Types of Retention: which type do you prefer?

[saved1994]

Four varieties of retention

Day N rolling retention. This is the percentage of users who returned to the application on day N from the time of installation or later. For example, one user returned to the application on the 14th day, and on the 45th day, and the second user returned on the 44th day, then when calculating the day 14 rolling retention, they will be considered as two users who returned on the 14th day.

Day N return retention. This is the percentage of users who returned at least once in N days. For example, when calculating day 21 return retention, all users who opened the application at least once on any day from the first to the twenty-first will be taken into account.

Day N bracket-dependent return retention. This is the percentage of users who returned to the application at least once from day M to day N, where parameter M limits the user return time. For example, if M is 14, then day 20 bracket-dependent return retention will show the percentage of users who opened the application from 14 to 20 days from the moment of installation. In fact, this is a special case of the day N return retention.

Day N full retention. This is the percentage of users who were opening the application every day until day N. For example, day 5 full retention is the percentage of users who opened the application on the 1st, 2d, 3d, 4th, and 5th day from the date of installation.

 

[teeeecher1]

@saved1994 Day N rolling retention as you called it is the most common, especially when graphed.

It helps show where the drop occurs and if there are any patterns in users behavior (eg. if the graph goes up on day 14 or day 30 it means there may be a fortnightly or month trigger for users to come back. For example an retention on an accounting app might have spikes weekly, fortnightly or monthly when people have to do payroll and come back to finish setting it up or try to use it for a pay run.