Most teams judge retention by a single number: Day 1, Day 7, maybe Day 30. That habit throws away the most useful information you have. Two apps can share an identical Day 1 retention and have completely different futures, because the shape of how their users leave is different. The curve tells you that. The point does not.
Learning to read the curve, not just a point on it, is the difference between knowing your retention and understanding it.
What is a retention curve?
A retention curve is a chart that plots the percentage of an install cohort still active against the number of days since install.
- It starts at 100 per cent on Day 0, when the whole cohort is present by definition.
- It declines as users drop off, steeply at first, then more slowly.
- Each point is a retention rate for that day, built from the same cohort logic covered on the Linkrunner retention documentation.
The curve is just those daily points joined up. Its shape, not any single dot, is what carries the meaning. If you need to settle how a cohort is defined first, our guide on how to define cohorts is the place to start.
How to read the shape, not the point
Every retention curve has three regions, and each says something different:
- The early cliff (Day 0 to Day 3). Nearly every app loses most installs here. A steeper-than-normal cliff points to an onboarding or first-session problem.
- The slope (roughly Day 3 to Day 14). How fast you keep bleeding users through the first two weeks.
- The tail (Day 14 onward). Whether the line keeps falling toward zero or levels off.
A single Day 7 number tells you the height of one dot on this line. It cannot tell you whether you are still falling fast or have already stabilised, which is the thing that actually predicts long-term value.
The flattening point and the retention floor
The most important feature of any retention curve is where it stops falling:
- The retention floor is the height at which the curve flattens into a roughly horizontal line. It represents your sticky core: the share of each cohort that has formed a durable habit.
- The smile is when the tail ticks back up, a sign that surviving users are re-engaging and even lapsed ones are returning.
- A curve that never flattens and keeps sliding toward zero means you have no durable base, however good the early numbers looked.
This is where the audit pattern bites. Two apps at the same Day 1 retention can flatten at wildly different floors, one at 15 per cent and one near zero, and the first is a real business while the second is a leaky bucket. The floor, not Day 1, is the number to protect.
Comparing curves by cohort and by source
A curve on its own is useful. Several curves overlaid are far more so:
- By cohort over time. Plot each month's cohort to see whether product and onboarding changes are lifting the floor.
- By acquisition source. Overlay curves per channel to see which source holds a higher plateau, not just which wins on a single day. That is the channel worth scaling, as we explore in cohort analysis techniques for growth teams.
- Watch the early slope by source. A sharper Day 0 to Day 1 drop from one channel usually means its users hit friction before reaching value.
Turning the curve into action
Different parts of the curve call for different fixes:
- Steep early cliff: work the first session. Cut steps between install and the first moment of value. Our onboarding metrics that predict long-term retention shows where to look.
- Fast mid slope: build early habits with behaviour-triggered nudges, not generic pushes.
- Low floor: the core loop itself needs work; re-engagement alone will not save a product users do not want to return to.
Frequently asked questions
What is a retention curve? It is a chart plotting the percentage of an install cohort still active against the number of days since install. It starts at 100 per cent on Day 0 and declines over time.
What does a healthy retention curve look like? It falls steeply in the first few days, then flattens into a horizontal plateau. The plateau shows a stable core of committed users, which matters more than the height of any single day.
What is the retention floor? The floor is the height at which the curve levels off. It represents the durable, habitual share of each cohort and is one of the strongest predictors of long-term value.
How do I compare retention curves by acquisition source? Overlay one curve per channel using cohorts of the same age. The channel whose curve holds the highest plateau over time is delivering the most durable users, regardless of which looked best on Day 1.
Read the shape, protect the floor
A retention curve turns a flat number into a story: a steep cliff, a slope, and a floor. The floor is the part that predicts your future, so that is the part to defend. Judge onboarding by the cliff, habit-building by the slope, and product-market fit by where the line settles.
If you want retention curves you can slice by cohort and by acquisition source without exporting data by hand, book a demo with Linkrunner and read your own curves the way this post describes.