Churn rate is the share of your customers, or your revenue, that you lost over a set period. That definition is easy. The hard part is producing a number that means the same thing in April as it did in January, so that when someone in a meeting says "churn is up," you are actually arguing about the business and not about two different spreadsheets. This guide walks through the formula, a worked example, the three variants people get wrong, a quick way to sanity-check your result, and the benchmark ranges that tell you whether your number is fine or a fire.
Before you begin: lock three decisions
You cannot calculate churn until you have settled three things. Skip these and every later number is negotiable.
- What you are counting. Customer churn (also called logo churn) counts accounts or users that left. Revenue churn counts the money that left. These answer different questions and, as you will see, often disagree.
- The time window. Monthly, quarterly, or annual. Pick one and label it. A "5% churn" with no period attached is not a metric, it is a rumor.
- The denominator. Who counts as "the customers you started with"? The usual choice is the count at the start of the period. The tricky question is what to do with customers who both signed up and left inside the same period. More on that below, because this is where most disagreements actually live.
The core formula
Customer churn rate for a period is:
churn rate = (customers lost during the period / customers at the start of the period) x 100%A worked example, with round numbers so the arithmetic stays out of the way. Say you begin the month with 1,000 active customers. During the month, 50 of them cancel or go inactive by your definition. Your monthly customer churn is:
50 / 1,000 = 0.05 = 5%That is the whole calculation. The numerator is who left, the denominator is who you started with, and the window is one month.
Revenue churn uses the same shape but swaps counts for money. Gross revenue churn is:
gross revenue churn = (recurring revenue lost during the period / recurring revenue at the start of the period) x 100%Here is why you compute both. Imagine those 50 lost customers were your smallest accounts. Your customer churn is 5%, but your revenue churn might be 2%, because the money walking out the door is a smaller slice than the logos. Reverse it, lose 50 of your largest accounts, and revenue churn can dwarf customer churn. Report only one and you can tell two honest-sounding stories from the same month. (There is a third flavor, net revenue churn, which subtracts expansion revenue from existing customers and can even go negative when your growing accounts outrun your leaving ones. Useful, but it is a different question, so keep it in its own column.)
The three variants people get wrong
Most churn arguments are not about the business. They are about one of these three definition slips.
Customer versus revenue churn. Covered above, and worth repeating because it is the most common mix-up in the room: someone quotes customer churn, someone else quotes revenue churn, and both think the other is wrong. Neither is. State which one you mean, every time.
The denominator problem. Take the same month. You started with 1,000 customers, lost 50, and also signed 200 new ones along the way. If your denominator is the 1,000 you started with, churn is 5%. If someone quietly switches the denominator to the 1,200 who touched the account at any point in the month, the same 50 losses now read as roughly 4.2%. Same reality, two numbers, and the gap grows the faster you are adding customers. Pick one rule for the denominator and never change it mid-comparison.
The time-window problem. A 5% monthly churn does not mean 60% annual churn. You do not multiply by 12, because each month you are churning a slightly smaller surviving base. Compounded, 5% a month works out closer to 46% a year. Short windows also exaggerate noise: one bad week can spike a weekly number that a monthly view would smooth out. Long windows do the opposite and can hide a recent slide inside a calm-looking annual average. The rule is simple. Compare like windows to like windows, and when you annualize, compound the survival rather than adding the losses.
Verify before you trust it
Before that number goes on a slide, run four checks. They take a minute and catch most embarrassing mistakes.
- Customers lost should never exceed customers at the start. If it does, your numerator is pulling in people who were not in the starting cohort. Recheck the denominator.
- Churn and retention should add up to about 100% for the same cohort and window. If retention is 90% and churn is 20%, one of them is measured differently. Find which.
- The same definition, period after period. If churn "improved" the same week someone changed the inactivity threshold or the denominator, you did not improve churn. You changed the ruler.
- When a number looks surprising, suspect the definition before the business. A sudden swing is far more often a pipeline change, a renamed event, or a new denominator than a real cliff. Confirm the number is real before you go explain it.
So what is "good"?
This is the question everyone actually wants answered, and it is the one where fabricated precision does the most damage. There is no single correct churn number, and any benchmark you see is a commonly-cited range, not a law. It varies widely by business model, customer segment, and company stage.
Roughly, and with all those caveats: established subscription software tends to see relatively low churn, often quoted in the low single digits to low double digits annually, with the healthiest enterprise products at the bottom of that band. Consumer and mobile products usually live much higher, because early retention decays fast and a large share of new users lapse within the first weeks. Early-stage companies swing more than mature ones, since a handful of accounts can move the percentage a lot. Treat any figure you read as a range that depends on the segment it came from, and compare yourself to your own trend and your own cohort before you compare yourself to a chart on the internet.
The number is the scoreboard, not the game
Here is the uncomfortable part. Everything above measures churn that already happened. The formula is a scoreboard, and the scoreboard records people who have already left. Calculating it correctly is table stakes. Most analytics tools will hand you the number.
The value is upstream of the number: seeing the accounts drifting toward the exit while they are still accounts, and doing something before the cancellation. That is the part worth building around. This heritage runs deep in live-service and games, where churn is measured in near real time and a bad week is not a slide, it is a crisis. At ThinkingAI, the closed loop is analyze, test, engagement, act, and it runs on infrastructure you host yourself, so the behavioral data never leaves your side. Agents detect the at-risk cohort and draft the intervention. People decide whether it ships. The churn number tells you the game is already lost for those users. The point is to still be playing.
So before you compare this quarter's churn to last quarter's, do the boring thing first: confirm both numbers are counting the same thing, over the same window, with the same denominator. A churn rate you can trust is worth more than a churn rate that is merely low.
Where to go next
Two follow-ups turn the number into something you can act on. The first is spotting at-risk users before they cancel, so the churn you calculate is churn you had a chance to prevent. The second is reading retention curves cohort by cohort, which shows you where in the lifecycle the leak actually starts rather than just how big it was. The calculation you just did is the entrance to both, not the destination.
