If you've lost more than 100 accounts, you have enough data to find your churn signature. Most teams don't look for it systematically, and that's a strategic error.
The churn autopsy is simple in concept: take your last 100-200 churned accounts, build a profile of each, and look for shared characteristics. Not the stated exit reason — that's the symptom. Look for the structural characteristics of accounts that churned: industry, company size, acquisition channel, time to first value, onboarding completion rate, usage patterns at 30/60/90 days, CSM engagement frequency.
What teams consistently find when they do this analysis:
Certain acquisition channels produce churny customers. Not because the channel is bad, but because the messaging in that channel attracts customers with a use case that your product doesn't serve well. The product isn't wrong. The message is.
Onboarding completion is one of the strongest predictors of first-year retention. Accounts that never completed the defined onboarding milestones churn at 2-4x the rate of accounts that did. The implication: onboarding completion should be a KPI with the same weight as NRR.
Accounts below a certain usage threshold in month 3 are essentially certain to churn before month 12. Knowing this threshold lets you intervene in month 3, not month 11.
Accounts in certain industries or company sizes churn at structurally higher rates than others. This is the ICP signal. If your product serves construction companies with 20-50 employees at dramatically lower churn than e-commerce companies with 200+ employees, your ICP should reflect that.
The churn autopsy isn't morbid. It's the most direct path to improving ICP, improving onboarding, and improving the accounts you serve.
Run it quarterly. Update your ICP quarterly. Iterate.