Churn rate: the formula, and why one business has three honest answers

Nobody argues about the numerator. The customers left; you can count them.

The argument is always about the denominator. Divide those losses by a different base and the same month reports a different churn rate — with nobody lying. That is why two people in one meeting can quote two numbers and both be right, and it is why churn matters so much to customer lifetime value, which takes churn as a direct input. Get the denominator wrong and every rupee of LTV downstream is wrong with it.

The churn rate formula, stated once

Churn rate = customers lost in the period ÷ the base of customers you are measuring against.

That is the whole formula. The useful part is the second half of the sentence, because "the base" has three common readings:

• Customers at the start of the period. The simplest. Ignores everyone who joined during the period.
• Average customers over the period. Usually the average of the opening and closing counts. Smooths out fast growth.
• A fixed cohort. Only the people who were already customers on day one, tracked forward. The strictest reading of "who left us".

Pick one, write it down, and use it every month. Switching bases between reports is how churn becomes a debate instead of a measurement.

Three denominators, three answers.

The same month, computed three ways

Illustration, not data. The numbers below are hypothetical, chosen to make the spread visible. Say a subscription business reports this June:

• Customers on 1 June: 1,000
• New customers joined in June: 200
• Customers lost in June: 60 — of which 48 were there on 1 June and 12 joined and left inside the month
• Customers on 30 June: 1,140

Same month. Same business. 4.8% to 6.0% — a quarter's difference in relative terms, entirely from the denominator.

Now carry it into LTV. Lifetime value for a subscription is roughly average revenue per customer ÷ churn rate. At ₹500 a month, the 6.0% figure implies about ₹8,300 of lifetime value and the 4.8% figure about ₹10,400. Nothing changed except what you divided by. The cohort method itself — and when it is worth the effort — belongs to our guide to cohort analysis.

Customer churn, revenue churn and net revenue churn

Counting accounts and counting money are different questions. If all your customers pay the same price they give the same answer. The moment prices differ, they diverge sharply — you can lose 2% of accounts and 20% of revenue in the same month.

Two rules that stop most confusion. Retention is only 100% minus churn when both use the same unit and the same denominator — net revenue retention is not one minus customer churn. And never quote net revenue churn to a retention team: expansion from big accounts can hide a steady bleed of small ones.

If you are deciding which of these belongs in a monthly report at all, that is a reporting question — our notes on choosing marketing KPIs and building a marketing dashboard cover the selection.

Why a published "average churn rate" is almost always unusable

We are not going to give you a benchmark number here, and that refusal is the point of this section.

Look at what real filers do with the same word. Netflix's 2004 annual report defined churn as "customer cancellations in the quarter divided by the sum of beginning subscribers and gross subscriber additions, then divided by three months" — a fourth denominator again, with gross additions included. A telecom glossary filed by VEON divides disconnections by "the average number of prepaid or postpaid customers for the period", and counts a customer who disconnects and reconnects on the same number as churn.

Both are defensible. Neither is comparable to the other. So a blog post reporting "the average churn rate in your industry" is usually averaging numbers built on different arithmetic — and it typically does not say which.

Five things a usable benchmark would have to disclose, and almost none do:

• The denominator. Start, average, cohort or start-plus-additions.
• The unit. Accounts, revenue, seats or licences.
• The period. Monthly and annual figures are not interchangeable — more on that below.
• The contract length. On annual contracts a customer can only leave at renewal, so monthly churn is near zero for eleven months and spikes in the twelfth.
• What counts as churn. Pauses, downgrades, failed payments and win-backs are each in or out, and the choice moves the number.

Regulators take this seriously for exactly this reason. The SEC's 2020 guidance on metrics in company reporting says it would generally expect a disclosed metric to be accompanied by "a clear definition of the metric and how it is calculated", plus why it is useful and how management uses it — and that a change in how it is calculated should itself be disclosed. If a filing has to say it, your monthly review can too.

On periods: monthly churn does not multiply. At 6% a month, a year of compounding leaves you with roughly 52% of the base gone, not 72%. Each month's loss comes off a smaller base than the last.

Churn is partly a marketing problem, not only a product one

A chunk of next quarter's churn was decided before the customer ever paid you. It was decided by the promise in the ad, the discount that closed them and the audience your targeting bought.

• Discount-led acquisition buys discount-led customers. They respond to the next offer too — someone else's.
• Chasing a low cost per acquisition widens targeting. The cheapest customers to buy are often the ones with the weakest reason to stay.
• An overstated promise churns on contact with the product. The ad wrote a cheque the onboarding could not clear.
• Soft offers bring volume that never activates. A free download and a free trial attract different intents.

Which means part of the fix lives in the media plan. Knowing which customer segments stay, and where in the customer journey people quietly disengage, changes what you buy — not just what you build.

How we read churn next to acquisition data (our own approach)

Digital Hangover's view, from our own client work. The pattern we see most often is a client whose acquisition cost looks perfectly healthy while growth has flattened anyway. The blended churn number rarely explains it. So we stop reading the blend.

What we do instead, in this order:

1. Split churn by the channel and campaign that acquired each customer. One number for the business is an average of things that behave nothing alike.
2. Group those customers by joining month and track each group forward. A stall is either a supply problem or a survival problem, and the groups tell you which.
3. Read each channel's offer beside its churn. Where the discount was deepest, the survival curve is usually steepest.
4. Re-price the channel. If a channel's customers leave sooner, its true cost of acquisition is higher than the dashboard says, because you buy the same customer again sooner.

Our working rule from this: we will not put a blended churn figure in a monthly review without the acquisition-channel split next to it. On its own it is the number most likely to send a client to fix the product when the problem is in the buying. We do not promise this reverses a stall — it tells you where to spend the next month looking.

What actually moves churn, honestly scoped

The tactical playbook for keeping customers is its own subject, and we have handed it to our guide on customer retention rather than repeat it here. What belongs on this page is the scoping — five things that are true before any tactic:

• Fix the definition first. You cannot move a number that changes shape between reports.
• Separate voluntary from involuntary. Failed cards and expired mandates are churn you lose by accident, and usually the cheapest to recover.
• Most cancellations are decided early. The first few weeks carry more weight than any later intervention.
• You can only change churn for future cohorts. Acquisition quality is already locked in for the customers you have.
• Some churn is structural. Seasonal products and one-time needs churn by design. No loyalty programme fixes that, and chasing it wastes budget.

For the "why did they leave" half of the question, asked rather than inferred, the usual instrument is a survey at cancellation and an ongoing NPS measure — both are leading signals, where churn is a trailing one.

What one written-down definition buys you

A churn number that means the same thing in January and in July becomes comparable to itself. That is the only comparison that was ever available to you — and it is enough to tell whether last quarter's changes worked.

It also makes LTV trustworthy, makes channel-level cost of acquisition honest, and ends the meeting where two people quote two numbers. One paragraph in a document, agreed once, does all of that.