SaaS Churn Rate: Which Average Is Yours

What a SaaS churn rate means

Churn rate is the share of a subscription base lost in a period. The formula is not the hard part and never has been. The hard part is that the same word covers two different counts, and both of them are published as though there were only one.

Customer churn counts people. Divide the customers lost in a period by the customers you had at the start.

Both formulas are easy to find, the overview Google returns included, and neither answers the question underneath them. Which of the two you are looking at is the question it leaves out.

Revenue churn counts money. Divide the recurring revenue lost by the recurring revenue you started with, so a downgrade registers and a cheap cancellation registers less.

A company with heavy small-account turnover and a stable enterprise base can report a frightening customer churn and a comfortable revenue churn in the same quarter. Neither number is wrong.

Every benchmark on this page counts departures and not money, and even inside that the unit shifts.

Recurly counts subscribers, and one account can hold several subscriptions. Check what a source counted before you hold your own number against it.

Why the published averages disagree

The published averages for SaaS churn run from 3.22 percent to 14 percent. That spread is among figures using the same annual window, counting departures and not revenue. That looks like a window problem and it is not one.

One guide carries its range from two other publishers, HubSpot and Stripe, and neither arrives with a sample. Recurly, the one row running on its own data, does not publish a sample either.

Split it by who is buying

Vena publishes the answer in a table that its own headline does not use. Once the customers are separated, the disagreement stops being a disagreement.

Vena Solutions, read 7 September 2026: the same metric, split by who is buying

1. 1Enterprise SaaS loses 10 percent of customers a year or less. Usage-based and freemium products lose 50 percent or more.
2. 2That is a factor of five inside one category, which is wider than the gap between any two publishers quoting an average for it.
3. 3The last two rows are monthly and the ones above them are monthly too, and the voluntary figure is larger than the overall B2B figure it should sit inside. Something in that table is inconsistent, so those rows are not used together here.

So a store of averages is not much use. Say your rate is 20 percent a year. That is twice the enterprise ceiling and below the small-business floor of 30 percent, so your number falls between two published bands and belongs to neither.

Separate the cancellations from the declines

The total splits two ways, and almost nobody splits it: customers who decided to leave, and customers whose payment simply failed. and the second group is larger than most teams assume.

On those figures roughly a third of the loss is a billing event. An expired card, a declined transaction, a bank that blocked a recurring charge.

A billing dunning queue, the screen involuntary churn arrives onWhere the half of churn nobody chose becomes a cancellation

The decline reason column is the only place the two kinds of loss are told apart. A row that runs out of retries becomes a cancellation and joins the churn number beside the customers who decided to go. Nothing on this screen is a product problem.

That third has nothing to do with your product and it does not respond to anything you would do about a product. It responds to card updater services, retry schedules and dunning emails.

A forum thread carries the same observation in its title, from an operator: most of their churn was failed payments. That is one person and not evidence, and it points at the same place the network data does.

What churn rate gets confused with

Three neighbouring numbers get used as though they were this one. Each answers a question churn cannot answer, and mixing them is how a healthy business reads as a failing one or the reverse.

Monthly and annual rates are the fourth confusion. A 1 percent monthly rate is not 12 percent a year, because each month is measured against a base the previous month already reduced.

State the window beside the number. Two figures with no window attached cannot be compared at all, and the ranges above show how far apart they can sit.

Where the term goes next

Churn is an input, not a conclusion. Two decisions sit downstream of it, and both change when the rate moves, which is why the number is worth measuring properly instead of benchmarking loosely.

The first is what a customer is worth. Lifetime value takes your churn rate as its denominator, so an error here scales into every acquisition decision you make from it, and the LTV to CAC ratio inherits it whole.

The second is where the money goes. A business losing half its customers a year is buying the same customers repeatedly, and no acquisition channel outruns that. Products in the freemium band above face this arithmetic first, which is one of the things whether product-led growth fits your price has to answer for.

Measure it monthly, split it into voluntary and involuntary, and record which of the two counts you used. Then compare yourself against your own segment row and not against a category average.

Suppose the acquisition side is the one you want moved instead: that is our SaaS SEO service.