Churn rate: the formula, the benchmarks, and the period everyone forgets
Churn rate is the percentage of customers or revenue lost over a defined period, calculated as losses during the period divided by the base at the start of it. The formula takes one line. The period attached to it decides whether the number means anything at all.
A 3% churn rate is either world-class or a business in serious trouble. Annual, it beats the median SaaS company. Monthly, it compounds to losing roughly 31% of your customers a year. Most articles quoting a churn benchmark never say which one they mean, and the two readings differ by a factor of ten.
Table of contents
- The churn rate formula
- Monthly and annual churn are not comparable
- What the benchmarks actually say
- Customer churn and revenue churn answer different questions
- A third of SaaS churn is failed payments
- Where the calculation breaks
- What churn rate cannot tell you
- Frequently asked questions
The churn rate formula
Two versions matter, and they are not interchangeable.
Customer churn rate counts logos:
Customer churn = customers lost during period / customers at start of period
Start January with 500 customers, lose 20, and customer churn for the month is 4%.
Revenue churn rate counts money:
Revenue churn = MRR lost during period / MRR at start of period
Start with $200,000 of MRR, lose $9,000 to cancellations and downgrades, and revenue churn is 4.5%.
Customers acquired during the period are excluded from both. This trips people up constantly. If you sign 40 new customers in January and lose 20, the churn denominator is still the 500 you started with, not 540 and not 520. Including new customers in the denominator flatters the rate by diluting it with accounts that never had time to leave, and the faster you grow the more it flatters.
That is also why a growing company can watch its churn rate look stable while retention quietly deteriorates. Growth hides churn in any calculation that lets new customers into the denominator.
Monthly and annual churn are not comparable
Churn compounds, so converting between periods is not multiplication or division. This is the single most common error in churn discussions and it changes conclusions by an order of magnitude.
A 3% monthly churn rate does not mean 36% annual churn. It means:
1 - (1 - 0.03)^12 = 30.6%
Just under 69.4% of your customers survive the year. At 5% monthly, 46% of them are gone in twelve months.
Running it the other way is more revealing. Recurly's July 2026 network data puts the median annual churn rate for SaaS at 3.22%. Expressed monthly, that is about 0.27%. So when an article tells you "3% to 5% monthly churn is normal for SaaS", it is describing a business churning ten to fifteen times faster than the median subscription company in a real payments network.
Both numbers get published. Both get quoted without a period attached. Before you compare your churn to any benchmark, find out which period the benchmark used, and if the source does not say, discard it.
What the benchmarks actually say
Recurly's churn benchmark research, drawn from its network of subscription businesses and updated with July 2026 data, reports median annual rates by industry:
| Industry | Total | Voluntary | Involuntary |
|---|---|---|---|
| SaaS | 3.22% | 2.16% | 1.06% |
| Business & professional services | 3.44% | 2.27% | 1.18% |
| Travel, hospitality & entertainment | 3.91% | 2.63% | 1.28% |
| Digital media & entertainment | 4.14% | 2.55% | 1.59% |
| Ecommerce | 4.25% | 2.87% | 1.38% |
| Education | 4.99% | 3.30% | 1.69% |
Across all industries the median sits at 3.60% total, split 2.34% voluntary and 1.25% involuntary. For software businesses specifically, Recurly puts the median overall annual churn at 3.04%, with top-quartile performers at 1.78% or below.
The more useful cut is by average revenue per customer, because it shows churn is not a single dial:
| ARPC band | Total | Voluntary | Involuntary |
|---|---|---|---|
| $10 to $25 | 4.29% | 2.99% | 1.30% |
| $25 to $50 | 3.84% | 2.73% | 1.11% |
| $50 to $100 | 3.15% | 2.41% | 0.74% |
| $100 to $250 | 2.87% | 2.40% | 0.46% |
| Over $250 | 3.07% | 2.90% | 0.18% |
Involuntary churn falls by more than seven times across that range while voluntary churn barely moves. Cheap subscriptions do not lose customers mainly because those customers are less satisfied. They lose them because low-value cards fail more often and are worth less effort to recover.
This is the same segmentation effect that makes a single net revenue retention benchmark misleading, and the reason a blended rate should always be broken out by cohort before you act on it. ChartMogul's SaaS Retention Report finds the pattern from the opposite direction: retention rises sharply with average revenue per account. Price point drives the number more than execution does, so comparing your churn to a median computed across every price band tells you almost nothing about your own performance.
Customer churn and revenue churn answer different questions
Track both, because the gap between them identifies who is leaving.
When revenue churn runs higher than customer churn, the accounts you are losing are larger than average. Five percent of customers walking out with 9% of revenue is an enterprise retention problem, and it is the more dangerous direction because large accounts are expensive to replace and usually took a long sales cycle to win.
When revenue churn runs lower than customer churn, you are losing small accounts. Eight percent of logos taking 3% of revenue is often a self-serve tier behaving exactly as self-serve tiers do. It still costs support load and it still damages word of mouth, but it is not a threat to the revenue line.
When the two track closely, churn is spread evenly across the base, which usually points at something structural: onboarding, a missing feature, a competitor.
