Churn

What is Churn?

Churn is the rate at which customers, or the revenue they represent, stop paying over a given period.

Two camps describe it slightly differently, and the difference matters. David Skok's widely circulated "SaaS Metrics 2.0" on For Entrepreneurs splits it into customer churn, the percentage of accounts lost, and MRR churn, the percentage of recurring revenue lost, noting you can lose a lot of small customers or a few of your biggest ones and get a very different picture from each measure. a16z's "16 Startup Metrics" (2015) frames the same split at the dollar level: gross churn is MRR lost in a month divided by MRR at the start of the month, and net churn subtracts upsell revenue from that loss before dividing. a16z is blunt about why the distinction exists: gross churn estimates the actual loss to the business, while net churn understates it because it blends upsells with the churn itself.

Produck uses churn the way both sources converge on it. Customer churn tells you how many accounts you're losing. Revenue churn (gross, not net) tells you how much money that's actually costing you, and it's the one to trust when you want an honest read on whether the business is leaking.

Why it matters for product-market fit

Churn is a lagging indicator. By the time it shows up in your dashboard, the account already decided to leave, weeks or months before the invoice stopped. That's why Produck treats churn as an output of the loop, not an input you manage directly.

In the Listen stage, Produck pulls in the feedback that predicts churn before it happens: a feature request that goes unanswered, or a support thread that never resolves. In Diagnose, that feedback gets tied to the account and the retention cohort it belongs to, so you can see which unresolved issues correlate with the accounts most likely to leave. In Decide, you're weighing a fix against everything else on the roadmap, and a rising churn signal is often the argument that moves it up. In Ship, the loop closes: you ship the fix, then watch whether churn in that cohort actually drops, which is the only real proof the fix worked.

This is also where churn connects back to activation. Accounts that never activate well are far more likely to churn early, so a churn problem often traces back to a broken first-run experience rather than anything about the mature product. And churn should never be read in isolation from your North Star Metric, since a team can hold churn flat while still losing ground on the metric that actually reflects value delivered. Produck's argument for why teams get this wrong in the first place is laid out in the PMF stack is broken: most tools measure activity, not whether the product is actually working for the people using it, and churn is the clearest place that gap shows up.

When it works, and when it doesn't

It works when

• You segment it by cohort and plan, so a churn spike in one pricing tier doesn't get buried in a healthy blended average
• You track gross churn alongside net churn, so expansion revenue from your best accounts doesn't mask losses elsewhere

It falls short when

• You only look at the trailing number, since by definition it can't tell you what to fix, only that something needs fixing
• You compare churn across companies without matching contract length and segment, since a monthly self-serve product and an annual enterprise contract produce very different baseline rates
• You treat a falling churn number as proof of product-market fit on its own, when it can just as easily mean you've stopped selling to the accounts likely to churn
• You let net churn substitute for gross churn in an executive update, since a healthy-looking net number can hide a real gross-churn problem underneath

How to apply it

1. Pull last quarter's churned accounts and split them into customer churn (count) and revenue churn (dollars), gross not net.
2. For each churned account, find the last piece of feedback they gave you, whether that's a support ticket or silence after a specific unresolved issue.
3. Group those accounts by the workflow or feature area where the friction lived, and check that group against your activation data from onboarding.
4. Rank the patterns by how much revenue they represent, not how many tickets they generated.
5. Pick the top pattern and turn it into a shipped fix, tracked against the cohort it came from.
6. Produck recommends re-measuring that cohort's churn 60 to 90 days after the fix ships as a practical starting window, and only then calling it resolved.

churn vs retention

Churn and retention measure the same movement from opposite ends. Churn counts what you lost. Retention counts what you kept, and a mature team should track both rather than inferring one from the other, since a retention number can look flat while gross churn quietly rises and gets offset by new signups. Read the full breakdown on the retention page.