Skip to main content

 I compare weekly N-day and unbounded retention for the same segment, same starting and return event and same period, I often get the longer retentions (W+3 here) with a lower unbounded than N-day retention. But If unbounded retention counts all the users who have returned on Week 3 of later, shouldn’t it be at least as many as the number of users who returned exactly on Week 3?

N-Day retention (weekly)
Unbounded retention (weekly)

 

Hi Sebastien,

Happy to explain this! It is possible to have a lower Unbounded Retention than N-day based on the way we calculate the values.

When computing the overall N-Day retention percentage for Day/Week/Month X, we only include users who have had a complete Day/Week/Month X in the denominator. The reason we do this is to avoid artificially deflating overall retention. In your specific case, the number in the All Users column for the overall segment is the number of users with a complete week 0. Any user who has not completed week 0 cannot be included. 

For Unbounded Retention, Day/Week/Month X represents the time range of X Days/Weeks/Months to the current time. Thus, a user has a complete Day/Week/Month X and is counted towards the overall Unbounded Retention as long as X Days/Weeks/Months has passed since the starting event for that user.

In the Unbounded Retention chart more users qualify for the chart due to the way Unbounded Retention is calculated. Since the volume of users is higher in the Unbounded Retention chart and the number of users who were retained remains the same, the percentage is expected to be lower here.

I would also note that week 3+ has a dotted line to indicate that users in this time range still have the possibility to convert. As time passes and more users pass week x the chart will have more users to include. 

Hope this helps let me know if you have more questions!


 

Reply


Terms & ConditionsCookie settings