Retention over time

In this analysis, we agggregate daily retention

\[ \frac{\text{number of new users on date for days from start}} {\text{number of users on the date - days from start}}, \]

by month, quarter, year.

Notice how the confidence intervals fully overlap, when aggregating, and the variability in examining daily retention. This demonstrates the underlying truth in the simulated data, that there is no true difference in retention probability from day to day. Again, without confidence intervals, it’s not reliable to interpret the differences in the data.