Churn explained simply

What churn actually measures

Churn is the share of customers — or revenue — that you lose over a fixed window. If 1,000 paid subscriptions are active on the first of the month and 50 cancel before the last, your customer churn for that month is 5%. The mirror metric is retention, and the two always sum to 100% over the same denominator. That sounds trivial until a product manager at Notion or a founder pitching Sequoia tries to compare their number to a competitor and realizes nobody is using the same definition.

The reason churn matters more than almost any other top-of-funnel metric is that it compounds. A consumer subscription product running at 8% monthly churn loses roughly 64% of any cohort within twelve months, no matter how good the acquisition engine is. You can pour money into Meta ads, hire a growth team, ship a referral loop — and still be filling a bucket with a hole in the bottom. This is why investors ask about churn before they ask about CAC: it caps every other growth lever you have.

The other thing to internalize early: churn is a result, not a cause. A spike next month was almost always created two or three months ago, by a price change, a botched onboarding update, or a competitor launch. By the time the dashboard turns red, the user who cancelled today decided to leave six weeks ago. Treat churn as a lagging indicator and pair it with leading signals like login frequency, feature adoption, and support-ticket volume.

The four flavors of churn

People say "churn" and mean four different things. Mixing them up is the single biggest source of misleading board slides. The two axes that matter are who decides (voluntary vs involuntary) and what you count (customers vs dollars, gross vs net of expansion).

Voluntary churn is the obvious one: a customer clicks "cancel subscription" in settings. The fix lives in product and pricing — better onboarding, better packaging, save offers at the cancel step. Involuntary churn is the silent killer: an expired credit card, a failed 3DS step, a Stripe webhook that quietly retried three times and gave up. On consumer subscription books, involuntary churn routinely accounts for 20-40% of total cancellations, and most of it is recoverable with a dunning sequence that costs almost nothing to run.

Gross revenue retention (GRR) caps at 100% — it can never exceed it, because it only counts losses. Net revenue retention (NRR) can exceed 100% when expansion (upgrades, seat growth, usage overages) is bigger than churn plus contraction. A SaaS company reporting NRR 120% with GRR 85% is telling you two things at once: their existing customers are growing fast, but their underlying logo retention is mediocre and someday the expansion engine will hit a ceiling.

Benchmarks by segment

There is no universal "good" churn rate. A 5% monthly figure that would get a B2B SaaS CEO fired is healthy for a consumer subscription app, and would be a triumph in mobile gaming. The benchmark table below is the rough range investors and operators use as a sanity check — not gospel, but a starting point for the question "are we in the right neighborhood?"

A few patterns worth pinning down. Enterprise contracts are sticky because switching costs are real — migration, retraining, procurement. SMB customers churn fast because they go out of business; the only fix is to win bigger logos. Consumer subscriptions look horrifying monthly but Netflix's 3-4% monthly is considered best-in-class. If your B2C product clears 5% with no expansion, you are doing fine — chasing 1% eats margin you don't have. Fintech is the middle case: neobanks like Chime or Revolut measure "active account" churn since the product is mostly free, and the number that actually drives revenue is share-of-wallet.

Why churn is the fastest lever on LTV

For a flat-ARPU subscription product, the canonical formula is:

LTV = ARPU / monthly_churn_rate

Plug in real numbers. At $50 ARPU and 5% monthly churn, LTV is $1,000. Cut churn to 3% and LTV jumps to $1,667 — a 67% increase from a two-point reduction. Cut it to 2% and LTV doubles to $2,500. No acquisition channel, no pricing experiment, no upsell motion produces leverage like this. It is the closest thing to compounding interest in growth work.

The reason churn beats acquisition as a lever is asymmetry of effort. Doubling new customers means doubling CAC spend or unlocking a new channel — both expensive and slow. Cutting churn by two points often means fixing a single onboarding step, plugging involuntary payment failures with a smart retry sequence, or sending a win-back email to dormant users. One engineer-week of dunning work can return more LTV per dollar than a quarter of paid-acquisition optimization.

The corollary: if your product is pre-product-market-fit, no amount of churn work will save you. Save-offers on a leaky product just delay the inevitable and pollute your cohort math. Fix the leak in the first 30 days first — that's an activation problem, not a churn problem. See our activation framework for product managers for how to separate the two.

The churn curve and what it tells you

Plot survival by tenure and almost every product shows the same shape: high churn in the first weeks, then a long flattening tail. The first 30 days do most of the damage. After month three or four, the surviving cohort is usually 5-10x stickier than the day-one cohort.

