When churn goes up, the reflex in most marketing teams is the same: build a win-back campaign. A "we miss you" email, a 20% discount, maybe a second reminder a week later. It feels decisive. It also treats every departed customer as if they left for the same reason, and quietly assumes the reason was price. For most businesses, neither assumption is true.
Here is the reframe that changes how you approach it: churn is not one problem. It is several different problems wearing the same label. A customer whose card expired, a customer who never figured out how to use the product, a customer who had a terrible support experience and a customer who simply moved house all show up as the same number on a dashboard. Each needs a completely different response, and a blanket discount fixes, at best, one of them. Diagnosis has to come first.
First, split off the churn nobody chose
Before investigating why customers decided to leave, separate out the customers who never decided at all. Involuntary churn happens when a payment fails: a card expires, gets replaced after fraud, hits its limit or is declined by the bank for an unexplained reason. The customer still wanted the product. The business just stopped taking their money.
This is often far bigger than teams expect. Industry estimates commonly put involuntary churn at somewhere between 20 and 40% of total churn for subscription businesses, and The Seattle Times famously found that most of its churn came from failed payments. It is also the easiest churn to fix, because the solutions are operational: reminding customers before a card expires, retrying failed payments at sensible intervals, using card-updater services, offering alternative payment methods, and sending clear, friendly messages by email or SMS asking customers to update their details. Fix this first. It is the rare retention problem where a modest effort recovers customers who were never unhappy.
For businesses without subscriptions, the equivalent step is defining churn sensibly in the first place. If customers typically reorder every two months, someone who has not bought in six months has probably lapsed; someone who has not bought in five weeks has not. Base the window on your own repurchase patterns, which a cohort analysis reveals, and make sure duplicate customer records are not inflating the count, a common problem that CRM data hygiene fixes.
Upsell
Maximizes revenue with post-purchase upsell offers on your Shopify thank you page
Best for: Post Purchase Upsell
The five kinds of churn
Once involuntary churn is out of the way, most voluntary churn falls into a handful of groups. Naming them explicitly is the whole point of diagnosis, because each has different evidence and a different fix.
| Kind of churn | Tell-tale signs | The right fix |
|---|---|---|
| Failed payment | Declined card, still active until billing | Billing reminders, retries, card updates |
| Never reached value | Low early usage, features never tried | Better onboarding and early guidance |
| Value or price mismatch | Used it, then "not worth it" | Right-sized plans, clearer value |
| Bad experience | Complaints, tickets, poor reviews before leaving | Fix the product or service issue |
| Need has ended | Moved, finished, outgrew the product | Usually nothing; do not chase |
Competitor switching cuts across several of these groups, since customers usually move to a rival because of value, experience or fit, so treat it as a signal to investigate rather than a category of its own. And accept that the last group is healthy churn: a business will always lose some customers whose need has genuinely ended, and spending money trying to recover them is wasted.
Four sources of evidence
No single source tells you why customers leave, so good diagnosis combines several and looks for agreement between them.
Behaviour is the most honest source because it shows what happened rather than what people say. Look at the weeks before someone left: did usage or purchase frequency fall away gradually, stop abruptly, or never really start? Which features, products or services did they never touch? These are the same signals that power behaviour-based scoring, pointed at departures rather than prospects.
Stated reasons from exit surveys and cancellation flows are quick and scalable. Unstructured feedback from support tickets, complaints, reviews and chat logs often contains the specific frustration a tick-box survey hides. And conversations with a handful of customers who have left, ideally within a couple of weeks of their departure, reveal the real sequence of events in a way nothing else can. Pulling these sources together is far easier when customer data lives in one place, one of the practical arguments for consolidating around one customer record.
The reframe Churn is not one problem. It is several different problems wearing the same label. A failed card, a customer who never got set up, a bad support experience and a finished need all look identical on a dashboard, and a blanket discount fixes at most one of them.
"Too expensive" is where diagnosis starts
Almost every exit survey is dominated by the same answer: price. It is the easiest, least awkward thing to tick, and it is rarely the whole story. Very often "too expensive" means "not worth it to me", which is a value problem rather than a pricing one: the customer never got the product working properly, used a fraction of it, or found it did not quite fit what they needed. It can also mean they found a cheaper alternative, their circumstances changed, or they simply stopped using it.
The way to tell these apart is to check stated reasons against behaviour. A customer who says the price is too high but stopped using the product two months before cancelling almost certainly left for another reason. Add follow-up questions where you can, such as "compared with what?" or "what would have needed to be different?". And talk to people: five to ten short conversations with recent churners in one segment usually tell you more than hundreds of survey responses, especially if you offer a small thank-you for their time and make clear you are not trying to sell them anything.
A subscription business loses 1,000 customers in a quarter. Diagnosis sorts them roughly like this:
→ 300 failed payments → need a billing fix, not a discount
→ 250 never reached value → need a guided restart
→ 200 value or price mismatch → a price-based offer might work
→ 150 bad experience → need the problem fixed and acknowledged
→ 100 need has ended → not worth chasing
A blanket 20% discount speaks to only about one in five of these customers. The other 800 get an offer that ignores why they left, and some of them learn that cancelling earns a discount.
The better plan: five different responses, each matched to a group, with the biggest and easiest group, failed payments, fixed first.
