Sales prompt of the day

Churn post-mortem: what was knowable, and when

An account left and the account team has a story. Reconstruct what the record actually shows and find the point where it was still saveable.

You are a retention analyst running a blameless churn post-mortem. I will paste the account history. Be direct — a comfortable post-mortem teaches nothing.

Produce:

A) THE TIMELINE: the account's history as evidenced in my material, marking each point where a signal appeared, what was done about it, and how long the gap was between the signal and any response. Separate what the record shows from what the team believes.

B) THE DECISION POINTS: the two or three moments where a different action plausibly changes the outcome, what the available information supported doing at that moment, and the honest entry "not visible at the time" for anything only obvious in hindsight.

C) THE PATTERN CHECK: which of these signals would be detectable in other accounts today from data we already collect, the specific check that would surface them, and the stated churn reason compared against what the timeline actually supports — these often disagree.

Inputs: [ACCOUNT HISTORY: USAGE OVER TIME, SUPPORT CONTACTS, STAKEHOLDER CHANGES, QBR NOTES, INVOICES OR CONTRACT CHANGES] · [STATED CHURN REASON] · [WHAT THE TEAM BELIEVES HAPPENED] · [WHAT WE MONITOR ACROSS OTHER ACCOUNTS TODAY]

Rules: Do not invent events, sentiment, or motives the record does not support — write "not in the record" and never assign blame to a named individual. Use roles and internal account IDs, never customer names or personal data. Keep confidential CRM, deal, and customer data out of consumer AI tools and follow your employer's AI-use policy. This reads the record; the account team's context still matters and should be added, not overwritten. Verify the timeline against the support, billing, and usage record before this post-mortem circulates.

Why it works — Churn post-mortems converge on the exit-interview reason, which is usually the most polite answer rather than the true one, and the same failure repeats. Separating the evidenced timeline from the team's narrative surfaces the disagreement, and demanding the honest "not visible at the time" label prevents hindsight from producing a lesson nobody could actually have acted on.