Churn rate
  |
  | *
  |  *
  |   *
  |    *
  |     **
  |       ***
  |          *****______
  +-----------------------> tenure (months)
        0    3    6    12

This shape changes how you spend effort. Early churn (months 0-2) is almost always an onboarding, activation, or expectation-setting problem — the user signed up, didn't reach the aha moment, and left. Late churn (months 6+) is a value-delivery problem — they got what they came for and have nothing left to do, or a cheaper alternative appeared. The fixes look completely different: onboarding redesigns and lifecycle emails for the first, depth-of-product roadmap and account-management motion for the second.

Common pitfalls

When a team reports churn for the first time, the most common error is conflating monthly and annual rates by simple multiplication. Monthly churn of 5% does not equal annual churn of 60% — the correct formula is 1 − (1 − 0.05)^12 ≈ 46%, because each month's churn applies to the shrunken base from the previous month. Boards that don't catch this end up benchmarking your monthly number against a competitor's annual number and concluding you're three times worse than you are.

A second trap is measuring churn only on paying users in a freemium product. If your free tier is the top of your funnel, free-user churn is the metric that actually predicts paid revenue 90 days out. Limiting the definition to paid subscribers hides the leak where it actually starts. The cleaner approach is to track engagement-based churn (no login in 30 days) for the free base and revenue churn separately for the paid base, then connect them through your activation funnel.

The third pitfall is ignoring cohort heterogeneity. New cohorts almost always churn faster than older ones — partly because the old ones are survivor-biased, partly because acquisition channels shift over time. If you report a single blended churn number, a deteriorating new-cohort curve can be masked for months by your sticky long-term base. The fix is to report churn by acquisition month, not just headline rate, and to flag any cohort that looks materially worse than the one before it.

The fourth, and most expensive, is treating involuntary churn as voluntary. When a user's card expires and the renewal fails, every product analytics tool will record that as a cancellation. Without a dunning system that retries and notifies the user, you lose customers who genuinely wanted to keep paying. Card-update prompts, smart retry logic, and account-update services from Visa and Mastercard typically recover 30-50% of involuntary churn — that's pure margin with no acquisition cost.

Related reading

• How to calculate churn in SQL
• How to calculate involuntary churn in SQL
• How to calculate churn risk score in SQL
• How to calculate gross revenue retention in SQL
• CAC explained simply
• Activation framework for product managers

If you want to drill metric-design questions like this every day, NAILDD is launching with hundreds of PM and analytics interview problems built around exactly these tradeoffs.

FAQ

Should I report monthly or annual churn?

Use the cadence that matches your contract length. SaaS products with monthly billing report monthly churn; enterprise products with annual contracts report annual churn (and break it down by renewal cohort). The fastest way to confuse a board is to report monthly churn for your SMB segment and annual for enterprise on the same slide without labelling — the SMB number will look ten times worse than it really is. Pick one cadence per segment and stick with it.

Customer churn or revenue churn — which matters more?

For investor conversations and unit-economic models, revenue churn is the number that drives valuation, because it accounts for the fact that losing one enterprise logo at $200k ARR is not the same as losing one SMB at $200. For product and CS team-level OKRs, customer churn is often more actionable, because it tracks the count of human relationships you need to repair. Most mature SaaS companies report both, and the gap between them tells you whether your high-value or low-value tier is bleeding faster.

What does it mean when net revenue retention is over 100%?

NRR above 100% means existing customers are paying you more over time than the ones you lose take away. The expansion can come from seat growth (a customer team grew from 10 to 25 users), usage growth (they processed more transactions), or upsell to higher tiers. Best-in-class B2B SaaS companies like Snowflake and Databricks have historically run NRR in the 130-170% range, which is what makes their growth so capital-efficient — they don't need to acquire as aggressively because the installed base is its own growth engine.

How is hard churn different from soft churn?

Hard churn is the cancellation event: payment stopped, subscription deleted, account closed. Soft churn is the behavioral leading indicator: login frequency dropped, the user stopped opening lifecycle emails, support tickets fell to zero. Soft churn typically precedes hard churn by 30-90 days and gives you a window to intervene before the customer is gone. Most CS teams build their health-score models on soft-churn signals because by the time hard churn fires, win-back is much harder than retention.

When is churn reduction the wrong priority?

Before product-market fit, focus on activation and value delivery, not churn. A leaky bucket needs to hold water at all before you patch the holes. If fewer than 30-40% of new signups make it through activation in the first 30 days, your problem is upstream of churn, and save-offer or dunning work will produce numbers that don't compound. Once activation is healthy, churn becomes the highest-leverage lever you have — in roughly the order: involuntary (dunning), onboarding (first-30-day fixes), expansion (upsell motion), then save-offers and win-backs.

Can churn ever be too low?

Surprisingly, yes. A consumer product with implausibly low churn often has a measurement bug — inactive users counted as active because the "active" definition is too loose. A B2B product with very low logo churn but flat revenue might be over-discounting to keep accounts on the books, which damages NRR even though the churn line looks great. If your churn rate is significantly better than every benchmark in your segment, audit the definition before you celebrate.