Download the graph data: enwiki202405.csv
This graph has versions that count in each month only users that made at least 1, 10, 100 or 1000 edits in the month.
What the graph shows?
The graph has 3 axes, horizontal, vertical and the color axis. Each horizontal line groups all users that had their first edit in the same month, each "pixel" of that horizontal line is how many users of that group had edited in a specific month. The graph is triangular because no user can have edited before the month of their first edit.
The user retention can be verified by the amount of users of that group that have kept editing months after their first edit, so the more red pixels in a horizontal line the more user retention for the users that had their first edit in that month.
Besides the original graph you can also see graphs that consider a minimum amount of edits in each month. If an user made less than that amount of edits in that month that user is not counted in that specifc month.
How the graph can be used?
We can see in the graph that some months have a slightly more user retention than others, that is perceived by horizontal and vertical lines where the colors are a little bit higher or lower in the color scale related to the colors of the lines arround. Those horizontal lines suggest that some event in that particular month has attracted (or moved away) new users with the potential to keep editing for long time, in other words, that can indicate whether an event attracted significant new long-term users. Vertical lines are more rare and more subtle, they indicate that some event in that month attracted (or moved away) old users.
The graphs that consider a minimum of edits in each month are useful to remove from each "pixel" users that made only eventual edits like an orthographic error fix. If an user keeps editing at the rate of 100 edits or more per month for example, that is a more robust collaboration than an user that make 1 or 2 edits per month. So it is possible to evaluate whether an event have attracted new users with a potential high level of collaboration.
About the data
The data consider non deleted edits of non bot users in any namespace. Users are considered bots if they have bot flag or if their user page is in a category linked to d:Q3681760, in wikis without that category it is also considered users that match the regex "[Bb][Oo][Tt]\b".
The edits imported from other wikis were not excluded from the data, unfortunalety there is not an easy way to do that.