Churn-Risk Save Playbook 2026

Set the trigger on a relative drop, not an absolute threshold. A score of 45 might be normal for a small team while a fall from 80 to 55 at a big account is a fire alarm.

• Score every account 0–100 on how much, how often and how deeply they use the product. Accoil does this on top of the events you already send to Segment, PostHog, Amplitude or Mixpanel — no new instrumentation.
• Trigger on a drop of 20+ points over 14 days. Tune the window to your usage cadence: weekly-use products may need 28 days.
• The trigger doesn't fire empty-handed. It carries the payload the rest of the play runs on: the current engagement score, the 14-day trend, the percentage of seats still active, and which features went quiet. That last field is the difference between "something changed" and "your team stopped running reports" — the hook every later step hangs on.
• Ignore accounts still in onboarding — they have their own play. This one is for established accounts changing behaviour.

A score drop tells you something changed; it doesn't tell you how much it matters. Before a human ever sees the alert, the payload picks up four more fields:

• ARR & plan tier, straight from your CRM or billing system.
• Renewal date — a drop 60 days before renewal is a different animal to a drop the week after signing.
• Account owner, so the alert can be routed rather than broadcast.
• Account segment, as defined in Accoil — enterprise vs. self-serve, ICP vs. long-tail. Segments are built from the same traits and calculated traits the score runs on, so the routing rule downstream reads one field instead of five.

The signal arrived with the "what"; this step adds the "so what."

Not every at-risk account can get a CSM's afternoon, and pretending otherwise is how save plays die. The decision reads the enriched record — ARR, renewal date, segment — and splits on a rule everyone understands: high-value accounts (top ARR tier) or anything renewing inside 90 days goes to a human; everything else gets the automated track. Write the rule down — the play only works if nobody has to make a judgment call at triage time.

Post to a shared channel like #cs-alerts and tag the account owner. The alert carries the whole payload — score trend, active-user percentage, what went quiet, ARR, renewal date — so the CSM can act from the message itself without opening a dashboard. A shared channel beats a DM: it creates gentle social pressure and lets a team lead spot anything that sits unclaimed.

This is the step that decides whether the play works. The CSM reaches out with the specific change as the hook — "I noticed your team stopped using the reporting module" beats "just checking in" every time. That specificity comes straight from the "features gone quiet" field the trigger delivered.

• 48-hour SLA from alert to first touch. Track it.
• Lead with the observed change, then listen: is it a champion who left, a budget review, a workflow that moved elsewhere?
• Log the reason code in the CRM. After a quarter you'll know your top three churn drivers — and which ones a save play can actually fix.

Lower-value at-risk accounts get a lifecycle sequence instead of a call — and the same payload personalises it: a "here's what you're missing" email keyed to the feature that went quiet, a case study matched to their segment, and an offer to book time if they want help. Suppress the sequence the moment their engagement recovers — nothing erodes trust like a "we miss you" email to an account that came back last week.

An account exits the play when its score climbs back above its own healthy baseline — not when someone marks a task done. If it doesn't recover within 30 days, escalate: high-value accounts get an exec sponsor touch, the rest get a renewal-risk flag the forecast can see. Either way the outcome is recorded, and your save rate, GRR impact and time-to-first-touch numbers come from this step.