A SaaS founder spent 3 months optimizing churn from exit-survey data. The real reason was in PostHog the whole time.
Had a churn problem at $4,200 MRR — 8 to 12 users a month quietly cancelling. Every cancellation genuinely hurt. Did the standard playbook: set up an exit survey in Tally on the cancellation page, one multiple choice question plus an optional text box. Responses came back consistent — almost always too expensive, missing features, or not what I needed. Treated those answers as ground truth and acted on them: kept pricing flat, roadmapped the features people mentioned. Felt data-driven. Felt like doing the right thing. Three months of this.
Acting on what users literally said on their way out was supposed to be the obvious data-driven move — keep price down, ship the missing features they mentioned, and churn should ease as the product matched what they said they wanted. The exit survey was the most direct signal we could get from leaving customers; trusting it felt rational.
Three months of roadmapping features and pricing restraint, and churn didn’t move at all. Same 8-12 cancellations a month. Sat staring at the same numbers thinking I was missing something obvious. Turns out I was. By the time someone fills out a cancellation survey they’ve already emotionally checked out — they pick the easiest box and want to get it over with. The real story for why they left happened weeks earlier, and they’ve already moved on by the time they answer. The survey was telling me what people SAID, not what they DID. Roadmap was being steered by a signal that was systematically wrong, and there was no way to know that just by looking at the survey CSV.
Got frustrated enough to ditch the survey and just message cancelled users directly. Not a sequence, not a template — opened Notion where we tracked cancellations, found emails, wrote “hey, saw you cancelled last month, genuinely trying to understand what happened, would you be willing to tell me?”. First reply came in 20 minutes. They didn’t mention price once. They’d gotten confused during onboarding, couldn’t figure out how to do the one thing they signed up to do, tried twice, gave up. Six weeks gone before I messaged them. Pulled PostHog session recordings — watched them hit the exact UI element that confused them, click around 90 seconds, close the tab. Did the same exercise for six more churned users over the next two weeks: PostHog recordings + Gmail support history + own database event logs, manually cross-referenced. Patterns weren’t subtle: 4 of 7 never got past the same onboarding step, 2 had unanswered support emails we’d closed without follow-up, 1 was a power user whose key feature we’d broken with an update we didn’t catch. None of this was hidden — PostHog had the recordings, Gmail had the threads, our own database had the event logs. We just never looked at it together. Fixed the onboarding issue in about a week. Single UI change. Churn dropped noticeably the following month.