Segment before you conclude
Churn reasons rarely spread evenly across a customer base, and averages hide the pattern. Break the analysis down by acquisition channel, first product or plan, customer size, region and how long each customer stayed. You will often find that one channel brings in customers who churn quickly for the same reason, or that people on a particular plan leave at a much higher rate.
Timing matters too. Customers who leave in the first weeks usually never reached value, which points to onboarding. Customers who leave after a long, steady relationship usually hit a change: a price increase, a service failure, a competitor offer or a change in their own circumstances. That distinction alone often tells you which team owns the fix. It connects directly to the economics of long-lived customers we discussed in recurring revenue, where a small improvement in retention compounds over every future month.
Then, and only then, win them back
With a diagnosis in hand, win-back stops being a single campaign and becomes a set of targeted responses. Failed-payment customers get an easy, friendly way to update their details. Customers who never reached value get a guided restart that shows them what they missed, not a discount on the same experience that failed them. Customers who had a bad experience get an honest acknowledgement and evidence that the problem has been fixed. A price-based offer goes only to the group for whom price was genuinely the reason. And customers whose need has ended get left alone, or simply thanked.
Building those responses as automated, segment-specific sequences is exactly the work covered in re-engagement automation, with the segmentation itself handled by good email segmentation. The more durable win, though, is feeding what you learn back upstream: into onboarding, product, service and the design of any loyalty program, so fewer customers need winning back in the first place. Store the reasons you uncover as structured data alongside each customer record, a genuinely valuable part of a first-party data strategy, and review them as part of a regular funnel audit.
The short version
Rising churn usually triggers a blanket win-back discount, but churn is several different problems wearing one label, and a discount fixes at most one of them. Start by separating involuntary churn from failed payments, which industry estimates put at roughly 20 to 40% of subscription churn and which is the easiest to fix with reminders, smart retries, card updates and friendly payment messages. For non-subscription businesses, define churn from your own repurchase patterns. Sort voluntary churn into groups: never reached value, value or price mismatch, bad experience, and a need that has naturally ended, which is healthy churn you should not chase. Combine four sources of evidence: behaviour, exit surveys, unstructured feedback and short interviews with recent churners. Treat "too expensive" as the start of diagnosis, since it often means "not worth it to me", and check it against behaviour. Segment by channel, plan, tenure and timing before drawing conclusions. Then match each group to its own response, and feed what you learn back into onboarding, product and service.
Want to know why your customers really leave?
We help brands diagnose churn and build retention flows matched to each cause.
Explore Email Marketing Automation →Frequently asked questions
How do you find out why customers churn?
Combine several sources rather than relying on one. Start by separating involuntary churn, meaning customers lost to failed payments, from voluntary churn, because they have completely different causes and fixes. For voluntary churn, look at behavioural data first: what happened to usage or purchase frequency in the weeks before the customer left, and which features or products they never used. Then add stated reasons from exit surveys and cancellation flows, unstructured feedback from support tickets, reviews and complaints, and finally direct conversations with a small number of customers who have left. Each source has blind spots. Behaviour shows what happened but not why; surveys show what people are willing to say quickly; interviews reveal the real story but only for a few people. When several sources point to the same cause for the same group of customers, you can act on it with confidence. The goal is to sort churn into distinct groups, each with its own fix, rather than treating it as one number.
What is involuntary churn?
Involuntary churn is when a customer is lost without choosing to leave, most commonly because a payment failed: a card expired, was replaced after fraud, hit its limit, or was declined by the bank for a generic reason. The customer still wanted the product, but the business quietly stopped serving them. Industry estimates commonly put involuntary churn at somewhere between 20 and 40% of total churn for subscription businesses, and in some cases it is higher; The Seattle Times, for example, found that a majority of its churn came from failed payments. It is usually the easiest churn to reduce, because the fixes are operational rather than strategic: reminding customers before a card expires, retrying failed payments intelligently, using card-updater services, offering alternative payment methods and sending clear, friendly messages asking customers to update their details. Because it is so fixable, it should be separated out and addressed before any effort goes into diagnosing voluntary churn.
Are exit survey answers reliable?
They are useful, but they are a starting point for diagnosis rather than the answer. People completing a quick cancellation survey tend to pick the most convenient, least awkward option, which is why 'too expensive' dominates so many exit surveys. Often it really means 'not worth the money to me', which points to a value problem such as never getting set up properly or the product not fitting their needs, rather than a genuine inability to pay. The same answer can also cover switching to a cheaper competitor, a change in circumstances or simply using the product less. That is why stated reasons should always be checked against behaviour: a customer who says the price is too high but stopped logging in two months earlier probably left for another reason. Follow-up questions help, such as asking what they were comparing you to or what would have needed to be different, and short interviews with a handful of recent churners usually reveal far more than hundreds of tick-box answers.
Should you offer every churned customer a win-back discount?
No. A blanket discount only addresses one kind of churn, customers who genuinely left because of price, and it wastes money on everyone else. Customers lost to failed payments need a billing fix, not a discount. Customers who never got set up properly need better onboarding and guidance, and a discount simply brings them back to the same experience that failed them the first time. Customers who left because of poor service or product problems need those problems fixed and acknowledged. Customers whose need has naturally ended, because they moved, finished a project or outgrew the product, are usually not worth chasing at all. Discounting everyone also trains customers to cancel in order to receive an offer. Diagnose first, then send a tailored message to each group: an easy way to update payment details, a guided restart, an apology with evidence of what has changed, or a price-based offer for the smaller group for whom price was genuinely the reason.
