Most teams do a cancellation survey, collect one sentence, and call the result a root cause. That is how the same leak survives quarter after quarter. “Too expensive,” “not using it,” and “priorities changed” may be accurate descriptions, but they do not yet tell you which promise broke, when confidence fell, what the customer tried, or which part of the system could have changed the outcome.
A churn autopsy is a disciplined reconstruction of a lost account. It combines the customer’s voice with the timeline of usage, onboarding, support, billing, product changes, and commercial context. The goal is not to make every loss preventable. The goal is to turn repeated losses into hypotheses that can be tested by an owner.
Choose a stage to see what the team should preserve, reconstruct, classify, and change. The workflow is centered deliberately: the autopsy is the operating instrument, not a sidebar note.
Freeze the customer’s segment, plan, tenure, usage, support history, invoices, expansion or contraction, and the exact cancellation date. An exit reason without the surrounding timeline is a story fragment, not an autopsy.
cohort → timeline → evidenceThe seven categories below are a practical taxonomy, not a universal ranking. Start with the pattern in your own cohort, then use the questions to move from a stated reason to corroborated evidence.
| Root cause | What the team may observe | Exact questions to ask |
|---|---|---|
| 1. Value never became concrete | Low usage, vague success language, weak executive sponsorship, or a renewal conversation that begins with “we never really got started.” | “What outcome did you expect to be different by this point?” “Which result would have made renewal an easy decision?” “Where did the expected value fail to become visible?” |
| 2. Activation stalled | The account signed but never completed a critical setup step, invited the right users, connected a required workflow, or reached a meaningful first win. | “What was the first step that felt harder than expected?” “Who was responsible for completing it?” “What did you do after the setup stalled?” |
| 3. Product or service friction | Repeated support contacts, workarounds, defects, slow response, missing integrations, or a growing gap between the promised and usable experience. | “What were you trying to do when the friction appeared?” “How often did it happen?” “What workaround did you adopt?” “What would have made the problem tolerable?” |
| 4. Misfit or overpromised fit | The customer bought for a use case the product was not designed to serve, or the sales promise outran delivery capacity and operating reality. | “What did you believe you were buying?” “Which part of that expectation was set before the sale?” “When did you realize the fit was different?” |
| 5. Commercial pressure | Budget cuts, procurement changes, price–value tension, payment issues, consolidation, or a new internal approval threshold. | “What changed in the buying environment?” “Was the decision about absolute price, expected value, timing, or internal priority?” “What alternative did you compare us with?” |
| 6. Champion or context loss | A sponsor left, ownership changed, a team reorganized, the project was paused, or the original business case lost its internal advocate. | “Who owned the result when you started?” “What changed in the organization?” “What would a new owner need to see to restart the work?” |
| 7. Trust or relationship erosion | Communication drops, unresolved commitments accumulate, the customer stops volunteering information, or the renewal becomes a surprise rather than a review. | “When did confidence begin to decline?” “Which commitment felt unfinished?” “What did you hesitate to tell us while you were still a customer?” |
Do not force one cause. A customer can leave because activation stalled, then describe the loss as price. Another can experience a product defect but stay because a strong champion absorbs the work. The autopsy should allow competing explanations until the timeline, behavior, and conversation point in the same direction.
Separate controllability from responsibility. A market shutdown may be real and uncontrollable, but it can still reveal that the business has no contraction plan. A champion departure may not be preventable, but it can reveal that value was never distributed beyond one person. The purpose of classification is not blame; it is to decide what kind of response is rational.
A useful exit conversation lowers the pressure to defend the decision. It asks for chronology, comparison, attempted workarounds, and the moment the account moved from recoverable to lost.
Start with permission and curiosity: “We are reviewing our own process, not trying to argue you back into the product. Would you walk me through the decision from the first moment you considered leaving?” Do not insert a counteroffer after every answer. A customer who feels managed will give you a cleaner-sounding, less useful story.
Follow the verbs. When the customer says “we stopped using it,” ask what they tried, who tried it, what happened next, and what they expected to happen. Concrete actions reveal where the system failed more reliably than adjectives such as disappointing, expensive, or complicated.
Use onboarding dates, support tickets, usage changes, invoices, renewal notices, and product releases as memory aids. Ask: “What was happening in the business during that period?” “Which event changed the project?” “What did your team do immediately afterward?” Timeline reconstruction helps distinguish an early fit problem from a later context shock.
Use evidence carefully. Data can show that usage fell; it cannot by itself prove why. A low-login account may have achieved the intended outcome, while a high-login account may be trapped in repetitive work. Combine behavioral signals with the customer’s account of meaning.
Ask what would have changed the decision, but do not treat the answer as a guaranteed intervention. “If we had fixed one thing three months earlier, what would you have wanted it to be?” “Would that have changed the renewal decision or only improved the experience?” The second question separates a meaningful lever from a polite wish list.
Ask what was already tried. “Did you raise this with anyone?” “What response did you receive?” “What did you do when the first solution did not work?” A root cause may be a known problem, but the deeper leak may be delayed acknowledgement, unclear ownership, or a recovery promise that was never closed.
Reflect the story back: “I heard three moments: the implementation stalled, your internal owner changed, and by renewal the price felt difficult to justify because the result was not visible. Did I miss anything?” Ask which part is most important and what you may quote internally without identifying the customer.
Record confidence. Mark each cause as stated, observed, corroborated, or unresolved. That small discipline keeps a vivid interview from overpowering quieter evidence across the cohort.
The system earns its keep when it changes what the team does before the next comparable customer reaches the same moment.
Start with a deliberately sampled set rather than only the loudest account. Include recent losses, high-value losses, early-life churn, involuntary churn where relevant, and a comparison group of retained customers from the same segment or cohort. The exact number depends on volume; the principle is to avoid mistaking one dramatic story for the whole base.
Ask what happened before the decision, what the customer tried, what alternative they chose, and what evidence exists in usage, support, billing, and onboarding records. Keep the first reason as a hypothesis. Raise confidence only when different evidence sources converge.
A controllable cause is a lever the company can plausibly influence, such as unclear onboarding, an unresolved defect, poor communication, or a packaging mismatch. An uncontrollable event may still reveal a preparedness gap, but it should not be counted as proof that a retention intervention would have saved the account.
Score each pattern on frequency, revenue exposure, controllability, confidence, and time to learn. A frequent low-revenue issue may deserve a product fix; a rare high-revenue issue may deserve an account-specific recovery plan; a low-confidence pattern may deserve more interviews before engineering work begins.
Record the cause hypothesis, supporting evidence, counter-evidence, affected segment, owner, intervention, leading indicator, lagging outcome, review date, and rollback condition. If a decision has no owner or review date, it is an observation, not an operating action.
Define the comparable population and the time window before launching the change. Watch the leading behavior first, then cohort retention, expansion or contraction, support load, and customer-reported value. A lower churn rate can be misleading if the mix of customers, pricing, or acquisition sources changed at the same time.
Preserve the cohort, reconstruct the timeline, ask the exact question, classify the hypothesis, and assign one correction to one owner.
Run The Autopsy Workflow