How to read a retention curve
A retention curve is a line chart that plots retention rate on the vertical axis against days since install on the horizontal axis. It starts at 100% on Day 0, when the whole cohort is by definition present, and declines as users drop off. The shape of that decline tells you far more than any single retention number, because it shows the pattern of how users leave, not just how many.
Nearly every retention curve falls steeply in the first few days, since most installs never return after their first session. What separates a healthy app from an unhealthy one is what happens next. A curve that keeps sliding toward zero means you have no durable user base. A curve that flattens into a horizontal line means you have found a group of committed users who keep coming back.
Why the flattening point matters
The point where the curve stops falling and levels off is often called the retention floor or the "smile" if it ticks back up. That plateau represents your sticky core: the share of each cohort that has formed a durable habit. Its height is one of the most important numbers in your business, because those users drive the bulk of long-term engagement, lifetime value, and word of mouth. Two apps can share the same Day 1 retention yet have completely different long-term value if one curve flattens at 15% and the other keeps decaying to near zero.
Reading the curve by acquisition source sharpens this further. When you overlay retention curves for different channels, you can see not just which source retains best on a single day, but which one holds a higher plateau over the long run. That is the channel worth scaling.
Using retention curves in practice
Compare curves across cohorts to see whether product and onboarding changes are lifting the plateau over time. Compare curves across acquisition sources to find the channels that deliver durable users. And watch the early slope for onboarding problems, since a sharper Day 0 to Day 1 drop usually points to a first-session experience that is losing users before they reach the app's core value.