Key takeaways
- State the period and the denominator. A rate without both is not comparable.
- Logo churn counts customers; revenue churn counts dollars lost with them.
- Net revenue retention adds expansion and subtracts contraction and churn.
- Compound, do not multiply. 5% monthly churn is 46% a year, not 60%.
- Cohorts reveal improvement that a blended rate hides.
What churn counts
Churn measures loss, and the first decision is what is being lost. Logo churn counts customers who cancelled. Revenue churn counts the recurring revenue they took with them. Account churn counts contracts, which can differ from both when one company has several accounts or one account has several products.
The second decision is the period. Monthly churn is standard for self-serve and SMB subscriptions; annual churn is standard for enterprise contracts. A monthly number and an annual number are not comparable, and a blended number that mixes contract lengths is hard to interpret.
The formulas
Churn rate = customers lost ÷ customers at the start of the period
Retention rate = 1 − churn rate | Customers lost = starting customers − ending customers + new customers
Annual churn from monthly = 1 − (1 − monthly churn)^12
Monthly churn from annual = 1 − (1 − annual churn)^(1/12)
Compounding is the part people skip. Five percent monthly churn is not 60% a year; it is 1 − 0.95^12 = 46.0%, because the base shrinks each month. Going the other way, 30% annual churn is 1 − 0.70^(1/12) = 2.93% a month, not 2.5%.
Worked examples and annualising
| Starting customers | Churned | Monthly churn | Retention | Annualised churn |
|---|---|---|---|---|
| 1,000 | 50 | 5.0% | 95.0% | 46.0% |
| 500 | 25 | 5.0% | 95.0% | 46.0% |
| 2,000 | 120 | 6.0% | 94.0% | 52.4% |
The third row shows why small differences matter: 6% monthly churn leaves 47.6% of the base after a year. Over three years, 0.94^36 = 10.8% survives. A single point of monthly churn is worth tens of percent of the customer base over a few years.
Annualising should be labelled as a projection. It assumes the rate holds for twelve months, which is rarely exactly true, and it is not the same as measuring churn over a full year.
Logo churn versus revenue churn
Suppose 8 of 120 customers leave in a month and they were paying $240 of the $6,600 opening MRR. Logo churn is 8 ÷ 120 = 6.7%. Revenue churn is 240 ÷ 6,600 = 3.6%. The business lost more customers than revenue, which usually means small accounts left.
The reverse pattern is more dangerous. Losing 2 customers who pay $2,000 of a $20,000 base is 2% logo churn and 10% revenue churn. A dashboard that reports only logo churn would describe that month as calm. Report both, and report the average revenue per lost customer next to them.
Gross versus net revenue retention
Gross revenue churn ignores expansion. Net revenue retention puts it back:
NRR = (opening MRR + expansion − contraction − churned MRR) ÷ opening MRR
With $75,000 of opening MRR, $3,000 of expansion, $2,000 of contraction, and $4,000 of churn, NRR = (75,000 + 3,000 − 2,000 − 4,000) ÷ 75,000 = 72,000 ÷ 75,000 = 96%. Gross revenue churn is 5.3%, net revenue churn is 4.0%, and the difference is expansion.
Above 100% the existing base grows without a new customer. Below 100% new sales are partly replacing lost revenue before they add anything. NRR is not a substitute for logo churn: a 105% NRR driven by price increases can coexist with heavy customer loss, and the customers who remain may be the ones with the least choice.
Cohorts and the denominator
The denominator can be the starting count, the ending count, or the average. Starting is the most common and the easiest to defend. Average is (start + end) ÷ 2, which produces a slightly higher rate when the base is shrinking: 100 churned from 1,000 at the start and 900 at the end is 10.0% of the start and 10.5% of the 950 average.
Pick one, disclose it, and keep it. A blended rate across all customers also hides cohort behaviour. A cohort view follows customers who started in the same month; if the January cohort retains better than the June cohort, the product or onboarding improved even when the blended number is flat.
Common mistakes
- Multiplying monthly churn by twelve instead of compounding it.
- Mixing logo churn and revenue churn in the same comparison.
- Counting downgrades as churn rather than contraction.
- Leaving involuntary churn, such as failed payments, out of the loss.
- Changing the denominator between periods and calling it a trend.
- Reporting a churn percentage without the customer count behind it.
Practical caveats
Small numbers are noisy. One cancellation among 40 customers is 2.5% churn and also a single event. Track the count and the percentage together, and use rolling three-month averages before acting on a movement.
Seasonality distorts monthly comparisons. A business that loses customers after the holidays will show a spike every January. Compare the same month year over year, or use a trailing twelve-month view for planning.
Distinguish voluntary from involuntary churn. Voluntary churn is a product or price decision by the customer; involuntary churn is a failed card or an expired trial. The fixes are different, and combining them hides whether the problem is value or billing operations. Reactivations should be reported separately from new customers so the base is not double counted.
Reporting churn with growth
Churn is half of a growth equation. Net customer growth = new customers − churned customers, divided by the starting base. Starting with 1,000, adding 80, and losing 50 gives 3% net growth. In dollar terms, a quick ratio of (new + expansion) ÷ (churned + contraction) = (4,500 + 1,500) ÷ (2,000 + 1,000) = 2.0 says the base is being refilled faster than it drains.
Report the four lines together: new, expansion, contraction, churn. The total explains the change in MRR and in customer count, and it prevents a flat month from being described as stable when it was actually a month of heavy replacement.
Frequently asked questions
Is churn calculated monthly or annually?
Either, as long as it is stated and applied consistently. Annualise by compounding, not by multiplying the monthly rate by twelve.
What is a good churn rate?
There is no universal figure. It depends on the segment, contract length, price point, and stage of the business. Compare against your own cohorts and close peers.
Does a downgrade count as churn?
No. A downgrade is contraction: the customer stays but pays less. Churn is a full cancellation, which is why revenue churn and contraction are reported separately.
How do I calculate net revenue retention?
Add expansion to starting recurring revenue, subtract contraction and churn, then divide by starting recurring revenue. A result above 100% means the existing base grew.
Should churn use starting or average customers?
Starting is simpler and more common. Average reduces distortion when the base moves sharply within the period. Pick one and disclose it.
What is involuntary churn?
Loss caused by failed payments, expired cards, or billing errors rather than a customer decision. It is often recoverable and should be tracked separately from voluntary cancellations.
Further reading