Jump to content

Research:Onboarding new Wikipedians/Logging

From Meta, a Wikimedia project coordination wiki

Notes on the funnel analysis for onboarding...

welcomecreation actions


E3Experiment's onboarding.js logs events on the welcomecreation page (shown by Special:Userlogin?type=signup after successful account creation), against Schema:Onboarding The user should be not-anon ("authenticated") and have a userName, since she just created an account.

  • welcomecreation-impression (user arrives on the page)
  • welcomecreation-cta-click (click the link to Special:GettingStarted)
    • now that the end of account creation is the GettingStarted page this link is gone, yet is still getting hits !!
  • welcomecreation-returnto (user clicks the [← No thanks, back to article] link)

Getting Started actions

This is the same page as welcomecreation, so events are duplicated

Events on Special:GettingStarted are exactly like Community portal clicks, so they are logged against Schema:OpenTask. There are some additional actions

  • gettingstarted-visit : user arrives on page.
    • TODO need to cleanly distinguish user seeing the page from account creation vs. visiting the page Special:GettingStarted.
  • gettingstarted-communityportal-click : user clicked "More articles..." link
  • gettingstarted-returnto-click users get the always the welcomecreation-returnto-click above.

TODO could have a special action or field when the user's gettingstarted-visit came from account creation as opposed to returning to it. You can work it out by looking at the referrer, but code can be more direct.

standard OpenTask actions


If the user clicks any of the pages (from by the opentask-short template) list, openTask logs the existing accept-edit-complete actions, but instead of the action having a task 'Add wikilinks' or 'Fix spelling and grammar', the action has a task of 'gettingstarted-edit'


It would be easy to apply OpenTask handling to the returnTo page, so that on click it's added to the openTask cookie and gets accept-edit-complete logging. But there are bugs:

  • must strip out " (page does not exist)"
  • don't do this for uneditable pages (Special: pages, etc.)
  • doesn't work for pages in namespaces, maybe don't do it if the page isn't in the main namespace