# How to Turn Health Score Alerts Into Playbooks People Follow

> What a customer success playbook needs, the four I start every team with, and how I measure response time, recovery rate and save rate.

Source: https://clarivoxx.com/articles/customer-health-score-playbooks-turn-alerts-into-action · Author: Abdessamad Ghanem · Clarivoxx · Customer Success · Updated 2026-03-05

Here is how I turn health score alerts into playbooks my team actually follows. A customer success playbook is a written, repeatable response to one specific health score signal. Each playbook has five parts: a trigger, one named owner, three to five ordered actions, a deadline (usually 24 to 72 hours for high-risk signals) and exit criteria. I start every team with four: adoption drop, champion loss, support escalation and renewal risk.

The playbooks sit on top of a [customer health score formula](https://clarivoxx.com/articles/customer-health-score-formula) and the [early-warning alerts](https://clarivoxx.com/articles/customer-health-score-alerts) built from it. To check which accounts would trigger a playbook today, score them in the free [health score calculator](https://clarivoxx.com/tools/health-score-calculator).

## Why don't health score alerts reduce churn on their own?

An alert tells you which account needs attention. It does not say who owns the response, what to do first, how fast, or how you will know it worked. Without those answers, the response depends on each person's experience and workload that day.

I saw this with a partner who went quiet a few months before renewal. The alert fired. The CSM assumed the account manager would call. The account manager assumed support was handling an open issue. Support thought it was a commercial matter. Three weeks passed before anyone picked up the phone. Nobody was careless. Nobody had been named.

Google's SRE teams found the same thing with on-call work: writing the steps down in advance in a playbook produced roughly a 3x improvement in time to recovery compared with improvising. Customer success is slower than incident response, but the logic holds.

## What should a customer success playbook contain?

Keep each playbook to one page with five parts:

1. **Trigger:** the exact signal, for example a 15-point drop in 30 days, or no login from the primary contact in 21 days.
2. **Owner:** one named role, not a team.
3. **Actions:** three to five steps, in order.
4. **Deadline:** usually 24 to 72 hours for high-risk signals.
5. **Exit criteria:** what closes it, such as the score back above the threshold or a renewal conversation held.

## Which playbooks should a team start with?

You do not need twenty. Start with four that cover the most common failures, and add more when your data shows a pattern.

| Playbook | Trigger | First action | Response window |
|---|---|---|---|
| Adoption drop | Product usage down 30% month over month | Personal outreach to the champion with a usage review | 48 hours |
| Champion loss | Primary contact leaves or goes silent | Identify and engage a new sponsor | 72 hours |
| Support escalation | Repeated critical tickets in 14 days | Executive check-in and a written recovery plan | 24 hours |
| Renewal risk | Low score within 90 days of renewal | Value review and renewal strategy call | 5 business days |

## How do I write playbook actions people actually follow?

Nobody reads a wall of policy text while an account slips away. Start each step with a verb and make it something you can finish in one sitting.

- **Weak:** "Consider reaching out to the customer to discuss engagement."
- **Strong:** "Call the champion within 48 hours. Share the usage trend, ask for the top blocker, and book a 30-minute working session."

Add a short email or call template to each playbook. The first step is where teams lose days, and a ready-to-edit message removes most of that delay. With French- and Arabic-speaking partners, I kept the template in each language so nobody waited for a translation before reaching out.

Wire each playbook to its alert:

- Each alert type links to exactly one playbook.
- The alert shows the owner, the deadline and a link to the playbook.
- If nobody acknowledges it within the window, it goes to the team lead automatically.

For renewal-risk accounts, a shared [customer success plan template](https://clarivoxx.com/customer-success-plan-template) keeps the recovery steps and the customer's goals in one place.

## How do I measure whether a playbook works?

```
Response time  = median hours from trigger to the owner's first action
Recovery rate  = flagged accounts whose score returns to a healthy band ÷ flagged accounts
Save rate      = at-risk accounts that renew ÷ at-risk accounts that reached renewal
Unacknowledged = alerts with no action inside the window ÷ total alerts
```

Worked example for one quarter. 50 accounts triggered a playbook. 32 of them returned to a healthy band: recovery rate = 32 ÷ 50 = **64%**. 18 at-risk accounts reached their renewal date and 14 renewed: save rate = 14 ÷ 18 = **78%**. 3 alerts got no action inside the window: 3 ÷ 50 = **6%** unacknowledged.

| Metric | Target | This month | Last month | Change |
|---|---|---|---|---|
| Median response time | <48 h | 31 h | 44 h | −13 h |
| Recovery rate | 60%+ | 64% | 57% | +7 pts |
| Save rate | 75%+ | 78% | 71% | +7 pts |
| Unacknowledged alerts | <5% | 6% | 11% | −5 pts |

A low recovery rate usually means the trigger comes too late or the actions are too weak. Slow response time is an ownership problem, not a strategy problem. To see what a better save rate is worth in revenue, plug it into the [renewal rate calculator](https://clarivoxx.com/tools/renewal-rate-calculator) and the [health score calculator](https://clarivoxx.com/tools/health-score-calculator) for the accounts behind it.

## What I've seen in practice

- **"The CS team" is not an owner.** Every time I saw a playbook assigned to a team, response times slipped. One role, one name.
- **Playbooks without exit criteria never close.** We once had 40 "open" adoption playbooks, half of them for accounts that had recovered months earlier. A clear exit rule fixed the backlog in a week.
- **Champion loss was the slowest to recover.** Of the four, it was the one that recovered least often for me. The fix was to map a second contact during onboarding, before anyone left.
- **Too many playbooks dilute all of them.** One team had fourteen. Nobody remembered which applied. We merged them back to five.

Review each playbook every quarter against real outcomes. This week, write the adoption-drop playbook on one page, name its owner, and link it to your trend alert. For a broader view that also covers expansion and renewal plays, see [my retention analytics playbook](https://clarivoxx.com/articles/the-retention-analytics-playbook).

## Frequently asked questions

### What should a customer success playbook contain?

Five parts: the trigger that activates it, a single accountable owner, three to five concrete actions in order, a response deadline, and exit criteria that define when the playbook is closed.

### Which customer success playbooks should a team start with?

Four cover the most common failure patterns: adoption drop (usage down about 30% month over month), champion loss, support escalation (repeated critical tickets) and renewal risk (low score within 90 days of renewal).

### How do you measure whether a playbook works?

Track response time after the trigger, recovery rate (flagged accounts whose score returns to a healthy band) and save rate (at-risk accounts that renew). If 14 of 18 at-risk accounts renew, the save rate is 78%.

### How many playbooks does a customer success team need?

Fewer than most teams think. Start with four, assign one owner each, and add new playbooks only when your data shows a recurring pattern that the existing ones do not cover.

## Sources

- [Introduction to Site Reliability Engineering](https://sre.google/sre-book/introduction/), Google SRE book (playbooks and time to recovery)
- [Monitoring Distributed Systems](https://sre.google/sre-book/monitoring-distributed-systems/), Google SRE book