Learn · customer-understanding

By Fieldwork Research Team · Published 2026-07-10

Why are customers churning?

Customers churn when the gap between what they expected and what they experienced becomes too wide to ignore. That gap is almost never visible in your analytics. Your dashboard shows you when churn happens, at what rate, and from which cohort. It does not show you what the customer was thinking when they decided to leave, what they tried before giving up, or what would have changed their mind. That information only exists in a conversation.

Why churn is harder to diagnose than it looks

The instinct when churn increases is to look at the data. Usage drops, engagement falls, a feature goes untouched. Analytics tell a clean story: customers who churned used the product less in the 30 days before cancelling. That observation is accurate and almost entirely useless for preventing the next wave of churn.

The problem is that reduced usage is a symptom, not a cause. Customers use a product less because something is wrong, but the analytics cannot tell you what that something is. It could be that the product no longer fits their workflow. It could be that a competitor solved a problem yours didn't. It could be that the person who championed the product internally left the company. It could be that the onboarding never worked and they never reached the value they were promised.

Each of these causes requires a completely different response. Building a re-engagement email sequence is the right answer to one of them and the wrong answer to all the others. Without knowing which cause is operating, every churn reduction initiative is a guess.

What analytics can and cannot tell you

Analytics are good at telling you what happened. They are structurally incapable of telling you why.

What you can learn from analytics:

When in the customer lifecycle churn tends to happen
Which features churned customers used or didn't use
Which segments have higher churn rates
What the last action was before cancellation

What you cannot learn from analytics:

What the customer was trying to accomplish that the product failed to deliver
What they told their colleagues when they decided to cancel
What a competitor offered that made switching feel worth the effort
What you could have done differently that would have changed the outcome

The second list is the list that determines what you actually do about churn. Getting it requires talking to customers. For a practical playbook, see how to interview customers at scale.

The three conversations that explain most churn

Churn rarely has one cause. But in most SaaS products and service businesses, the same three underlying patterns appear repeatedly when you talk to customers who left.

The value was never fully realised. The customer signed up with a specific expectation of what the product would do for them. Something in the onboarding, the setup, or the early experience prevented them from reaching that value. By the time the renewal came around, they were paying for a product they'd never fully used. They didn't churn because the product was bad. They churned because they never experienced it being good.

This pattern is almost invisible in analytics because usage data doesn't distinguish between "uses the product confidently" and "opens the product, gets confused, closes it again." Both look the same in a session log.

The problem the product solves stopped being a priority. Markets change, companies reorganise, budgets shift. A product that was essential six months ago becomes a line item to cut when the business pressure changes. The customer didn't churn because of anything you did or didn't do. They churned because their situation changed.

This pattern matters because it's often misdiagnosed as a product quality problem. If you don't know this is the cause, you'll spend resources improving a product that was already good enough.

A competing solution became more attractive. Not necessarily cheaper. Faster, simpler, better integrated with the rest of their stack, or simply better at the specific thing that matters most to them. Customers rarely switch because a competitor is categorically better. They switch because the gap between what they have and what they could have became worth the switching cost.

This pattern is the most actionable. It tells you exactly what to compete on.

How to find out which pattern is operating

Exit surveys capture some of this but not enough. A customer who just cancelled is not in a reflective mood. A dropdown asking "why did you cancel?" produces responses that are socially acceptable rather than honest. "Too expensive" is the most common exit survey response across every industry. It is also the least accurate: price is rarely the real reason and almost never the only one.

The conversations that produce real churn insight are structured qualitative interviews with customers who have recently left, conducted by someone who has no stake in the answer. The interview needs to:

Start with their experience before the product, not with the product itself. What were they trying to accomplish? How were they doing it before?

Reconstruct the timeline of the relationship. When did they first feel something wasn't working? What did they try? Who else was involved in the decision to leave?

Ask about the moment of decision, not the outcome. What was happening in the week they decided to cancel? What was the specific trigger?

Ask about alternatives. What are they doing now? What does that solution do differently?

The answers to these questions produce a picture of churn that is specific enough to act on. Not "customers are churning because of price" but "three of the eight customers we spoke to never successfully integrated the product with their existing workflow because the setup required technical knowledge they didn't have, and by the time they realised that, they'd already paid for two months."

That is a solvable problem. A price objection is not.

What this looks like in practice

A subscription software company is seeing monthly churn increase from 2.1% to 3.4% over a quarter. The team runs exit surveys and the most common response is "too expensive." They consider a pricing change.

Before committing, a researcher runs 10 interviews with customers who cancelled in the past 60 days. The brief is specific: understand what they were trying to accomplish, when things started feeling wrong, and what the decision to leave actually looked like.

Seven of the ten interviews reveal the same pattern. Customers were using the product for a specific workflow that changed when their company adopted a new internal tool. The product and the new tool didn't integrate. Customers spent two months trying to make it work, then gave up.

None of them mentioned price in the interview. When asked directly, three said price was a factor but only because they couldn't justify paying for something that no longer fit their workflow.

The fix is an integration, not a price change. The exit survey would have led to the wrong decision. The interviews led to the right one.

Frequently asked questions

How many churned customers do I need to interview to understand why churn is happening?

For a well-scoped churn study, 8 to 12 interviews with recently churned customers will typically surface the primary patterns. The goal is saturation. The point at which new interviews stop producing new explanations. For most products, that happens within 10 conversations if the recruitment criteria are tight (recently churned, within a specific customer segment, having used the product beyond the first week).

Should I interview customers who are about to churn or customers who already have?

Both, ideally, but for different reasons. Recently churned customers can tell you what actually happened. The complete story with the outcome known. At-risk customers (identified by usage signals) can tell you what is currently happening, which allows intervention before the decision is made. Churned customer interviews diagnose the problem. At-risk customer interviews create the opportunity to solve it in real time.

Why do exit surveys consistently underreport the real reasons for churn?

Exit surveys ask customers to self-report their reason for leaving at a moment when they are disengaged and often mildly frustrated. The options provided are usually too broad, too polite, or too product-focused to capture the real cause. "Too expensive" is easy to select and hard to dispute. "I never figured out how to make this work with my existing tools and eventually stopped trying" requires a text box and emotional honesty that most exit survey moments don't invite.

How do I get recently churned customers to agree to an interview?

A direct, personal request sent within two weeks of cancellation converts better than a generic survey link. The message should acknowledge that they left, express genuine curiosity about their experience without defensiveness, and make the time commitment clear and small (15 to 20 minutes). An incentive of $50 to $75 for a 20-minute conversation is appropriate and signals that you value their time.

What do I do with churn interview findings once I have them?

Map each finding to a specific point in the customer journey where an intervention could have changed the outcome. "Never reached value in onboarding" maps to an onboarding improvement. "Workflow integration broke" maps to a technical integration or a better setup guide. "Problem deprioritised internally" maps to multi-stakeholder engagement earlier in the relationship. Each finding should produce a specific, ownable action, not a vague initiative to "improve retention."

Can AI-conducted interviews work for churn research?

Yes, particularly for structured churn studies where the research question is specific and the interview topics are defined in advance. AI-conducted interviews are consistent across all participants, remove the awkwardness of a company employee interviewing a customer who just cancelled, and can run at a scale that makes it feasible to speak to every churned customer above a certain contract value rather than a sample. The analysis still requires human interpretation, but the data collection is both more consistent and more scalable than moderator-led sessions.

Related on Fieldwork

Last updated: 2026-07-10