What is customer health score?
Also known as: Health score, Account health score
A customer health score is a composite metric that summarizes how likely an account is to renew, expand, or churn, calculated by weighting signals such as product usage, engagement, support history, and relationship strength into a single value.
Formula
How to calculate customer health score
Health score = Σ (Signal score × Signal weight)- Signal score
- Each input normalized to a common scale, typically 0–100
- Signal weight
- Relative importance of that input; weights should sum to 1
Worked example
Product usage 80 (weight 0.4), stakeholder engagement 60 (weight 0.3), support sentiment 90 (weight 0.2), invoice timeliness 100 (weight 0.1). Score = 32 + 18 + 18 + 10 = 78.
In practice
What you need to know about customer health score
The signals that actually predict
Most health scores over-weight product usage because it is the easiest data to get. Usage is necessary but insufficient: an account can use a product heavily and still churn when the champion leaves or the budget owner changes. The signals with the highest predictive value are usually relationship signals — is your champion still there, are multiple stakeholders engaged, did anyone respond to the last three emails — and these live in conversations, not event logs.
Why most health scores get ignored
A score that drops from 82 to 64 without explaining why produces no action. The CSM cannot tell whether to call the champion, escalate to support, or do nothing, so they learn to ignore the number. Explainability is not a nice-to-have; it determines whether the score changes anyone’s behavior. Every score change should be traceable to the specific signal and the underlying evidence that moved it.
Validate against outcomes, then recalibrate
A health score is a hypothesis about what predicts churn. Test it: take the accounts that churned last year and check what their score was 90 days prior. If churned accounts were scoring green, your weights are wrong. Most teams set weights once during implementation and never revisit them, which is why so many health scores have no predictive value.
How to improve it
Improving customer health score
Start with three signals, not fifteen
Complex models are harder to debug and rarely more accurate. Begin with usage, stakeholder engagement, and support sentiment. Add signals only after validating that the existing ones predict.
Weight by segment
Login frequency may matter enormously for a daily-use product and not at all for a quarterly reporting tool. A single global weighting applied across segments produces noise in both.
Score trajectory, not just level
An account at 65 and falling is more urgent than one at 55 and stable. Direction of travel is often more predictive than absolute position.
Make every score clickable
A CSM should be able to click a score and see the specific evidence behind it — the email that went unanswered, the meeting where a concern was raised. Without that, the score is a number, not intelligence.
FAQ
Customer health score questions, answered
What should go into a customer health score?+
A useful score combines product usage, stakeholder engagement breadth and recency, support ticket volume and sentiment, and relationship signals such as whether your champion is still in role. Weight these by segment, since what predicts churn for a daily-use tool differs from a quarterly reporting product.
Why is my customer health score not predicting churn?+
The most common cause is over-weighting product usage while ignoring relationship signals. Accounts churn because a champion left, a budget owner changed priorities, or value was never realized by the people who decide — none of which appear in login data. Validate your weights by checking what churned accounts scored 90 days before they left, and recalibrate.
How often should health scores update?+
Continuously, or at least daily. A score refreshed monthly is a historical record, not an early warning system — the point of a health score is lead time, and batching destroys it.
Should health scores be shown to customers?+
Generally no, at least not as a raw number. Scores encode internal judgments and incomplete data, and a customer seeing themselves marked "at risk" changes the conversation unproductively. Share the underlying findings and the plan instead.
Where Aartha fits
Aartha builds health from a cited Customer Memory Graph, so every score change traces back to the specific meeting utterance or email behind it — not an opaque number.
See how it worksRelated terms
Keep reading
Churn rate
Churn rate is the percentage of customers who stop paying for a product during a given period. It is calculated by dividing the number of customers lost during the period by the number of customers at the start of that period.
Churn analysis
Churn analysis is the practice of examining lost customers to identify why they left, which segments are most affected, and which signals preceded the loss — so that retention effort can target causes rather than symptoms.
Customer retention rate
Customer retention rate is the percentage of customers you keep over a given period, excluding new customers acquired during that period. It is the complement of customer churn rate.
Net promoter score
Net promoter score (NPS) is a customer loyalty metric based on a single question: how likely are you to recommend this product to a colleague, on a scale of 0 to 10. It is calculated by subtracting the percentage of detractors (0–6) from the percentage of promoters (9–10), producing a score between −100 and +100.
Product adoption
Product adoption is the extent to which customers actively and habitually use a product to accomplish their work. It is measured through breadth (how many users), depth (how much of the product), and frequency (how regularly) rather than by logins alone.
Your next account move is already in the signals
Know the metric. Know why it moved.
Aartha keeps a cited, time-aware memory of every account — so a health change or a churn signal comes with the evidence behind it.