Is My Churn Normal? Leaky Bucket vs Dead Product

Churn is “normal” when your retention curve flattens. After the early drop-off, a stable group of users keeps coming back month after month. That flat tail is a leaky bucket, and a leak is fixable. Churn is fatal when the curve never flattens and slides toward zero.

Plot the cohort retention curve

Plot the percentage of a signup cohort still active at day 1, 7, 14, 30, 60, 90.

A leaky bucket curve falls steeply in the first week or two, then the slope levels into a near-horizontal tail. The users in that tail are your real product-market fit. You churn the rest, but you have a durable base to grow on.

A dead-product curve drops and keeps dropping. There is no stable group, because nobody forms a habit.

The retention number at day 90 matters far less than whether the curve flattened before it got there. A consumer app that flattens at 4% can be a healthy business. A B2B tool sliding through 30% with no plateau in sight is in trouble.

Why the plateau is the whole game

A retention plateau is the point where your curve stops falling and goes roughly flat. The survivors in it keep coming back indefinitely.

Growth is a bucket: you pour new users in the top, and churn drains them out the bottom. If the curve plateaus, the drain has a bottom. Pour in faster than the leak, and the base compounds.

If the curve never plateaus, every user you acquire eventually drains out. You can pour harder by spending more on acquisition. The moment you stop pouring, the level drops to zero.

The benchmark tool’s growth model works the same way. New cohorts retain at your month-1 rate. A stable base of long-tenured survivors churns at a much lower “mature” rate, about 10 times lower than fresh cohorts.

When your curve has a plateau, that stable base is real and it compounds. When month-1 retention is low and nothing graduates into the base, the model decays the base at up to ~10% a month until it hits zero.

Leaky bucket vs dead product: the side-by-side

Teams see scary churn, assume a dead product, and either give up or paper over it with more ad spend. Most of the time it is a leaky bucket. A leaky bucket is good news, because leaks have locations.

What “normal” churn looks like

Once you have confirmed a plateau, the numbers tell you whether your leak is normal or oversized. Business model, motion (freemium vs enterprise), and segment move “good” around a lot. Compare B2B to B2B and consumer to consumer.

B2B SaaS cohort retention (share of all signups still active)

Directional context, not targets: compare like-for-like (B2B to B2B, same motion and segment). B2B figures are cohort retention (Amplitude for Day 90; Day 1 and Day 7 are estimates). Pendo’s returning-user rates (50 to 70% Day 1, 40 to 60% Day 7, 25 to 35% at 90 days) measure something else.

Sources: Amplitude B2B Technology Product Benchmarks (Day 90, median to 90th percentile). Day 1, 7 and 30 are estimates kept consistent with it.

Consumer app retention (from first signup/use)

Directional context, not targets: compare like-for-like (consumer to consumer, same category).

Sources: Adjust, AppsFlyer, Amplitude, Mixpanel, UXCam.

A consumer curve flattening at 1 to 4% is a leaky bucket with a wide mouth. The model is expected to lose most users and make the math work on the survivors.

For the early-week drop, see why retention drops after day 1. For the long-tail plateau and what day-90 tells you, the day-90 retention guide goes deeper. The full ranges live in the retention benchmarks article.

How to tell which one you have: a 3-step diagnosis

1. Plot the cohort curve, not the blended number. A single “monthly churn rate” blends fresh users (churning fast) with your tenured base (churning slow) into one misleading average. Split by signup cohort and watch each cohort age.

2. Look for the flattening point. If you can draw a flat line through the tail, even a low one, you have a bucket. If the tail still slopes down all the way out, you have a drain. Real plateaus sometimes only appear at day 60 to 90, so give it time.

3. Check whether the plateau is moving. Compare your newest cohorts to your oldest. If the plateau is rising cohort over cohort, your fixes are working. If it is flat, you are stable but not improving. If it is falling, a previously-fine product may be dying.

How to plug a leaky bucket

If you have confirmed a plateau, the leak has a location. The work is finding which cohort or moment loses people.

• Fix the early cliff first. Most of the lost volume is in the day-1-to-day-14 fall. If users churn before reaching the core value, that is an onboarding or activation problem, and the cheapest leak to plug. (More on the first-week cliff in why retention drops after day 1.)
• Segment the curve. Often one segment is a leaky bucket and another is a dead product. Plot curves per acquisition channel, plan, or use case. Stop spending on the segments that never plateau, and double down on the ones that do.
• Raise the plateau, don’t just widen the mouth. Pouring in more users hides a leak for a while but does not seal it. Move your month-1 retention from, say, 30% to 35%, and every future cohort graduates a bigger survivor pool into the stable base.
• Confirm it is a leak. If no segment ever flattens after a proper cohort analysis, stop plugging. The fix there is the product or the audience, which is a product-market fit question.

Teams skip the cohort curve and cannot tell a leak from a drain. Then they spend months “improving retention” with email cadences and notifications on a product whose curve has no floor. Do the boring chart first.

See where your curve lands

Plotting cohort curves by hand is tedious. The free benchmark tool at benchmark.scilla.studio charts your retention against B2B and consumer ranges and runs the cohort growth model described above. Slide your month-1 retention and K-factor to see whether your base compounds or decays. In six minutes you know which bucket you hold.