Churn meaning: what is churn, and how to calculate churn rate

What does churn mean?

What is churn? In business, churn means customers or users leaving: they cancel a subscription, do not renew a contract, or simply stop coming back. A customer who leaves has churned. Counting the people who leave gives customer churn, and the churn rate formula below turns it into a percentage: the share of the customers you had at the start of a period who left during it.

Outside business the word still means a butter churn, or to stir something hard. The business sense borrows the image of constant movement: some people arriving, others leaving, the total hiding how much is changing underneath.

Churn matters because growth hides leaks. If you add 200 customers a month and lose 150, the headline number still goes up, but most of your effort is going into replacing people. Churn makes that loss visible, and it caps how big you can get: at some point the customers you lose each month equal the ones you win.

Churn rate formula

Churn rate = customers lost during the period ÷ customers at the start of the period × 100%

Two rules keep the number honest:

• Only the starting customers can churn. People who sign up during the period are not in the denominator, and if some of them also cancel in the same period, report them separately (often called early churn) rather than adding them to the count.
• Fix the period before you compare. Monthly churn and annual churn are different numbers, and monthly figures from a month with five weeks of billing are not comparable with a four-week month.

Some companies divide by the average number of customers in the period instead of the starting count. T-Mobile, for example, defines churn in its Form 10-K for 2025 as customers whose service was deactivated as a percentage of the average number of customers during the period, divided by the number of months in the period. It also nets out customers whose service is restored within a set time and leaves out customers who had service for less than a minimum period. None of these choices is wrong, but two churn rates are only comparable if they use the same ones.

A worked example

A subscription product starts March with 1,200 customers paying $60,000 in monthly recurring revenue (MRR). During March:

• 60 customers cancel. Together they paid $9,000 a month.
• 40 customers downgrade, losing $2,000 of MRR.
• 70 customers upgrade, adding $4,500 of MRR.
• 150 new customers sign up.

The 150 new customers appear nowhere: they were not customers on 1 March, so they cannot have churned from that base. The gap between 5% customer churn and 18.3% revenue churn says the customers who left were bigger than average, which a customer count alone would hide. If upgrades had added more than $11,000, net revenue churn would be negative.

Types of churn

Splitting voluntary from involuntary churn is worth doing early. Involuntary churn is usually fixed with payment retries, card updater services and reminder emails, not with product changes, so mixing the two sends you after the wrong problem.

Activity churn needs a definition, because nobody "cancels" a free app or a website. Pick an inactivity window that matches how often people should use the product (for example 7 days for a daily app, 30 for a weekly one, 90 days without a purchase for a store) and write it down next to the number. Change the window and the churn rate changes with it.

Monthly vs annual churn

Churn compounds, so monthly churn cannot be multiplied by 12. The customers you still have keep churning from a smaller base each month:

Annual churn = 1 − (1 − monthly churn)¹²

Multiplying 5% by 12 would suggest 60%; the real figure is 46%. Going the other way, an annual churn of 20% is about 1.8% a month, not 1.67%.

Churn vs retention, attrition and dormancy

Cohort analysis is the usual way to look at both at once: group customers by the month they started and follow what share is still active, and what share has left, in each later month.

What is a good churn rate?

There is no single good number. Churn depends on price, contract length, who the customers are and how churn is defined. A business selling annual contracts to large companies loses far fewer customers per month than a low-priced consumer subscription, and an app measuring activity churn with a 7-day window will report a higher rate than one using 30 days.

Public companies show the range. T-Mobile's 2025 10-K reports monthly postpaid phone churn of 0.93% for 2025, up from 0.86% in 2024, and prepaid churn of 2.72%. Postpaid customers are on longer-running plans than prepaid ones, and their churn is about a third of the prepaid rate at the same company.

The useful comparisons are your own: this month against previous months, and one cohort, channel or plan against another, all calculated the same way.

Why customers churn

Most churn traces back to a few causes:

• They never reached the value they signed up for. Users who do not complete setup or never use the core feature leave first.
• The product stopped fitting. Their needs grew, shrank or changed.
• Price. A price rise, a cheaper alternative, or a budget cut.
• Payment failed. Involuntary churn from expired or declined cards.
• A bad experience. Bugs, slow support or a failed migration.

The rate tells you how many left; it does not tell you why. That comes from looking at what churned users did, and did not do, before they left.

How to measure churn with product analytics

Billing data tells you who cancelled. Product analytics shows who is drifting away before they cancel, and who leaves products that have no subscription at all. In TapCub:

• Lifecycle analysis splits active users in each period into new, returning and resurrecting, and draws users who were active and then absent in the next period as negative dormant bars. You choose which event counts as activity. See Lifecycle in the docs.
• Slipping away counts users inactive for at least 7, 14, 30, 60 and 90 days as of the end of the range. It is the early-warning list for activity churn: people in the 14-day bucket have not churned yet by a 30-day definition.
• Composition (for apps) splits each week's active users into new, returning and continuing, and shows last week's active users who did not come back as negative churned bars, with a change ratio of this week's active users to last week's. See Composition.
• Product Engagement Score includes a growth part based on the quick ratio, (new + recovered) ÷ dropped users: above 1 the active base is growing, below 1 it is shrinking. See Product engagement score.
• Churn-risk list. One button on the People page creates a "Churn risk" segment, refreshed daily: people who purchased in the last 90 days and have had no pageview in the last 14. It then opens a new automation that runs for that segment every Monday morning, for example to send the list to your team.

Because churned users keep their history, you can segment them by acquisition channel, country, device or any property you send, and compare their first weeks with users who stayed.

How to reduce churn

1. Find where it is concentrated. Split churn by plan, channel, cohort and customer size before changing anything. A churn problem is usually a problem in one segment.
2. Shorten the path to first value. Compare what retained users did in their first days with what churned users did, and build onboarding around the difference.
3. Recover failed payments. Retries, card updates and reminders address involuntary churn directly.
4. Act before the window closes. Reach users in the 14-day slipping-away bucket while they still remember the product, instead of after they count as churned.
5. Ask the people who leave. A one-question cancellation survey gives the reasons the data cannot.

Common mistakes

• Putting new customers in the denominator. Using the end-of-period count, or adding new sign-ups, makes churn look lower in fast-growing months.
• Mixing monthly and annual plans. Annual customers can only churn at renewal, so blending them with monthly ones understates monthly risk. Report them separately.
• Multiplying monthly churn by 12. Churn compounds; use 1 − (1 − monthly churn)¹².
• Changing the inactivity window. Activity churn with a 7-day window and with a 30-day window are different metrics. Keep one definition per report.
• Only tracking the rate. The rate says how many left. Segmenting who left, and what they did before leaving, says why.

Related guides: Churn Rate vs Retention Rate and SaaS Metrics. See also customer churn analysis and the leaky bucket.