Research:Onboarding new Wikipedians/Preliminary data
Analysis for the period of December 13th - January 10.
Early data samples, Dec. 20th - 21st
[edit]Results for the first 25 hours of data collection (December 20 21:53 - December 21 22:53 UTC). This data is not yet deduplicated by token, meaning they don't represent a unique number of users and shouldn't yet be used for conversion rate analysis.
| Impressions on Special:GettingStarted | Clicks on "No thanks, take me back" | Clicks on tasks | Edit attempts | Successful saves |
|---|---|---|---|---|
| 3118 | 1114 | 171 | 114 | 38 |
| Impressions | Clicks on tasks | Edit attempts | Successful saves |
|---|---|---|---|
| 4708 | 118 | 11 | 8 |
Due to lack of SQL importing and the fact that instrumentation was in flux during the period prior to Christmas, we deferred metrics analysis until post-December 26th.
First funnel analysis
[edit]We completed early rounds of basic funnel analysis, calculating conversion rates for unique users for each step. We did conversion from the total percent of impressions, and also as a percent retained from the previous step in the funnel. The first round was for December 26-31st, the second for Dec. 26-January 10.
You can view this analysis, but the data should not be trusted as conclusive. We used revisions pulled from this round, which are not subject to instrumentation errors, for qualitative analysis. However, as you can see from comparisons of impressions and clicks in later analysis, we are likely underestimating clicks on any links from Special:GettingStarted by a large amount, thus making step-by-step funnel retention analysis dubious at best at this stage.