Growth gets the applause. But churn decides the ending. A business adding customers through a leaky bucket isn’t growing. It’s running to stand still, and paying acquisition costs for the privilege.
📌 TL;DR: Churn rate = customers (or revenue) lost in a period ÷ what you started with. Customer churn counts logos; revenue churn weighs them, and revenue churn is the one the business feels. Falling churn compounds as powerfully as growth does.
What Is Churn Rate?
Churn rate is the percentage of customers (or revenue) lost over a period. Start the quarter with 400 customers, lose 20, and quarterly customer churn is 5%. It’s the core measure of retention health. And the denominator discipline matters: define who counts as ‘active’, and keep the definition stable.
Customer churn vs revenue churn
Customer churn counts departures equally: losing a $100/month customer and a $10,000/month customer are each ‘one churn’. Revenue churn weighs them, and reveals what the business actually feels. The two diverge constantly: healthy logo churn among tiny accounts can coexist with alarming revenue churn if big accounts are quietly leaving. Track both, act on revenue.
What actually reduces churn?
- Seeing it early: usage decline, support silence, and champion departure precede cancellation by months; churn-risk signals are a data problem before they’re a success problem.
- Onboarding that lands: most churn is decided in the first ninety days, cancelled much later.
- The right customers: churn is often an acquisition problem wearing a retention costume: customers who never fit were always going to leave.
- Honest exit data: knowing WHY customers leave beats any dashboard of THAT they left.
Churn is one of the three business-metric siblings: it drains the lifetime value that acquisition paid for, and cohort analysis is how you see which customer vintages churn differently, and why.
The churn math traps
Churn looks like division and hides half a dozen ambushes. The denominator question: customers at period start, or an average across the period? Growing businesses get flattered by start-of-period denominators. Pick one convention and freeze it. The new-customer question: someone who joined and left within the month: churned, or excluded? Either answer works. Mixing them across reports doesn’t.
The pause question: downgrades to free tiers, seasonal pauses, failed payments in dunning: each needs a ruled status, because ‘sort of churned’ corrupts every downstream metric. The reactivation question: a churned account that returns: new customer or resurrection? Your CLV math cares about the answer.
And the aggregation trap: monthly churn doesn’t multiply into annual by simple arithmetic: 3% monthly compounds to roughly 31% annually, not 36%. Small confusion, large planning errors.
Diagnosing churn like an analyst
A single churn number is a smoke alarm. Diagnosis needs cuts. By cohort: are newer vintages retaining better (is anything we changed working)? By segment: which sizes, industries, or plans leak fastest (is churn actually an acquisition-targeting problem)? By tenure: where’s the cliff? Month two (onboarding), month thirteen (first renewal), or random (product-market drift)? By reason: involuntary payment failures vs deliberate departures. The fixes share nothing but the metric.
The diagnostic that changes the most minds: churn by ICP fit. Score churned accounts against your ideal profile, and watch a chunk of ‘retention problem’ reclassify as ‘we sold to the wrong companies’. That finding moves budget from save-campaigns to targeting, where it belongs.
The revenue-churn refinements
- Gross revenue churn: MRR lost to departures and downgrades; the leak itself.
- Net revenue churn: the leak minus expansion from surviving accounts; negative net churn (expansion outrunning loss) is the compounding engine every subscription business chases.
- Logo vs revenue divergence: healthy logo churn with ugly revenue churn means big accounts are leaving; the reverse means the long tail is; the pair diagnoses what the single number obscures.
Real-World Examples
Numbers stick better with faces on them. So here are three churn stories you’ll recognize.
A SaaS company loses 16 customers in a quarter. Fifteen are small subscription accounts worth $99 a month. One is an enterprise deal worth $8,000 a month. Logo churn says 4%, nothing scary. But revenue churn is triple that, because the one enterprise cancel outweighs the other fifteen combined. Same quarter, two very different stories.
A gym chain panics every January when ‘churn’ spikes. Except half those members didn’t cancel. They paused for the season and came back in March. Once pauses got their own status, real customer churn turned out flat year-round. The metric was broken, not the retention.
And a subscription box brand found a third of its MRR loss came from expired cards, not unhappy customers. A dunning flow and a card updater recovered most of it in a month. No product change required.
From measuring churn to predicting it
Mature retention work shifts tense: from ‘how much did we lose?’ to ‘who is about to leave?’. The raw material is behavioral: usage declining against the account’s own baseline, champions going quiet, support sentiment souring, invoices slipping. Assembled per account and fed to a scoring model (or even honest rules), these signals convert churn from an autopsy metric into a work queue. Accounts get ranked by risk, then routed to humans while the renewal is still alive.
Two disciplines keep prediction honest. Measure the intervention, not the model. A brilliant risk score that changes no outreach saves nothing; hold out a control group and prove the save-motion moves renewals. And audit the inputs. Risk scores computed on stale contact and usage data predict yesterday’s risk; the data freshness underneath the model IS the model’s ceiling.
The one-page churn review
A monthly ritual worth institutionalizing: one page. It shows gross and net revenue churn, logo churn by segment, the cohort view of newest vintages, top five churn reasons from exit data, and involuntary-churn recovery rates. Twenty minutes in the revenue meeting, same format every month. Trends surface, arguments shrink, and the retention conversation stays anchored to evidence rather than the loudest recent loss.
Frequently Asked Questions
What is churn rate in simple terms?
The percentage of customers or revenue you lose over a period, the speed of the leak in your bucket. Lose 20 of 400 customers in a quarter, and quarterly churn is 5%.
How do you calculate churn rate?
Customers lost during the period ÷ customers at the start of the period × 100, with a stable definition of who counts as active. Revenue churn swaps customer counts for revenue amounts.
What is a good churn rate?
It varies enormously by market, price point, and customer size. Enterprise contracts churn far less than SMB subscriptions. The honest benchmark is your own trend: churn falling beats any external comparison.
What is negative churn?
When expansion revenue from existing customers exceeds revenue lost to departures and downgrades, putting net revenue churn below zero. The business grows even with zero new sales; the strongest signal of product-led health there is.
What is involuntary churn?
Loss from payment failure rather than decision: expired cards, failed transactions. It’s the cheapest churn to fix (dunning flows, card updaters) and routinely a meaningful share of the total; separate it from deliberate churn before diagnosing anything.
Can churn be predicted?
Usefully, yes: declining usage, champion silence, and payment friction precede most deliberate churn by months, and models ranking accounts by these signals reliably beat unranked outreach. The proof standard is saved renewals against a control group, not model accuracy.