Analyze churn for [PRODUCT AND PERIOD] using the data and notes below: [CHURN DATA AND DEFINITIONS]. Customer feedback, support themes, and cancellation reasons are [CUSTOMER EVIDENCE]. Our customer segments, lifecycle stages, and material product changes are [SEGMENTS AND CONTEXT]. The business question I need answered is [DECISION TO MAKE]. The available retention options and constraints are [OPTIONS AND CONSTRAINTS]. First, state the working definitions you are using for customer, logo churn, revenue churn, voluntary churn, involuntary churn, cohort, and observation window. If the supplied data cannot support one of these measures, say so instead of estimating it. Produce a decision memo with these sections:
1. Data-quality and comparability checks: denominator changes, missing cancellation reasons, duplicate accounts, annual-plan timing, refunds, involuntary payment failures, and cohort maturity. 2. Churn picture: totals and rates only where calculable; breakouts by segment, acquisition source, tenure, plan, and cohort when data permits. 3. Evidence-backed drivers: distinguish observed patterns, customer-stated reasons, and hypotheses. Cite the specific field, count, or note behind each conclusion. 4. Priority actions: no more than five experiments or operational fixes. For each, name the target segment, intervention, owner type, success metric, guardrail, and earliest point to measure it. 5. Customer-response guidance: empathetic language and clear next steps for cancellation or save conversations, without pressuring customers or promising unavailable outcomes. Do not infer causation from a correlation, treat a small sample as representative, or call feature requests the root cause without supporting evidence. Do not invent benchmarks or data. Self-check: list every conclusion that is a hypothesis, and flag any metric whose denominator or date range is unclear. Ask up to 3 clarifying questions only if a required input is missing.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [PRODUCT AND PERIOD] | Enter the product, market, and date range covered by the analysis. | Pulseboard, a B2B reporting SaaS, April through August 2026 |
| [CHURN DATA AND DEFINITIONS] | Paste a table, CSV excerpt, dashboard export, or metric definitions with dates and denominators. | Monthly account export with account ID, plan, MRR, signup month, cancellation date, payment-failure flag, source, and cancellation reason; churn means paid accounts ending service in a calendar month |
| [CUSTOMER EVIDENCE] | Paste anonymized cancellation reasons, survey responses, support tags, and account notes. | 34 cancellation forms and 81 support tickets; recurring notes mention difficult CSV imports and lack of scheduled email reports |
| [SEGMENTS AND CONTEXT] | Describe meaningful customer groups, lifecycle stages, pricing changes, launches, outages, or policy changes. | Solo consultants, 2-10 person agencies, and 11-50 person teams; new pricing began June 1; import redesign shipped July 15 |
| [DECISION TO MAKE] | State the decision this analysis should inform. | Choose the two retention initiatives for the next quarter |
| [OPTIONS AND CONSTRAINTS] | List viable interventions, resources, product limits, and customer-policy constraints. | One product squad, two support specialists, no discounts longer than three months, and no outreach to accounts that requested no contact |
How to use
- Export enough raw data to identify the date range, account denominator, segment, and cancellation status.
- Remove names, emails, and sensitive free text before pasting data or support notes.
- Check whether the analysis separates payment failures, voluntary churn, and immature cohorts.
- Follow up with: “Turn the top two actions into an experiment brief with eligibility rules, event tracking, and a stop condition.”
Variations
Cancellation reasons
Use when you have mostly survey answers or support notes rather than a full data export.
Analyze these anonymized cancellation responses for [PRODUCT]: [RESPONSES]. The response period and total cancellations are [PERIOD AND DENOMINATOR]. Build a coded theme table with theme definition, count, representative paraphrase, confidence, and overlapping themes. Separate stated reasons from inferred causes, identify ambiguous or unusable responses, and recommend up to three follow-up questions for future cancellation surveys. Do not make rate claims if the denominator is incomplete. Self-check that the themes are mutually understandable and that no quote or customer fact was invented.
Save playbook
Use when creating a fair response process for customers considering cancellation.
Create a cancellation and retention-response playbook for [PRODUCT] and [CUSTOMER SEGMENT]. Known churn drivers are [EVIDENCE]. Available remedies are [APPROVED REMEDIES], and prohibited practices are [LIMITS]. Provide a decision tree for support: acknowledge, diagnose, offer eligible remedy, confirm cancellation if wanted, and record structured reason codes. Include example language for each stage. Do not pressure customers, hide cancellation steps, make unapproved promises, or use a discount without eligibility criteria. Check that each remedy maps to an evidence-backed problem.
Cohort review
Use when comparing retention across signup cohorts.
Review retention by cohort for [PRODUCT] using [COHORT TABLE]. Define the cohort entry event as [ENTRY EVENT] and the retained event as [RETAINED EVENT]. Explain which cohorts are mature enough to compare, calculate only metrics supported by the table, and identify changes in acquisition mix, pricing, or onboarding that could confound comparisons: [CONTEXT]. Return a cohort table, three careful observations, and three hypotheses to test. Do not attribute changes to a launch without a comparison. Flag missing denominators or inconsistent periods.
Tips
- Use an account-level export for logo churn and a revenue-level export for MRR churn; combining them hides different problems.
- Tag payment failures separately from voluntary cancellations because dunning, card updates, and product dissatisfaction need different interventions.
- Read cancellation text alongside tenure and plan data; the same complaint from new trials and long-term customers often means different things.
- Measure a retention experiment against an eligible comparison group, and include a guardrail such as refund rate, support load, or discount cost.
FAQ
Can AI identify the cause of churn from a spreadsheet?
It can identify patterns and generate hypotheses, but a spreadsheet rarely proves cause. Use customer interviews, controlled tests, and cohort comparisons to validate a proposed driver.
What data is most useful for churn analysis?
At minimum, include account ID, start date, cancellation or inactivity date, plan, revenue, customer segment, and cancellation reason. Product-usage events and support history improve interpretation.
Should we try to save every customer who cancels?
No. Build clear eligibility rules and respect a customer's decision to leave. Some churn is a poor fit or an outcome that cannot be fixed responsibly with a retention offer.