Revenue churn also has a subtlety customer churn does not. Downgrades count. A customer dropping from $500 to $200 contributes $300 of revenue churn while remaining a customer, so revenue churn can rise in a month where nobody cancelled at all. Tracking gross revenue churn separately from contraction keeps that distinction visible.
A third of SaaS churn is failed payments
Involuntary churn is the portion of churn caused by a payment failing rather than a customer deciding to leave. Recurly defines it precisely: the subscriber loses access because a payment fails, and it reflects nothing about their intent to stay.
Run the numbers from the table above. SaaS median annual churn is 3.22%, of which 1.06% is involuntary. That is 33% of all SaaS churn caused by payment failures rather than dissatisfaction. Across all industries the share is 35%.
Most churn advice ignores this entirely and treats churn as a product and customer success problem. A third of it is a billing problem, and it is the third most likely to be recoverable, because those customers wanted to keep paying. Expired cards, insufficient funds, issuer fraud rules and failed 3D Secure challenges all produce identical outcomes in your metrics and completely different outcomes in reality.
The practical response is dunning: retry schedules, card account updater services, pre-expiry notifications and a grace period before access is revoked. Stripe's revenue recovery documentation covers the mechanics. The measurement response matters just as much. If your churn number does not separate voluntary from involuntary, you cannot tell whether last month's spike was a product failure or a payment processor changing its retry behaviour, and those call for opposite responses.
Where the calculation breaks
Choice of denominator. Start-of-period is standard. Some teams use the average of start and end, which produces a lower number and is not comparable to anyone else's. Neither is wrong, but mixing them across months makes your own trend meaningless.
Annual contracts make churn lumpy. A customer on an annual contract can only churn on their renewal date, which is one reason annual recurring revenue moves in steps rather than smoothly. If most of your contracts were signed in Q1, most of your churn lands in Q1, and monthly churn rates for the rest of the year look artificially healthy. Cohort-based renewal rates describe this far better than a monthly average does.
Mid-period signups. A customer who signs up on the 20th and cancels on the 28th never appears in a start-of-period denominator, so a wave of instant cancellations can be invisible in the churn rate while being extremely visible in your support queue.
Small samples. At 50 customers, one cancellation is 2% churn. The month-to- month variation is noise, not signal, and reading it as a trend produces confident conclusions about nothing.
Reactivations. A customer who cancels in March and returns in May can be counted as churn plus new, or netted out. Both conventions exist and they produce different churn rates from identical events.
Trials and freemium. A trial that ends without converting is not churn unless you counted the trial as a customer. Counting trials as customers makes churn rates look catastrophic and makes them incomparable to anyone reporting on paying customers only.
What churn rate cannot tell you
Churn rate is a lagging aggregate, and it arrives after the decision that caused it was already made.
It cannot tell you when the cause happened. A customer who cancels in August usually decided in June, often after a specific event: a failed integration, a support ticket that went unanswered, a competitor's launch, a price change. The churn number timestamps the cancellation, not the cause.
It cannot tell you which customers are next. The accounts that will churn in November are, today, still paying and showing contraction, declining usage or a support pattern that a monthly percentage cannot express.
It cannot tell you why. A churn spike from a payment provider migration and a churn spike from a botched release look identical in the aggregate.
The gap between "churn went up 0.4 points" and "these eleven accounts left because the SSO integration broke on 14 July" is the gap between a metric and a decision.
Frequently asked questions
What is a good churn rate for SaaS?
For the median SaaS business in Recurly's July 2026 network data, annual churn sits at 3.22%, and top-quartile software businesses run at 1.78% or below. Judge yourself against your own price band rather than the overall median, since involuntary churn alone varies by more than seven times between sub-$25 and $250-plus subscriptions.
How do I convert monthly churn to annual churn?
Use 1 - (1 - monthly rate)^12. Churn compounds, so multiplying by twelve
overstates it. A 3% monthly rate is 30.6% annual, not 36%.
What is the difference between churn rate and retention rate?
They are complements of each other for the same base and period: retention = 1 - churn. Where they diverge is expansion, which retention metrics like net revenue retention include and churn rates do not. NRR can exceed 100% while revenue churn is positive.
Should I use customer churn or revenue churn?
Both. Customer churn tells you how many relationships you are losing, revenue churn tells you what they were worth, and the gap between them tells you whether your churn problem lives in the enterprise segment or the self-serve tier.
Is involuntary churn really that significant?
Yes. It accounts for roughly a third of total churn in SaaS and about 35% across all subscription industries in Recurly's data. It is also the most recoverable category, because those customers did not choose to leave.
Ready to find out why churn moved, not just that it did?
Churn rate tells you the percentage. It does not tell you which accounts left, what happened before they did, or whether the cause was a product change, a pricing change or a payment processor quietly tightening its retry rules.
GainSignal separates voluntary from involuntary churn, connects each churn event to the accounts and the changes that preceded it, and surfaces contraction while those customers are still paying you. The finding is not "churn is up". It is "these accounts left, this is what changed first, this is what it cost".