I corrected the dependencies of this task to reflect the fact that the test was restarted, see T112585 for more information.

Working on T115274 because that will greatly improve the quality of the analysis in this task and the analysis of the results of the test we're launching soon.

Report in progress: https://github.com/wikimedia-research/CompletionSuggestionTest/blob/master/analysis/Report.md

Aiming to finish on Wednesday.

Pre-final draft: {F2752600}

Awaiting review from @Ironholds

• in the opening para, some kinds of pages, not kind
• R is software? ;p
  
  I'm not comfortable saying it's our recommendation we switch to the new method. We can see that users have a greater outcome of getting results. This does not equate to a better outcome. The new approach may increase the number of results, for example, while also decreasing the quality of those results. I'd like us to look at least at clickthrough rates as a basic satisfaction heuristic before we recommend anything.

Final draft: