Subscription churn - how to predict and prevent it

Churn is rarely sudden. It's almost always preceded by signals — a drop in usage, silence, a failed payment, a lower score. The problem is that most companies learn about a departure only when the customer has already clicked "cancel". And that's the worst moment: the decision is made, the relationship is over, and the cost of acquiring that customer has just been lost. This article shows how to read the signals early, how to separate voluntary churn from technical churn, and how to build a simple system that keeps the customer before they leave.

What churn in subscriptions is

Churn is the share of customers or revenue that leaves in a given period. It sounds simple, but a single number can mislead you, because churn has several faces.

It's worth separating at least two views. Customer churn tells you how many customers left. Revenue churn (MRR) tells you how much money left — and that's usually the more important number, because losing one big customer hurts more than losing several small ones. On top of that comes the difference between gross churn (pure outflow) and net churn (outflow minus upsell and plan upgrades among customers who stayed). A company can lose customers and still grow in net revenue if the rest of the base buys more.

Look at churn in a cohort view too — how customers acquired in a given month behave in the following months — not just a single averaged value. Averaging hides the most important thing: at which point in the lifecycle the customer most often drops off. More on which metrics are actually worth tracking is in MRR, churn, LTV — subscription metrics.

Voluntary vs involuntary churn — two different diseases

The most important distinction in the whole topic: churn splits into voluntary and involuntary, and each is treated completely differently.

Voluntary churn is a conscious customer decision — they cancel because they stopped seeing the value, their situation changed, or they found an alternative. Involuntary churn (technical) is losing a customer who didn't want to leave at all — their payment simply failed: the card expired, funds ran out, the bank declined the transaction.

Why is this distinction key? Because confusing the two means treating the wrong disease. If you send a discount offer "to keep them" to a customer who dropped off because of an expired card, you waste margin on someone who never wanted to leave. And if you only retry the charge for a customer who lost the value, they'll leave anyway — just a month later.

Involuntary churn: the cheapest to recover, the most often ignored

Involuntary churn is usually the quietest and easiest-to-recover leak in the whole base — and yet most companies have no process for it at all.

The tools that genuinely reduce it:

• Smart retries — re-attempting a failed charge in planned windows (e.g. after a few hours, after a day, after a few days), not hammering the gateway on repeat. The timing is chosen to hit the moment funds are back on the account.
• A dunning sequence — automatic communication to the customer (email/SMS) telling them the payment failed, with a simple path to update the method. Most recoveries are simply a well-built reminder sequence.
• Method updates and alternatives — making it easy to swap an expired card and offering a method that doesn't expire.

Here's where the payment method itself gives an edge. Recurring BLIK payments don't expire like a card and fail less often on bank declines, so at the source they generate less involuntary churn than card-only subscriptions — especially on the Polish market. How they work is covered in Recurring BLIK payments — what they are. The full recovery playbook (retry windows, dunning copy, how much MRR you can recover) is in Recovering failed payments (retry/dunning).

Voluntary churn: why the customer really leaves

Voluntary churn almost always comes down to one thing: the customer stopped seeing value proportional to the price.

The most common causes worth recognizing:

• Weak onboarding — the customer never reached a first real result ("first value"), so the next charge looks like paying for nothing.
• A drop in usage — the product fell out of habit; the less frequent the use, the closer the cancellation.
• No visible result — the customer can't point to what the subscription gave them.
• Price vs perceived value — it's not "too expensive", it's "too expensive for what I get out of it".
• A change of situation or a competitor — some churn is natural, but some can be recovered with the right offer.

How to predict departures: early-warning signals

Departures are preceded by signals — usually weeks, sometimes months before the cancellation. The art is to collect and read them before the customer makes the decision.

Four groups of signals worth watching:

• Engagement — a drop in frequency and depth of use, longer gaps, fewer key actions.
• Payments — failed charges, plan downgrades, a drop in purchase frequency. These are some of the earliest and most objective signals.
• Satisfaction — a low score, especially detractors in an NPS survey. The moment of measurement is key — the most is said by NPS collected after a payment, not at random.
• Contact — either a sudden rise in complaints and tickets, or complete silence after earlier activity.

Customer Health: combine the signals into one risk picture

A Customer Health Score is a single indicator that combines scattered signals into a risk ranking — letting you act on a number rather than a hunch.

A simple version you can roll out without a big project:

• Pick 3–5 signals you actually have at hand (e.g. usage, failed payments, the latest NPS score).
• Give them weights and combine them into one score.
• Split the base into tiers — e.g. green (stable), yellow (watch), red (at risk).
• Assign an action to each tier so the score triggers a response right away.

How to spot a specific customer on the edge and what to do about it, I expand in Customer Health — spotting at-risk customers.

How to keep the customer: a four-step playbook

Stopping churn is a process, not a single action. Four steps that close it:

• 1. Segment the risk. Separate at-risk customers from stable ones using Customer Health — you can't respond sensibly to the whole base at once.
• 2. React to the cause. Differently for involuntary (retry, dunning, method swap), differently for voluntary (a value reminder, contact, a tailored offer).
• 3. Offer an alternative to cancellation. "Cancel" doesn't have to be the only option — a pause, a plan change or a billing-cycle change save some customers.
• 4. Close the loop. Measure whether the interventions actually lower churn, and iterate. Without this step you don't know what works.

Pause and winback: two levers that recover LTV

The two most underrated retention levers are pause and winback — both recover revenue that would otherwise simply be lost.

Pause instead of cancellation stops some of the departures that were inevitable at that moment — the customer doesn't break the relationship, just suspends it and returns. It lowers the friction of the "stay or leave" decision. The mechanics and impact on LTV are laid out in Pause instead of cancellation.

Winback is the deliberate recovery of former customers — accounting for the reason they left, the right moment, and a tailored offer. A former customer already knows the product, so the cost of winning them back is often lower than acquiring a new one. How to set it up is in Winback — how to recover former customers. Both translate directly into LTV.

Payment as the earliest and cheapest warning system

Of all the signals, payment data is the earliest, most objective and cheapest to collect — you have it anyway, at every billing cycle. A failed charge, a downgrade or a drop in frequency are hard facts, not declarations. That's why it's worth treating payment not as the end of a transaction, but as an early-warning system against churn.

This closes the logic I describe in the pillar piece Payment as a tool for retention, loyalty and LTV growth and in the piece on how payments affect customer retention. If you're building subscriptions or recurring dues in a practical way, start with recurring payments and the subscription model.

The most common mistakes in fighting churn

• Looking only at one averaged churn number instead of net revenue churn and cohorts.
• Confusing voluntary with involuntary churn and treating the wrong disease.
• Reacting only after the "cancel" click, when the decision is already made.
• No alternative to cancellation — only "stay" or "leave".
• Collecting NPS at a random moment instead of a point that actually says something.
• No closing of the loop — interventions without measuring effectiveness.

FAQ

What is churn in subscriptions?
Churn is the share of customers or revenue that leaves in a given period. It's worth counting it not as one averaged value, but by separating customer churn from revenue churn (MRR) and gross from net.

How does voluntary churn differ from involuntary?
Voluntary is a conscious cancellation by a customer who stopped seeing value. Involuntary is losing a customer through a failed payment (expired card, no funds). The first is treated with value, contact and an offer, the second with retry and dunning.

How do you predict that a customer will leave?
By early-warning signals: a drop in usage, a low NPS score, failed payments, a downgrade, silence from the customer. Combined into one Customer Health picture, they let you act before the decision is made.

What is a Customer Health Score?
It's a single indicator combining selected signals (usage, payments, satisfaction) into a risk ranking. It splits the base into tiers — e.g. green, yellow, red — and assigns a specific action to each.

How do you reduce involuntary churn?
With smart retries in planned windows, a dunning sequence with an easy method update, and a payment method that doesn't expire — like recurring BLIK payments.

Does a pause really reduce churn?
Yes — a pause gives the customer a third option instead of the binary "stay or leave". Some people who would have cancelled at that moment suspend the subscription and return, which saves LTV.

How do payments help predict churn?
Payment data — failed charges, downgrades, a drop in frequency — are the earliest and most objective risk signals. You have them at every billing cycle, so they work like a cheap early-warning system.