Churn Measurement
A working framework for measuring churn in subscription and repeat-purchase stores: which denominator to use, customer vs revenue churn, and how rolling windows change the number.
Churn Measurement
The methodology for computing churn — choosing denominator, window, and whether you're counting customers or revenue — so the number is decision-useful.
Churn measurement is the set of decisions that sit underneath a single churn percentage: which customers are in the denominator, whether you're counting accounts or euros, whether the window is a calendar month or a rolling 30 days, and whether you net out reactivations and expansion. Two analysts looking at the same Shopify subscription data can publish churn numbers that differ by a factor of two and both be defensible — so the framework matters more than the formula.
Done well, churn measurement gives operators a number that moves when the business moves. Done badly, it produces a vanity figure that's stable, comfortable, and useless for forecasting retention or sizing a winback budget.
Most retention dashboards report churn as a single number. The problem is that the number depends on at least four choices the dashboard rarely surfaces: who counts as a customer, when they count as churned, what unit you're measuring, and over what window. Change any one and the figure moves.
This page is the methodology layer above the Churn Rate metric itself. If you want the basic formula and a quick computation, use the Churn Rate Calculator. If you're deciding between framings, keep reading.
The denominator problem
The single biggest source of disagreement is the denominator. The naive formula — customers lost this month divided by customers at the start of the month — quietly assumes nobody joined mid-period. For a fast-growing subscription brand acquiring 800 new customers a month on a base of 5,000, that assumption understates churn by 10-15% because the new joiners barely had a chance to churn.
Two cleaner alternatives: use the average of start-of-period and end-of-period customers, or restrict the denominator to customers who were already active at the start of the window (a cohort view). The cohort view is the most honest for a growing brand because it isolates retention from acquisition mix. Pick one definition and lock it — the worst outcome is silently switching denominators between quarters.
Customer churn vs revenue churn
Customer churn counts heads: how many subscribers cancelled. Revenue churn counts euros: how much recurring revenue walked out the door. They diverge whenever your customer base isn't homogeneous — and in a subscription store with tiered boxes or variable basket sizes, it never is. A €19 entry-tier customer churning hurts the customer count just as much as a €89 premium customer, but only one of them moves the P&L. See Customer vs Revenue Churn for the full breakdown.
Then there's gross vs net. Gross churn is the lost revenue without any offset. Net churn subtracts upgrades, add-ons, and reactivations from the same cohort. A healthy DTC subscription business often shows 6% gross monthly churn and 3-4% net once you credit back the customers who skipped a month and returned, or upsized their box. Report both — net alone hides whether you're plugging the leak or just pouring water in faster.
The skipped-month trap
Subscription brands that let customers pause or skip a delivery face a classification choice: is a 60-day inactive customer churned, or paused? Pick a threshold (90 days is common for monthly cadence, 180 for quarterly) and apply it consistently. Brands that flatter their churn number by treating long pauses as 'still active' are setting up a nasty correction the first time finance does a real cohort analysis.
Windows, rolling averages, and what you compare against
Monthly churn is the default reporting cadence, but a single month of data is noisy for any brand under ~3,000 active subscribers. A trailing 90-day rolling rate smooths the spikes without lagging so far that you miss a real shift. For board-level reporting, annualised churn (1 − (1 − monthly)^12) is more legible but exaggerates small monthly errors — a 1-point monthly miss becomes a 10-point annual one.
Whatever cadence you choose, compare like with like. Churn rate against the previous period is the operating view. Churn against the inverse retention figure is the sanity check — see Churn vs Retention Rate if you ever need to reconcile the two. And benchmark cautiously: a 5% monthly churn rate is excellent for a €30 beauty subscription and catastrophic for a €250 annual prepay.
How window choice changes the same churn signal
Monthly (raw)
90-day rolling
Churn measurement FAQ
Churn rate = customers lost during the period ÷ customers at the start of the period. For a growing brand, use the average of start and end customers in the denominator, or restrict to a starting cohort. Apply the same definition every period — consistency matters more than which variant you pick.
Take the customers active on the first day of the month, count how many cancelled or hit your inactivity threshold by month-end, and divide. Decide upfront whether paused or skipped subscriptions count as churned — a 90-day inactivity rule is the common cut-off for monthly cadence boxes.
Report both. Customer churn tells you about product-market fit and onboarding; revenue churn tells you about commercial impact and whether high-value customers are leaving faster than the average. If you only have bandwidth for one on the leadership dashboard, revenue churn is more decision-useful.
Gross churn is lost revenue without any offset. Net churn subtracts reactivations, upgrades, and expansion within the same cohort. A subscription brand might run 6% gross and 3% net monthly — the gap is the size of your retention plays. Best practice is to publish both side by side.
Pick an inactivity threshold and apply it consistently. For monthly subscriptions, 60-90 days without a shipment usually counts as churned. For quarterly cadence, 180 days. The wrong move is leaving a customer 'active' indefinitely just because the account isn't formally cancelled — that flatters churn and produces a nasty correction later.
They're complements, not synonyms. Retention rate = 1 − churn rate when both use the same denominator and window. In practice, teams often calculate them differently (different cohorts, different inactivity rules) and they don't reconcile. See Churn vs Retention Rate for how to align them.
For consumable monthly boxes (beauty, food, supplements), 5-7% monthly churn is typical, 3-5% is strong, under 3% is excellent. For higher-priced, lower-cadence subscriptions (€80+ quarterly), expect 8-12% per quarter. Benchmarks are only useful within the same price tier and cadence.
Use monthly for operational reviews, 90-day rolling for trend analysis, and quarterly or annualised for board reporting. Single-month churn is too noisy for brands under ~3,000 subscribers — the variance can swing 2-3 percentage points on random shipping delays alone.
Define a purchase window based on your typical reorder cadence — usually 2-3x the median repeat interval. A customer is 'churned' once they pass that window without ordering. For an apparel brand with a 90-day median repeat cycle, 180 days of inactivity is a reasonable churn flag.
Almost always denominator drift or threshold inconsistency. Finance is probably using a cohort-based denominator and a different inactivity rule than your subscription platform's default. Lock a single methodology in writing, document the choices, and have one source of truth — usually the data warehouse, not the billing tool.
Start tracking the metric that matters
Free for 30 days. No credit card. Connect your site in 4 minutes and see the one metric driving revenue.