Churn in a SaaS company almost never looks like a decision. It looks like a slow drop in contact, a QBR that keeps getting rescheduled, a champion who stops replying, and then a non-renewal notice that arrives as a surprise to everyone except the customer.
The uncomfortable part is that the pattern is usually visible in hindsight and rarely looked for in advance. What follows is a structured way to look, using only information your company already has: tickets, email, call notes and the CRM. It takes about an hour, and it works best when you run it against the accounts you’ve already lost rather than the ones you still have.
Question one: when did contact actually change?
Take three accounts that churned in the past two years. Go into the email, ticket and meeting history and find the month where the rhythm changed. Fewer inbound questions. Slower replies from their side. QBRs that slid from scheduled to ad hoc. Product usage that thinned out, if you track it.
In most cases that month sits well before the cancellation notice, further back than anyone expects. The gap between the two is your warning window. It’s also the answer to whether early detection would have been useful, and it usually is.
Question two: who were they actually talking to?
Look at who the customer copied and who they wrote to directly. In companies where retention is quietly slipping, there’s a common shape. Early on the customer writes to the founder or their account executive. Later they write to their CSM. Later still they open tickets in the general support queue.
The customer didn’t decide to downgrade the relationship. The company handed them down the chain, usually for good operational reasons, and nobody told them it was happening or why.
Customers rarely leave over one missing feature. They leave over the feeling that they've been quietly reassigned.
Question three: what happened at the last handoff?
Every transition is a moment where the customer reassesses. Sales to implementation. Implementation to customer success. One CSM to the next when books get rebalanced. One renewal to the next.
For each churned account, find the last handoff before the contact pattern changed. Ask what the customer was told about it, by whom, and when. Usually the honest answer is that they were told nothing and worked it out from the change in who replied to their tickets.
Question four: how did the first ninety days go?
Go back to the start of the relationship. What did the customer get in their first ninety days, and how quickly did they see something that felt like value? In SaaS terms, what was their time to value?
Openings set tolerance. A customer whose first ninety days were organized, communicative and visibly useful will forgive a rough quarter two years later. A customer whose onboarding was slow and improvised has less patience banked, and they spend it faster. This is why onboarding work produces retention results that look out of proportion to the effort involved.
Question five: what did they ask for more than once?
Search the tickets and the email for repetition. A question asked twice is a gap in your explanation. A question asked three times is a customer teaching themselves that asking your company doesn’t resolve things.
Repeat tickets are the single most underused retention signal in B2B software, because each one on its own looks trivial. The pattern isn’t trivial, and it’s easy to find once you decide to look for it.
Question six: who inside the company wasn’t surprised?
Ask the team. In most cases at least one person, usually a support agent or a CSM close to the day-to-day work, will say some version of “I thought that might happen.”
That person had the signal and no route to raise it. Not because your company discourages it, but because nothing existed that made raising it a normal act. Building that route is often the cheapest retention improvement available, and it costs nothing but a standing item on an existing agenda, like the weekly customer success meeting.
Run this on the losses first. There's a strong instinct to apply this to current accounts and skip the post-mortem. Resist it. The churned accounts are where the pattern is unambiguous, and the pattern is what makes the exercise useful on the customers you still have.
What to do with the answers
You’re not looking for six problems. You’re looking for the two or three exit points that show up in more than one story. Churn clusters, and the clustering is what makes it addressable.
From there the work is small and specific. Document the handoff that keeps appearing. Rebuild the first ninety days. Put a short review on the calendar with a named owner and the two or three signals you now know to watch. None of that requires a new customer success platform, and most of it can be built and running inside a month.
The goal isn’t to prevent every cancellation. Some customers should leave, and a company with no churn at all is usually underpricing. The goal is to stop being surprised, because a cancellation you saw coming is a conversation, and one you didn’t is a non-renewal notice.