You can see the churn.You can’t see why.

Those are Customer Success platforms. They watch your accounts continuously, score health, and route alerts to your CS team. They cost $15K–$80K/year, require deployment work, and assume you have a CS team operating them. The Autopsy Kit is the diagnostic layer that lives upstream of all of them. It tells you WHY customers are churning, organized by actionable reason category, with matching win-back sequences. Most teams need the diagnostic FIRST — knowing the disease before buying the monitoring system. The kit is complementary to a CS platform if you already have one, and a replacement for the missing diagnostic if you don't.

Helpful but not required. The cohort retention analyzer accepts a simple CSV — date, customer ID, MRR/usage signal, status. If you don't have an analytics platform, the kit ships with a 'minimum viable retention dataset' template that pulls the same signal from Stripe, Chargebee, or any SQL query you can run against your billing database. If you do have Mixpanel or Amplitude, the kit's CSV format mirrors their cohort export schema directly.

Both. The script ships in two variants: SaaS (annual or monthly subscription, MRR-based) and services (retainer, project-based, or recurring services). The questions are structurally identical because the diagnostic logic is the same — what changed, what alternatives were considered, what would have needed to be true to stay — but the vocabulary and the example follow-ups are calibrated for each business model. Agencies, consultants, and recurring service businesses use the services variant.

The classifier uses confidence scoring across all five reason categories rather than forcing a single label. A typical output looks like: "Primary: competitor (62% confidence). Secondary: price (28%). Tertiary: fit (10%)." This matters because most churn is multi-causal — a customer leaves for a competitor because that competitor is cheaper because the original product never demonstrated enough value to justify its price. The classifier surfaces the layered reasoning so you fix the right thing at the right layer.

Real working tool. It's a Claude artifact (HTML/React) that runs inside any Claude.ai conversation. You upload your CSV (date, customer ID, status, MRR), and it produces cohort heatmaps, retention curves by acquisition month, survival analysis by churn-reason category (once you've classified), and a 'leakiest cohort' diagnostic. Open-source-style — runs entirely in your browser, no data leaves your Claude session. The full kit includes the artifact code plus a deployment guide.

Some, not all — and the kit is honest about which. "Life event" churn (the champion left, the company got acquired, the team was laid off) typically recovers at 15-30% with a re-discovery sequence aimed at the new buyer. "Neglect" churn (onboarding failure, integration broke, key user left) typically recovers at 20-40% with a re-onboarding offer. "Price" and "fit" churn recover at 5-15% — meaningful at volume but not magic. "Competitor" churn recovers at 3-10% — hardest. The win-back library is structured to give you realistic expectations per category rather than promising universal recovery rates that don't exist.

30-day no-questions refund. Run one exit interview using the kit, classify the result, and pull up the matching win-back sequence. If it doesn't give you a clearer picture of why that specific customer left than your existing process would have, email RedHub AI support and we refund. Refunds across the catalog are countable on one hand to date — the diagnostic value is self-evident the first run.