@GGalofre-WMF

I've added the requested workflow steps to the funnel to visualize editor behavior immediately after title population. This helps us understand the high drop-off at this stage by tracking how users respond to specific notices (i.e. title conflict, unsupported subject, and subject covered).

Results are included under the "Investigating the drop off at the title population step" section of the report.

I focused these visualizations on this step, as including these specific validation paths in the full Sankey diagram created too much clutter and was hard to read.

See a summary of findings for Junior Editors below (overall results as well as breakdowns per experience level, per platform, and per wiki are included in the report).

• The single largest drop-off in the workflow still occurs immediately after the title is populated. 37.1% of all Junior Editor sessions (2,904) abandon the process here ("Drop: Exit at Title Input") before advancing to topic selection and without triggering any validation notices.

• 15% of article guidance attempts by Junior Editors are shown one of the validation notices (.e. title conflict, unsupported subject, and subject covered).

• The "Unsupported Subject" notice is the most common notice shown after the title is populated. It appears in 8.0% of total Junior Editor sessions. However, the majority of users (60.1%) who see this warning bypass it by clicking "Start Writing" rather than abandoning the flow.

• The "Subject Covered" notice was shown in 5.1% of Junior Editor sessions. More than half (59.4%) of the users who saw this notice chose to "Improve," successfully redirecting their effort to an existing article.

• "Title Conflict" notices are the least common friction point, occurring in only 1.9% of Junior Editor sessions.

Let me know if you have any questions.

Thaks for this @MNeisler!

Looking at "valid title" vs "drop:exit at title input":

• Does it mean that for 2634 users in "valid title" their title search was valid, producing wikidata match results?
• Does it mean that for the 2904 users in "drop:exit at title input" their title search did not produce a valid result? or that the 2904 users in "drop:exit at title input" exited the workflow regardless of the result of their title search?

• Do we know how "immediately" they left?

If 2904 users in "drop:exit at title input" left after no results were shown, that points to the wikidata mechanism being a potential cause. If they left without even having time of scanning any workflow behaviour / results, that may just mean they didn't want to create an article to begin with.

Does it mean that for 2634 users in "valid title" their title search was valid, producing wikidata match results?

"Valid title" here just means the article creation attempt did not include any of the three instrumented warnings (title conflict shown, subject covered shown, or unsupported subject shown) and proceeded to the select a topic step. It does not necessarily mean Wikidata match results were shown. I replaced "Valid title" with "No Warnings Shown" in the sankey diagrams to help clarify what this specific proportion of users represents. Let me know if any other adjustments would be useful.

To know if Wikidata results are driving the drop-offs at the title step, we would need to pull in an additional event.

Do we have an event instrumented here that indicates if the initial title search actually produced a match? We currently log if a user selects a suggested Wikidata topic further down the funnel (the "Suggested Topic" step), but is there an earlier event we can add right at the Title Populated step? I see an event action called wikidata_connect but this event appears to show up much later in the funnel based on associated timestamps when I looked through a raw sample of events.

Does it mean that for the 2904 users in "drop:exit at title input" their title search did not produce a valid result? or that the 2904 users in "drop:exit at title input" exited the workflow regardless of the result of their title search?

The latter. These are users that did not receive one of the three instrumented warnings (title conflict shown, subject covered shown, or unsupported subject shown) and dropped off before proceeding to the topic selection step. It does not necessarily mean that their title search did not produce a valid result - just that they were not shown any of the three warnings and did not proceed to the topic selection step.

Do we know how "immediately" they left?

No. Because we currently track progression events rather than explicit exit events, we do not have a timestamp for when the user actually closed the window or abandoned the workflow. We would need to instrument a specific "abandon" event to measure the exact time spent on this screen before dropping off.

Thanks for that further clarification!

If we instrument a "wikidata result shown" and "abandon" timestamp, once we've switched from experiment to instrument in TR and Simple EN, we'd then be able to monitor (via a dashboard or visualization method that today still does not exist) this information starting after instrumentation happened, on a week by week basis, right?

And with that, we'd be able to see:

• how many users were shown a wikidata match and how many were shown no results.
• how long each group took to leave on average.

Something like:

Out of those 2904 users, 50% were shown no results and took 2 sec to leave, 50% were shown at least a result and of those 40% took 5 sec to leave, 60% took less than a second to leave.