Customer churn is the rate at which customers stop paying you over a given period. It is the inverse of retention: if you keep 90% of last year’s customers, you churned 10%. The number drives every Customer Success org because compounding churn caps growth no matter how fast you acquire — at 3% monthly churn, half your customer base is gone in under two years.
What churn is not
Churn is not the same as a single canceled contract, and it is not one number. The word collapses two distinct measurements that move independently:
- Logo churn counts accounts lost. If 100 customers start the period and 8 leave, logo churn is 8%.
- Revenue churn counts dollars lost. If those same 8 accounts were your smallest, revenue churn might be 2%; if one was a whale, it might be 15%.
A team can have low logo churn and catastrophic revenue churn, or vice versa. Report both. Reporting only logo churn is how SMB-heavy books hide a concentration problem; reporting only revenue churn is how a few big renewals mask a leaky long tail.
The formulas
Logo churn = Accounts lost in period / Accounts at start of period
Gross $ churn = (MRR lost to cancels + downgrades) / MRR at start of period
Net $ churn = (MRR lost − expansion MRR) / MRR at start of periodNet revenue churn can be negative when expansion outruns losses — that is the same fact GRR/NRR express, just sign-flipped (NRR = 1 − net revenue churn). See NRR vs GRR for the retention framing.
Voluntary vs involuntary
The other split is why the dollar left, and it changes who owns the fix:
- Voluntary churn — the customer chose to leave. Lost a champion, didn’t see value, found a cheaper tool, got acquired. This is a CS and product problem. The fix is value delivery, adoption, and relationship depth.
- Involuntary churn — the customer didn’t choose to leave; a payment failed. Expired card, insufficient funds, a hard decline the dunning sequence never recovered. This is a billing-ops problem, and it is the cheapest churn to fix. For self-serve B2B SaaS, involuntary churn commonly runs 20-40% of total churn (Recurly’s published dunning benchmarks); recovering even half of it with smart retries and card-updater services moves the headline number more than any CS play.
If you are not separating these two, you are aiming retention budget at the wrong problem. A QBR does not fix an expired Visa.
Why it compounds
Churn is multiplicative, not additive. Retention of R per period over n periods leaves you with R^n of the cohort:
| Monthly gross churn | Customers left after 12 months | After 24 months |
|---|---|---|
| 1% | 89% | 79% |
| 2% | 78% | 62% |
| 3% | 69% | 48% |
| 5% | 54% | 29% |
The gap between 1% and 3% monthly looks small in any single month and is the difference between a durable business and a treadmill over two years. This is why investors price NRR so heavily: low churn lets every dollar of new ARR stack instead of plugging a leak.
Benchmarks
Rough B2B SaaS norms (KeyBanc and SaaS Capital annual surveys):
- Annual gross revenue churn: under 10% is healthy for SMB-focused SaaS; under 5% is good for mid-market; under 3% is enterprise-grade. Median private B2B SaaS lands around 12-14% gross logo churn annually.
- Net revenue retention (the expansion-inclusive mirror): 100-110% is solid, 120%+ puts you in the top decile. Below 100% means churn is outrunning expansion.
Segment matters more than the headline. SMB books churn 2-4× faster than enterprise books on logo, so a blended company number hides which motion is bleeding.
Common pitfalls
- Measuring churn without cohorting. A blended rate averages a leaky new-customer cohort against a sticky old one. Cohort by acquisition quarter and segment, or you cannot tell whether onboarding or aging is the problem.
- Counting involuntary churn as voluntary. This sends CSMs to “save” accounts that only needed a card update — wasted motion, and the real failed-payment rate stays invisible. Tag the reason at cancellation.
- Annualizing monthly churn by ×12. 3% monthly is not 36% annual; it is 1 − 0.97^12 ≈ 31%. Multiplicative decay, not linear. Get this wrong on a board slide and the model is off by a tenth of the base.
- Ignoring down-sell. A customer who drops from 50 seats to 5 didn’t churn as a logo but vaporized most of their revenue. Gross revenue churn catches this; logo churn does not.
Related
- NRR vs GRR — the retention-side mirror of churn
- Gainsight — health scoring and churn-risk workflows
- ChurnZero — renewal and churn-signal automation
- Vitally — usage-based churn-risk scoring