Advanced Retention Functionality

Retention Criteria - On Retention (formerly, N Day Retention)

"On" Retention calculates the percentage of users who come back on a specific time unit (e.g. day, week, month).

After a user completes the “A criteria”, they will belong to a time unit cohort that corresponds with when they did the "A event" and the Retention report will start the clock counting from the time that particular user performed that "A event". When using "On" Retention, that same user must return to fulfill the “came back and did B criteria” on a specific and exact time unit relative to when they did the A criteria to be counted as retained.

For example, day 5 retention is the percentage of users who “came back and did B” exactly on the fifth day after they did the A event.

"On" retention is useful to understand high-level usage patterns.  We also recommend using "on" retention when your product relies on users returning in each and every time unit.

Retention Criteria - On or After Retention (formerly, Unbounded Retention)


"On or After" Retention calculates the percentage of users who come back on a specific time unit (e.g. day, week, month) or any time unit afterward. A user can fulfill the "came back and did B criteria" on a specific time unit or any time in the future to be counted as retained in "On or After" Retention.

"On or After" Retention gives you a better sense of how long you are holding on to your users in an absolute sense. It will tell you how many users used your app and then ever returned to find more value. In other words, it's the opposite of the overall churn of your user base.

We think "On or After" retention is a better fit for most businesses and it is the default calculation for the Retention report. This is because it’s more common that a product doesn’t require users to come back in each time unit to achieve the product’s value proposition, and more often than not when doing retention analysis we don’t only value the user coming back in a specific time unit as long as they eventually do. However, if this is important to you and your business, please utilize "On" Retention.

Custom Retention Brackets

Not all retention analysis can be done with a set, repeated cadence and it is not always important to measure how a user came back relative to each and every particular time unit.

Custom bracket retention gives you the flexibility to split up your retention buckets into customized intervals of multiple days, weeks, or months to create meaningful touchpoints as you see fit.

Use Cases

For example, let's answer an important mobile gaming question: How often do my users came back and play a game between 15 and 30 days after sign up to represent mid-term players?

**In mobile gaming, industry-standard intervals typically examine retention for users that came back and play a game on Day < 1, Days 1-3, Days 4-7, Days 8-14, and Day 15-30. 

Typical "exact day" retention buckets don't quite solve for this use case because users should be able to play a game on any day between day 15 and 30 and be counted in a single bucket to understand how many users were retained in that entire period. Exact day retention will log users in each individual day bucket (15, 16, 17, ..., 30) which, while more granular, doesn't give a retention percentage for the entire interval. Instead, we want to know how many users played a game at all in any of the days 15-30.

Free trials: If you want to understand how many users come back to use the product during a 7-day free trial and also who came back and kept using the product in the 30 days after the free trial is up, you could do something like this:


In between cadences: Are you somewhere in between a DAU or WAU product? You can group daily buckets to understand how your users retain every 3 days (bi-weekly).


Setting Custom Brackets

First, go into the time unit drop-down in the Retention query builder and select Custom


Then, select the size of each bracket for the given time unit. In this example, we are selecting the number of days that are in each custom bracket. Apply.


Each of these intervals is its own retention bucket, and if the user does the "come back and did B" event anytime inside that bucket they are counted as retained. Now I will be able to understand what percent of users come back to play at least one game in any day between day 15 and day 30 from signing up. 


You can use the Frequency report to learn how frequently users return to use your product or service.

To generate Frequency reports, click Reports, Retention, and select the Frequency option under the toggle for retention.


Mixpanel groups unique users in time-incremented buckets when they first complete an action, and then groups those same users in subsequent buckets when they return and perform the same or different actions.

The buckets measure the number of unique time buckets a user was active. Depending on the interval chosen, this will one of:
  • unique hours in a day
  • unique days in a week
  • unique months in a year
This will show how engaged users are with an action. In the example below, you can see how many users did the event "Complete Purchase" on unique days in a week. For users who did "Complete Purchase" on the week of Jan 9, 2020, 13.04% of them did "Complete Purchase" on 4 unique days that week.

Frequency Criteria

Cumulative Frequency: calculates the number of users who did the event in at least X unique intervals. This shows overall engagement up to that up to each frequency interval.

Non-cumulative Frequency: calculates the number of users who did the event in exactly X unique intervals. This is useful for determining exact engagement and for further investigating specific groups of engaged users.

Did this answer your question?



Please sign in to leave a comment.