A customer success playbook is a defined, repeatable sequence of actions a CSM runs when a specific trigger fires — an onboarding kickoff, a usage drop, an expansion signal, an approaching renewal. The unit is the play: trigger → steps → owner → success criterion → exit. A playbook is the library of plays; a play is one entry. The point is to convert what your best CSM does on instinct into something every CSM executes the same way, and that a CS platform can fire automatically so a human only does the part that needs a human.
What a playbook is not: it is not a process doc, and it is not a help-center article. A process doc describes how onboarding works in general; a play says “when an account reaches day 14 with under 30% of seats activated, the CSM sends the activation-nudge sequence and books a 30-minute enablement call, and the play exits when seat activation crosses 60%.” A play has a trigger, a measurable exit, and an owner. If it doesn’t, it’s documentation, not a play.
The four play families
Most CS motions reduce to four trigger-based families. Build these first; everything else is a variation.
- Onboarding plays. Triggered by a closed-won deal or go-live date. Goal: get the customer to first value (TTV) before the honeymoon ends. Steps: kickoff, success-plan definition, enablement, milestone check-ins. Success criterion: the customer hits the defined activation milestone (seats live, first workflow shipped, first outcome realized).
- Risk plays. Triggered by a health-score band drop (green→yellow, yellow→red), a champion departure, a usage decline, a support escalation, or a sponsor going dark. Goal: diagnose and intervene before the signal becomes a non-renewal. Success criterion: the triggering signal reverses (usage recovers, a new champion is identified).
- Expansion plays. Triggered by a seat-utilization ceiling, a new-use-case signal, a power-user cluster forming, or a budget-cycle date. Goal: route a qualified expansion to sales or self-serve. Success criterion: a CSQL (customer success qualified lead) is created and accepted.
- Renewal plays. Triggered by a renewal date minus a lead time (commonly 90/120/180 days by ACV band). Goal: secure the renewal at or above current ARR. Success criterion: renewal closed, no surprise churn.
How to structure a play
Every play is the same six fields. Hold the line on all six — a play missing the exit criterion or the owner is the one that runs forever or never completes.
| Field | What it answers |
|---|---|
| Trigger | The exact condition that fires the play (a threshold, a date, an event) |
| Segment filter | Which accounts this applies to (ACV band, product, region) |
| Steps | The ordered actions, each tagged human or automated |
| Owner | The single role accountable (CSM, CS Ops, AE) |
| Success criterion | The measurable condition that exits the play as a win |
| Timeout / escalation | What happens if the criterion isn’t met in N days |
A worked risk play, fully specified:
Play: Usage decline — mid-market
Trigger: weekly active users down >25% vs trailing 4-week avg
Segment: ACV $25K-$100K
Steps: 1. [auto] Slack alert to owning CSM + health score recompute
2. [human] CSM reviews usage by feature, identifies the drop
3. [human] CSM sends re-engagement email within 48h
4. [human] book a working session if no reply in 5 days
Owner: CSM
Success: weekly active users recover to within 10% of prior avg
Timeout: 14 days → escalate to CS manager, flag for QBR agendaWhat to automate vs. keep human
The automation rule: automate the trigger detection and the low-judgment steps; keep the relationship steps human. A CS platform (Gainsight, ChurnZero, Vitally, Catalyst) watches the signals and fires the play — that part should never be manual, because manual signal-watching is where accounts slip. Inside the play, the split is by judgment:
- Automate: trigger detection, health recompute, alert routing, task creation, templated low-stakes emails (onboarding nudges, survey sends, NPS follow-ups), data hygiene.
- Keep human: the diagnosis (“why did usage drop?”), the renewal conversation, the expansion pitch, anything where a wrong automated message damages trust. A red-account play should create a task and brief the CSM, not auto-send a “we noticed you’re churning” email.
Sequence the build: instrument the triggers first (you can’t fire a play on a signal you don’t capture), then write the plays, then automate the obvious mechanical steps, then layer in the judgment-heavy ones only where the template genuinely holds.
Common pitfalls
- Plays without exit criteria. A play that never measurably completes clutters the CSM’s task list and trains the team to ignore plays. Guard: every play has a single measurable success criterion and a timeout that escalates — no open-ended “monitor the account.”
- Over-automation of relationship moments. Auto-sending a renewal or save email reads as a form letter at exactly the moment the customer needs to feel seen. Guard: classify each step human or automated explicitly; default relationship and money conversations to human.
- One playbook for all segments. A 12-seat SMB and a 4,000-seat enterprise account need different lead times, different owners, and different step depth. Guard: every play carries a segment filter; maintain per-segment lead times (renewal plays especially — 90 days for SMB, 180 for enterprise).
- Trigger sprawl. Fifty overlapping triggers fire constantly, the CSM drowns in tasks, and play quality collapses into noise. Guard: cap active plays per CSM and tune trigger thresholds against historical fire rate — if a trigger fires on half the book, it’s miscalibrated.
- No feedback loop. Plays ship once and ossify while reality drifts. Guard: review play win rates quarterly (success criterion met within timeout / total fires); retire plays under ~30% win rate and re-specify the trigger.
When the framework breaks down
Playbooks assume enough volume that repeatability pays off. A pure high-touch book of 8-12 strategic accounts at $1M+ ACV each doesn’t run plays — it runs bespoke account plans, and forcing those CSMs into a play library adds overhead with no payoff. The playbook model earns its keep in the tech-touch and mid-touch bands where one CSM owns 40-200 accounts and consistency, not customization, is the constraint.
Related
- Customer health score — the signal layer most risk and renewal plays trigger on
- Customer onboarding — the play family with the tightest TTV stakes
- Customer churn and expansion revenue — what risk and expansion plays exist to move
- Gainsight and ChurnZero — platforms with native playbook automation