Customer health score
A health score turns scattered signals into a single, actionable read on each account. Done well, it focuses CSM time; done badly, it is a dashboard no one trusts.
Quick answer — What is a customer health score? A composite measure, often color-coded, that combines usage, support, relationship, sentiment, and commercial signals to indicate how likely an account is to renew and expand.
What is a customer health score?
A customer health score is a composite rating, often red/yellow/green or a 0–100 number, that summarizes how likely an account is to renew and grow. It combines signals such as product usage, support activity, relationship strength, and commercial factors.
Purpose: The goal is not a pretty number. It is to tell a CSM where to spend the next hour.
Common inputs
| Category | Example signals |
|---|---|
| Product usage | Active users vs licensed seats, key feature adoption, login frequency, trend over time |
| Outcomes | Whether the customer's stated goals are being met |
| Support | Ticket volume, severity, unresolved escalations |
| Relationship | Executive sponsor engaged, champion present, multi-threaded contacts |
| Sentiment | NPS, CSAT, survey feedback, conversation tone |
| Commercial | Time to renewal, payment issues, contract size, expansion history |
How to build one
- Start from churn. Look at customers who churned and customers who renewed. Which signals differed in the months before?
- Keep it simple. Five to eight inputs you trust beat twenty you cannot explain.
- Weight by predictive value. Usage is often the strongest signal, but test it against your own data.
- Segment it. A healthy enterprise account looks different from a healthy SMB account. See segmentation.
- Define the action. Each status should trigger a play, not just a color.
- Validate and revisit. Check quarterly whether red accounts actually churn more often.
An illustrative scorecard
This is a simple example to show the structure, not a recommended model. Your weights should come from your own churn data.
| Input | Weight | Green | Yellow | Red |
|---|---|---|---|---|
| Active users vs licensed seats | 25% | >70% | 40–70% | <40% |
| Key feature adoption | 20% | Core features in use | Partial | Not used |
| Executive sponsor engaged | 15% | Met in last 90 days | Met in last 180 | No contact |
| Open support escalations | 15% | None | 1–2 minor | Major unresolved |
| Progress toward customer goals | 15% | On track | Behind | Not defined |
| Sentiment / NPS | 10% | Positive | Neutral | Negative |
Score each input 2/1/0, multiply by weight, and sum into a 0–100 score. Then define the status thresholds and the play that each triggers.
What each status should trigger
| Status | Meaning | Action |
|---|---|---|
| Green | Healthy and on track | Maintain cadence; look for expansion and advocacy |
| Yellow | Warning signs | CSM investigates within a week; adds an action to the success plan |
| Red | Material risk | Run the at-risk playbook; manager review |
A 60-day implementation plan
- Weeks 1–2: pull data on churned and renewed accounts for the last 12–24 months.
- Weeks 3–4: find the signals that differed in the six months before renewal.
- Weeks 5–6: draft the score and test it retroactively: would it have flagged the churned accounts?
- Weeks 7–8: launch to the team with training and defined plays.
- Quarterly: re-validate against real outcomes and adjust weights.
Keeping it trustworthy
- Require a note when a CSM overrides a score.
- Review false positives and false negatives monthly.
- Document the formula where everyone can see it.
- Cut inputs that do not predict anything.
Common mistakes
- Scoring on activity (calls held) rather than customer outcomes.
- Over-weighting lagging indicators like NPS that arrive too late.
- Letting CSMs override scores without recording why, which destroys trust in the model.
- Building it once and never testing it against actual renewals.
Where AI helps
Modern platforms can learn weights from historical churn and surface the signals driving each score, and language models can read conversations for risk. See AI-powered health scoring and churn prediction with AI. Treat these as a stronger starting point, not a replacement for judgment, and watch which platform supports your approach.
Frequently asked questions
See what CSMs really earn
Compare your pay against real CSM submissions — free, anonymous, takes 2 minutes.