Retention Curve Creator


Fit a regression curve to your data


Enter your app's observed retention - from any cohort you like - to generate a model.


     

*Provide at least two observations. To exclude a data point, set it to 0%.

**A D30 retention observation is optional, but including that data point can greatly increase the accuracy of long-term predictions.


The Details


Day 1 (D1) retention is the percentage of your users that return to the app the day after they install. In general, day n (Dn) retention is the percentage of users who return exactly n days after they install.


The retention curve is typically a function of the form r(n)=anb, where each app has its own values for the coefficient a and the exponent b that best fit their data. Dn retention is modelled (predicted) by r(n) for any value of n.


For example, if a game's retention curve is defined by the function


r(n) = 0.396 * n -0.472


then D120 retention is predicted to be


0.396 * 120 -0.472 = 0.041 = 4.1%


In other words, if 1,000 people install today, 41 are expected to use the app exactly 120 days later.


The summation of the retention curve from D0 to Dn defines the expected player days; i.e., the average number of distinct calendar days a user interacts with your app over that timeframe. A curve fit to this data is used to predict both lifetime value and daily active users. For this particular retention curve, the function


PD(n) = 0.0 * n 0.0


best fits the data. Click to expand the graph in a new tab.


a player-days curve fit to data