Reference · Anytime

Customer Success Playbook

Situational reference for every recurring customer scenario. Pick a situation on the left — get triggers, stakeholders, do's & don'ts, conversation starters, orchestration, and signals to watch.

🔥
Phase · Nurture (emergency re-entry)

Unhappy Customer — Escalation Recovery

Trigger
Customer has raised escalation, sent formal complaint, or expressed dissatisfaction to leadership.
Objective
Make the customer feel heard, take full ownership, and demonstrate recovery through evidence.
Stakeholders to align
  • Ops / Ground team
  • VP / SVP / Business Head
  • CXO (if senior escalation)
  • CS + Engineering + Product
Tools you'll use
  • Escalation Dashboard
  • RCA Template
  • 30-60-90 Recovery Plan
  • Action Tracker
  • Model/Prompt Change Log (for AI-platform accounts)
Do
  • Respond within 4 hours of any escalation signal.
  • Start with listening. No slides on the first recovery call.
  • Escalate internally within 2 hours — Code Red treatment.
  • Deliver written RCA within 48 hours with root cause, fix, and prevention plan.
  • Assign a senior named CS contact with direct number.
  • For AI-platform incidents (hallucination, regression, prompt-injection, runaway cost): include model/version, prompt diff, tool-call trace, eval-set delta, and a guardrail change in the RCA — not just an infra timeline.
Don't
  • Do not be defensive or explain before acknowledging impact.
  • Do not use generic email templates.
  • Do not discuss commercial matters until operational issues are resolved.
  • Do not let more than 24 hours pass without a visible update.
  • Do not silently roll back a model or prompt without written notice — AI customers treat undeclared model changes as a trust breach.

Conversation starters

"We want to first acknowledge the frustration — this is not the experience you should have, and we own that completely."
"Rather than explaining what happened, let us focus on what we are doing next. Here is our RCA and 30-60-90 plan."
"We have assigned [name] as your dedicated contact from today — every update comes from them directly."
"On the model regression you flagged: here is the version diff, the eval-set delta, the guardrail we have added, and the rollback path. We will keep both versions live until you confirm stability."

Orchestration timeline

  1. 1.Hour 0–2: Acknowledge — do not solve yet, acknowledge impact
  2. 2.Hour 2–4: Internal Code Red — engineering, product, CS leadership looped in
  3. 3.Hour 24–48: Written RCA with named owner and fix timeline
  4. 4.Day 7: Recovery plan review — joint ownership confirmed
  5. 5.Day 30: Milestone check — is operational trust restored?
Positive signals
  • Customer raises new ticket through normal channel
  • VP re-engages in scheduled connects
Anti-signals
  • Escalation re-opened after appearing resolved
  • Customer copies senior leadership on routine comms
Target outcome

Customer feels heard. Operational trust restored before any commercial conversation.

Cadence: Daily for 7 days → weekly for 30 days → monthly thereafter