Free tool

Retention curve calculator.

Paste the number of users still active in each period after signup and get the curve, plus the only reading that matters: does it flatten, or does it keep falling. A curve that flattens means a base that compounds. A curve that never does means acquisition is filling a bucket with a hole in it.

Your cohort

Start with the size of the cohort, then one number per period. Commas, spaces or one per line all work. Paste a column straight out of a spreadsheet.

Match this to how often someone would naturally use the product.

Share of the cohort still active 7 weeks
Retention curve: the share of one cohort still active in each period after signup. 100% flattening near 35% week 0 week 7

Your curve is flattening, settling around 35%. The last three periods lost 1.4, 0.8 and 0.6 points, so roughly a third of every cohort looks like it stays. That is a real retained base: each new cohort adds to it rather than replacing the last one.

Retention by period
Period Still active Retained Change
Week 0 1,000 100.0%
Week 1 620 62.0% -38.0 pp
Week 2 480 48.0% -14.0 pp
Week 3 410 41.0% -7.0 pp
Week 4 380 38.0% -3.0 pp
Week 5 366 36.6% -1.4 pp
Week 6 358 35.8% -0.8 pp
Week 7 352 35.2% -0.6 pp

Settling near

35%

Leaving each week

1.7%

Lost by the end

64.8%

Reading it

The shape matters more than the level.

It flattens

The curve falls steeply, then settles. That flat part is a group of people for whom the product genuinely works, and it accumulates: every new cohort adds another layer to the base. This is the shape you are looking for, and the level it settles at matters less than the fact that it settles.

It keeps falling

Roughly the same percentage leaves every period, so every cohort eventually reaches zero. More acquisition does not fix this, it just makes the leak more expensive. Activation and the core product are where the work is, and paid acquisition against this shape loses money reliably.

It falls off a cliff in period one

A very steep first drop with a decent tail usually means an activation problem rather than a product problem: people arrive, do not reach the point where it becomes useful, and leave. Worth separating the people who completed a first meaningful action from those who did not, and drawing two curves.

It smiles

Rare and excellent. The curve dips then rises, because returning users pull the number back up. If you see this, whatever caused it is worth understanding in detail.

One curve for the whole product is the blended average of every audience you have. The version that changes decisions is one curve per acquisition channel, because the people who arrive from a recommendation and the people who click an ad are not the same people.

Questions

Retention, asked plainly.

What is a retention curve?
A retention curve plots what share of one cohort of users is still active after each period. You take everyone who signed up in a given week or month, then count how many are still around one period later, two periods later and so on. It always starts at 100 percent and falls; what matters is whether it stops falling.
What does it mean when a retention curve flattens?
A flattening curve means you have found a group of users for whom the product genuinely works. They stop leaving, and that share of every future cohort accumulates into a base that compounds. It is the single strongest signal of product-market fit available from usage data, and it is the difference between growth that adds up and growth that leaks away.
What if my retention curve never flattens?
If the curve keeps falling at roughly the same rate every period, eventually every cohort reaches zero. More acquisition will not fix that: it makes the leak more expensive. This is the case where working on activation, onboarding or the core product beats working on marketing, and where paid acquisition is close to guaranteed to lose money.
What is a good retention rate for SaaS?
It depends on what you sell and how often people need it. As a rough guide for subscription software, a weekly curve that flattens somewhere above 30 percent is healthy, and above 40 percent is strong. The absolute level matters less than whether the curve flattens at all, because a flat 20 percent beats a slowly declining 35 percent over any long enough period.
Should I measure retention weekly or monthly?
Match it to how often someone would naturally use the product. Weekly for something used most days, monthly for something used a few times a month, and monthly for anything billed monthly. Measuring a weekly curve for a product people open twice a quarter produces a frightening chart and no information.
Why does retention by acquisition channel matter?
Because channels bring different people. A single blended curve averages the audience that found you through a genuine recommendation with the one that clicked an ad out of curiosity, and those two retain very differently. Splitting the curve by the channel that brought each user is usually the moment the paid acquisition question answers itself.

Pasting one cohort into a box gives you one curve. The version that decides where your marketing budget goes is one curve per acquisition channel, computed every day without you exporting anything. That is what Ripples does. One script tag, your billing provider and your Google Ads account, and these numbers are computed from real data instead of typed in from memory.

See how that works

Other free tools

UTM builder CAC payback period calculator LTV:CAC ratio calculator ROAS calculator

Retention curves, split by where they came from.

One script tag and a billing connection. Free until $1K MRR.