Why I keep coming back to this chart
Of all the charts I look at running my own businesses, the retention curve is the one I trust the most and the one I’m most careful not to over-read. It’s a simple idea: take a group of people who all started using something on roughly the same day, and track what share of them are still around on day 1, day 7, day 30, day 90. Plot that as a line and you get a curve that almost always drops fast at first and then, if the product has real staying power, levels off.
That leveling off is the whole story. A retention curve that keeps sliding toward zero means you’re renting your users, not keeping them. A retention curve that flattens into something close to a straight line means you’ve found a group of people who stick, and you can build on that. Everything else in this article is about learning to tell those two shapes apart, and about what the chart won’t tell you no matter how long you stare at it.
What a retention curve actually is
Start with a cohort. That just means a group defined by a shared start date, usually “everyone who signed up in the same week” or “everyone who made their first purchase in the same month.” You do not mix cohorts together, because the whole point is to watch one group age.
For that cohort, you calculate the percentage still active at fixed intervals after they joined. “Active” has to mean something specific and consistent, logging in, opening the app, making a purchase, whatever action actually indicates use. If you change the definition of “active” halfway through, your curve is comparing two different things and the shape becomes meaningless.
Plot percentage retained on the vertical axis and days (or weeks, or months) since joining on the horizontal axis. Day 0 is always 100%, since everyone in the cohort was active on the day they joined by definition. From there the line almost always falls.
The shape is the message
Most retention curves share a family resemblance: a steep drop in the first few days, a slower decline over the following weeks, and then, for products that hold on to people, a flattening tail. The steep initial drop is normal. Some portion of any group of new users or customers was never going to stick around, they were curious, they were comparison shopping, they hit a wall setting up. What matters is not that this early drop happens, but how far it goes before it slows down, and whether it ever actually stops.
A curve that keeps declining in a straight line, with no flattening, is the concerning one. It means you’re losing a consistent share of whoever’s left, week after week, with no floor. Run that forward and eventually almost nobody from that cohort remains, no matter how large the group started. This shape usually points to something structural, the product doesn’t deliver enough ongoing value to justify staying, or it solves a one-time problem rather than a recurring one.
A curve that flattens, sometimes called an asymptote, means you’ve found a segment of users who keep coming back. The height at which it flattens matters as much as the fact that it flattens. Leveling off at 40% is a very different business than leveling off at 8%, even though both are technically “retention curves that flatten.” For a solo operator, that flattening percentage is often the most honest number you have about whether the thing you built has a durable core.
There’s also a shape you’ll see occasionally called a smile curve, where retention actually ticks back up after an initial dip. This shows up in products people forget about and then rediscover, or in things with seasonal or habitual re-engagement built in. It’s real, but it’s less common than people assume, and I’d treat a smile curve you weren’t expecting as a prompt to check your active-user definition before you celebrate it.
A concrete way to picture it
Say you’re running a small subscription tool and you look at everyone who signed up in the first week of a given month, a hypothetical cohort of 100 people, just to make the shape concrete. By day 1, some have already bounced, they signed up and never came back. By day 7, more have quietly stopped. By day 30, the decline usually slows down noticeably. Whoever is still active at that point has, in effect, decided the tool is worth keeping around, and that group tends to erode much more slowly from there.
The exact percentages will differ for every product and every audience, and I’m not going to hand you numbers as if they were universal, because they aren’t. What is universal is the pattern of checking: does the line keep dropping at roughly the same rate it started, or does it visibly bend and flatten. That bend is the thing to look for before you look at anything else.
Where people misread this chart
The most common mistake is comparing cohorts of different sizes or different acquisition sources as if they were the same group. A cohort of 10 people who found you through a personal referral will behave differently than a cohort of 500 who arrived through a paid ad, and averaging them together hides both signals. If you can, split your retention curves by acquisition channel before drawing conclusions. A blended curve can look mediocre while hiding one channel that retains well and one that doesn’t.
The second mistake is confusing retention with churn and treating them as simple opposites without checking the math. Retention and churn should sum to 100% for a given period, but only if you’re measuring the same cohort over the same window with the same definition of active. If your churn number comes from billing cancellations and your retention number comes from product logins, they’re not describing the same population, and reconciling them takes real work, not a formula.
The third mistake, and the one I’ve made myself, is reading a short curve as if it were a long one. A cohort that looks great at day 7 can still fall apart by day 90. Early retention tells you whether onboarding worked. It does not tell you whether the product holds up over months. Don’t extrapolate a flat 7-day curve into a promise about 6-month retention, the two are related but not the same claim.
What the curve can’t tell you
A retention curve shows you what happened. It does not tell you why. Two products can have identical curves for completely different reasons, one because it delivers consistent value, the other because switching costs are high and people haven’t gotten around to leaving. You can’t tell the difference from the chart alone. That takes talking to the people who left, or the people who stayed, and asking.
It also can’t tell you what to fix. A steep early drop could mean the onboarding is confusing, the pricing is wrong, the product doesn’t match what the marketing promised, or the audience was never a good fit to begin with. The curve tells you where the leak is, roughly, day 2 versus day 60, but it’s a starting point for investigation, not a diagnosis. I treat it as a prompt to go look closer, never as a verdict on its own.
And it’s worth saying plainly: nothing about a retention curve, however good it looks, guarantees future results. A cohort that retained well last quarter can retain worse next quarter if the market shifts, a competitor launches, or you change something about the product without realizing the effect. Use the curve to understand what happened and to ask better questions, not as a forecast you can bank on.
If you want more of this kind of plain-language breakdown of the metrics and methods behind running a small data-driven business, you can find the rest of what we cover on the home page.
Get new guides and videos first — join the Telegram channel.