@Pginer-WMF @PWaigi-WMF Here is the summary of the first iteration of the analysis:

As the Contribute menu entry point was enabled during the second and third weeks of November 2024 on more wikis, there was a spike in its usage to access the dashboard immediately after enablement. Post the initial spike, the usage dropped during the first two weeks of December 2024 (8 daily sessions on average), and followed by an increase during the last two weeks (12 daily sessions on average). Although there is a spike in usage after the deployment, it is not substantial compared to the number of new wikis (over 80) where the entry point was enabled on.

~24% of the sessions were started with sources that are independent of the dashboard open entry points (article is not predetermined), and users of these sessions had a chance to interact with the features on the dashboard home. When a similar analysis was conducted in July 2024, only 10% of the sessions started with an independent source, which is a ~14% increase in users opening the dashboard without an article predetermined.

Overall, users who open the dashboard independently (i.e. without an article being predetermined) have higher transition rates to the next steps of the funnel compared to users who reach the translation start step from an external entry point, where an article has already been selected. The difference is 10–15 percentage points for editing steps (such as opening the editor and making at least one edit) and almost double for publishing steps. For example, in 33% of the sessions where users select an article from the dashboard home and start a translation, were able to successfully publish the article, whereas for sessions where the article is predetermined from an external entry point, it is only 17%.

For the period between 1 October and 31 December, the instrumentation for the custom suggestions menu was deployed during the last week of November. Since then, user interactions with the custom suggestions menu have been low, averaging around 3 clicks per day to open the custom suggestions menu.

• There were approximately 50 sessions in which users clicked to open the custom suggestions menu Of the selected filters, collections (such as WikiProject lists or campaign lists) were the most selected, followed by topic areas.
• However, we observed that more people are now reaching the dashboard directly (without an article being predetermined), compared to before – although this should have increased the interactions with the custom suggestions menu, that doesn't seem to be the case.

For users reaching the dashboard directly, in cases where an article is not predetermined,

• 74% of users start a translation by selecting an article from the search results.
• After users reach the dashboard, they search for an article and proceed to the translation start step.
• For new users, the New Translation button—which leads to search and is located at the top of the screen—often has been the first-click action.
• After search, suggestions based on recent edits are the most used source, followed by suggestions provided in the absence of a seed article.
• These suggestions are present when users open the dashboard and are selected without further customization.

The full report is available at https://analytics.wikimedia.org/published/reports/content_translation/custom_suggestions_analysis_iter1_T381391.html

At the moment, we don't have enough data to fully answer some of the questions mentioned in the ticket. For example, number of translations started based on new custom suggestions is in the order of tens, whereas other sources are in the order of thousands, which makes comparison difficult to draw any conclusions. After reviewing these initial insights, we can do a follow-up after there is more data.