QA summary of test data by 2023-11-30 updated in the description.

Instrumentation approved by legal and privacy review.

In T348613#9303544, @mpopov wrote:

@cchen @PWaigi-WMF: can you please confirm that Legal & Security reviewed and approved the data collection / measurement plan for this experiment?

conflict.

@SBisson I've submitted updates to the schemas.

Below is the deletion and upload data from June 2023 - August 2023, the numbers exclude bot and mobile uploads.

@SBisson thanks for reviewing the requirements. I will go ahead and update the ticket and the schema.

Completed the review of 4 schemas in the google sheet:

• wikistories consumption event
• wikistories contribution event
• mediasearch interaction
• searchpreview

Future data will be updated in this sheet https://docs.google.com/spreadsheets/d/1qR6yPFktt-DTfETFJD50a2ooVbPe6Ad9m1S3fEdK2kE/edit#gid=0

Below are the deletion rates of media uploaded within a given month that is flagged for deletion within 30 days of being uploaded in May, June and July 2023.

The users who click the ads will be split 50/50 into control and experiment groups for this experiment. Our main metric will be users' median time spent on site, and we decided that the smallest change in it that we would consider meaningful is 20%.

Blocked by https://phabricator.wikimedia.org/T344071

@AUgolnikova-WMF I add list of moderators that made deletions in the past 6 months.

According to the GroupRights page on Commons, only admins (sysops) can delete pages, so we are searching for admins (sysops) user groups for the moderators.

Adding data for the overall deletions rate for the last 12 months by the end of June 2023, including deleted files and outstanding deletion requests

update a summary of June to July search previews data with the additional wikis added, the detailed number can be found in this dashboard

• Opt-out Rate: Very few users opt-out the search preview across wikis, the opt-out rates are all below 0.01%.
• Number of Previews Opened: The numbers of previews opened on mobile are twice as much as on desktop for idwiki, ptwiki, huwiki. For ruwiki and ukwiki, these numbers are almost triple on mobile. On nlwiki, the number of previews is 15% less on mobile than on desktop. And on cawiki and nowiki, these numbers are similar on desktop and mobile.
• % of Sessions with Preview Opened: The percentage of search sessions where the user clicked at least one search preview are almost doubled on mobile compared to on desktop across all the wikis. Idwiki has the highest open rate on desktop (~7.4%) and ptwiki has the highest open rate on mobile (~14.1%). Ruwiki has the lowest open rate among all the wikis on both desktop (~3.6%) and mobile (~8.1%).
• Average number of Search Previews opened: The number of previews opened per session on mobile is 30% higher on mobile than on desktop in all 3 wikis.
• Click-through rates: the click through rates of all the sections in the search preview are still higher on mobile than on desktop across all 8 wikis.
  • Snippet section has the hightest click-through rate, while all other elements have click-through rate below 5%.
  • Interwiki Commons started to have a large increase in click-through rate from March on ptwiki and idwiki compare to previous months.
  • ptwiki has the highest click-through rates on the section on both desktop and mobile except the mobile interwiki commons and interwiki links.
  • Ukwiki and huwiki have the lowest click-through rates on most of the sections among all the wikis.
• Average Click Position: The average click positions are 0 (first result) on desktop and 1 (the second result) on mobile.

Adding data for deletion requests in the queue:

I queried the notification data, looks like GROUP BY echo_notification.notification_bundle_hash works for the case. i will close this ticket.

Thanks for the examples. The queries i used before for the notification data is

@matthiasmullie thanks for the details! to confirm, for combined suggestions, multiple events entries have the same event_id as well right?

Month over month deletion rate from January 2023 to May 2023:

@AUgolnikova-WMF Please find the baselines for upload wizard improvements below. The data is from April 2023 and May 2023, the detailed number can be found in this notebook

@AUgolnikova-WMF I see the deletion rates increase over time because a file may be deleted 2 months or longer time after uploading. In this case, we can set a timeframe of deletion for the metrics similar to revision. For example, total number of filed deletion requests within 30 days (or 60 days) of uploading monthly

Added new lists to the spreadsheet with users:

• have their email feature enabled
• have their mail confirmed
• not from any block groups
• not include users from the previous list

@Sneha Please find the list of users here
https://docs.google.com/spreadsheets/d/1GHynurGnIbVPqni5sdI0Mk9tglwDwJB6DW0M7yv5hDU/edit#gid=1733538878

Between Spanish and Arabic communities, which one is more active on Commons?
There's no direct way to determine what language users are using while doing the uploads. Per discussion with Sneha, I took a look at the number of files with captions in Spanish and Arabic. In Commons, there are 373K files with Spanish captions and 42K files with Arabic captions.
I also compared user counts. For those who made uploads in April, ~27K of them have accounts in Spanish language projects and 8.7K of them have accounts in Arabic language projects.
Spanish has a larger user and content basis on Commons in this case.