How To Keep Your Users

Lee recommends tracking groups of new users over time rather than blending the whole user base together. Define cohorts, an action that signals real product value, and a time period that matches intended usage. For example, Instagram might track users who view several posts, while Google Photos tracked users who viewed a photo full-screen.

A cohort chart counts how many users return in each later period, then divides by the cohort’s starting size to show retention percentages. Lee says the key signal is whether the curve flattens, since even a modest stable share can accumulate into a large user base. Early Google Photos cohorts settled at roughly 20% to 40% weekly retention.

Lee warns against choosing time periods that are too broad or actions that are too shallow. At Bump, he widened weekly retention to monthly and then quarterly, making the numbers look better even though the product was meant to be used frequently. He also cautions that opening an app, clicking a notification, or merely paying may not show that a user is getting value.

Refresh retention charts every week or two, then investigate weak cohorts by changing the product, reaching better matched users, improving onboarding, or strengthening network effects. At Google Photos, a push to attract Gen Z brought in users with poor retention because the product’s memory and reminiscence use case fit them less well. Cohort slices by country, device, or customer type can reveal where the mismatch lies.

The strongest pattern is retention curves that flatten and then rise, with newer cohorts improving. Lee’s layer cake chart shows active users by month and by the cohort they came from; thick layers from older cohorts indicate that growth includes users who stayed, not only recent signups. If curves do not flatten, he advises talking with customers to understand what needs to change.