The client is gone; the lessons don't have to be. Run a blameless post-mortem on the
relationship: what actually drove the departure, when it became visible in the data, and
what to instrument so the pattern is caught earlier next time.
1. Confirm the departed client (look it up) and establish the timeline: relationship start,
termination notice date, final date. Review the full history, working backward from the
end; state which portion the analysis fully covers if searches cap.
2. Reconstruct the cause chain from evidence:
- The proximate trigger, if visible (a final incident, a renewal dispute, a stakeholder
change).
- The underlying conditions: chronic recurring issues without root-cause fixes, SLA
erosion, declining sentiment arc, dropped commitments, relationship concentration in
one departing contact.
- External factors the tickets hint at (competitor mention, internal-IT hire, business
downturn) — labeled hypothesis where inferred.
Present as a dated narrative: what happened, in what order, one representative example per
pattern.
3. Mark the point of no return as best the data shows — the moment after which no thread
suggests the client was still persuadable — and what distinguished before from after.
4. Extract lessons, each phrased as a process change, not a regret ("recurring issues at
this frequency need a forced RCA", "a single point of contact going quiet for 60 days
should trigger AM outreach").
5. Build the missed-signals list: each early warning visible in the data with its date, how
many months before termination it appeared, and the detection rule that would catch it
(frequency threshold, sentiment drop, silence window). Where a rule maps onto Client Risk
Scan or Sentiment Decline Watch signals, say so — the autopsy should feed the scanners.
6. Close with two or three recommended changes — monitoring rules, process fixes,
account-coverage changes — each traceable to a lesson.
Guardrails: blameless and internal — the autopsy names process failures, never people; if an
individual's handling contributed, describe the missing guardrail or escalation path, and
performance evaluation goes to the manager. Evidence discipline: every cause claim cites
thread evidence (one representative example, plain language, no ticket IDs); external-factor
speculation is clearly labeled hypothesis. Do not rewrite history to a tidy single cause —
churn is usually plural; report the chain as the evidence shows it, including genuine
unknowns. The former client is still owed professionalism: no disparaging commentary, and
nothing here is ever shared externally or reused in marketing. Result-cap honesty on every
historical claim ("based on the final 18 months of history").
Magic Library