Page MenuHomePhabricator

Drop support for the old Impact module
Open, LowPublic

Description

Since we stopped A/B testing the new Impact module in T336203: Positive reinforcement: Deploy the new Impact module to all Wikipedias on all Wikipedias, we should start the process of dropping the support of the oldimpact Homepage variant, which includes dropping A/B testing itself and dropping the old Impact module code as well.

(note we're A/B testing the new Impact module, but for technical reasons, the treatment group is actually called oldimpact and is given the old Impact module in the code; hence the confusing title)

Checklist
  • Confirm with @KStoller-WMF that we will not need the old Impact module in the future
  • Drop the oldimpact homepage variant (makes the old Impact module class unused)
  • Drop the old Impact related code
  • Possibly rename NewImpact to just Impact, since the distinction makes no sense to keep

Event Timeline

I think we should be adding this to our workstream fairly soon, since the new Impact module is meant to be deployed to all Wikipedias very soon (Nov 01, to be precise).

@KStoller-WMF Can you confirm we will not need to switch back to old Impact for product reasons in the future? Or should we keep the capability in our code in the meantime?

Thanks for adding this task!

From my point of view, there is no need to switch back to the old Impact module in the future. But I can imagine two scenarios where we might consider switching back:

  • There is a major community complaint about the change
  • The new impact data doesn't populate or the feature breaks in some way.

The first scenario seems unlikely. If engineers think the second scenario is fairly unlikely as well, then I think we should move this task forward. Perhaps we should just wait a sprint or two first?

Thanks for the quick re! I think in the unlikely case the second scenario happens, we should prefer fixing the bug, rather than reverting to old Impact. Waiting one sprint sounds good to me.