
• "What brings you to Wikipedia?" is easy to answer. nobody struggled to place themselves in the 3 options.
• the welcome plus "Personalize your Home" moment reads as confirmation that the account now exists, not as a request for information. no one pushed back on being asked, and a couple of people compared it to picking topics when signing up for a streaming service.
• seeding interests from a real article works when suggestions come back from the system. people connected the suggestions to the article they had opened. the failure mode is the empty state: some finished setup with 3 interests but received 0 suggestions, so they had nothing to review.
• every input path on the interests screen got used: searching for a specific article, searching for a broader topic, and tapping the related-article chips. none went unused.
• dropped directly on the Newcomer Homepage with no instruction, everyone understood it as a place to contribute. the recurring gap is expectational: several people also wanted a reading-oriented Home (saved articles, favorites, bookmarks).
• the navigation banner works for getting back to the Home. getting back to the seed article is the consistent breakdown in both rounds: the search button either goes unnoticed or does not come as a natural pattern. The "User page / Talk" tabs under "Hello, [username]!" invited taps in both rounds.

acceptance criteria

• responses immediately influence at least one homepage element: met! through suggested edits. interests are used to align the suggestions and keep the batch small. nothing changes visually in the module itself.
• skippers receive a sensible default homepage state: met by design! dismissing the dialog lands the newcomer on the default Newcomer Homepage. not observed in testing, since nobody skipped in either round. worth watching in the experiment.
• design for a multivariate test, multi-step and one-step: met! the multi-step version is what we tested. the one-step version is the same three steps stacked in a single dialog or page, with no new content or logic.

decision on components

we observed no meaningful gaps between the two builds beyond touch targets being a bit small in the codex version, so keeping stock codex in the experiment is acceptable. the right move is to bring these learnings back into codex rather than writing and maintaining new growth-specific components.

follow-ups

• give people a one-tap way back to the seed article.
• better handle the empty or thin suggested-edits state, so the Home always says something useful.
• make the seed-article behavior more visible.
• reading-focused content: across both rounds people expected the Home to do something for readers too, especially those who chose "reading and exploring". we are in discussion with the Reader Experience team about a lightweight reading module on the Newcomer Homepage. as few as 3 reading suggestions could go a long way, and it could be populated with the same morelike API we already use for suggested edits.

next

writing the design spec for implementation, so we can run the experiment in production T432812.

assets

prototype: https://aminalhazwani.github.io/protowiki3/pr-preview/pr-4/no-distractions
figma explorations: https://www.figma.com/design/sIvFfvFgrw59lEjtJHlX62/T430418-account-setup--aka-welcome-survey-?node-id=0-1&t=yB8ATFSdsxazdDDe-1