Page MenuHomePhabricator

Answer analytics question: What percentage of users who arrive on the homepage (logged out) log in?
Closed, ResolvedPublic

Description

What percentage of users who arrive on the homepage (logged out) log in?

Note that we can't exactly identify a user, but we can identify visits, so:

  • I've added "Homepage Anon" and "Homepage Login" visit segments in matomo
  • Homepage Anon includes all visits that include an unauthenticated homepage request (/)
  • Homepage Login includes all visits that include a homepage login request (/oauth/login/?next=/)
  • Dividing "Homepage Login" by "Homepage Anon" over the same time period will give you the fraction of visits in which an anonymous user who hit the homepage also the login button.
  • Bots are excluded from these counts

Event Timeline

Note that currently this matches on exact url, so we'll probably want to tweak this to account for users that come in from the echo notification with extra url parameters.

I've updated the visitor segments to no longer match on the Bot/Not-Bot variables since those will no longer be available with the bot traffic filtered out. The hourly cron task will need to run again before the data gets updated, and data from before the update got deployed will include bot traffic.

Moving this back to 'In Progress' as I'm actually still tweaking the homepage login segment.

I realized that our frontpage login button isn't using the standard next parameter to set the post-login route, so we were missing some logins there. That's a completely reasonable choice for the homepage, since login redirects to / when there's no next. We're capturing those now.

I've verified that logging in from non-homepage locations (such as the browse and partner pages) is still excluded.

Note that currently this matches on exact url, so we'll probably want to tweak this to account for users that come in from the echo notification with extra url parameters.

@Samwalton9 There are a few ways we can account for this.

  • We can exclude the notification paramaters globally, so that they are removed from the stored data. This will make echo notification hits look just like any other hits so we can use our visitor segments as is. This is super simple, but would preclude gathering notification-specific stats.
  • We can update the visitor segements to use patterns instead of exact match. This will take a little more time since you have to test text pattern matching pretty carefully to ensure that we get the correct data.
    • If we do this, we could also add separate notification/non-notification segments using patterns if we want separate counts

We can exclude the notification paramaters globally

Let's do this :)

markasread and markasreadwiki added to the global list of query URL parameters to exclude

Data is looking more reasonable (24 logins for 78 visits, 30%), and not seeing any obviously-bot behaviour.