Aartha logoAartha

Customer health score calculator

Score each signal 0–100 and set how much it should count. The point of this tool is less the number than the sensitivity: change the weights and watch the same account move from green to yellow. That is why health scores set once during implementation and never validated tend to have no predictive value.

Your numbers

Breadth and frequency of meaningful use — not logins.

%

Most teams over-weight this because it is the easiest data to get.

Breadth of engaged contacts, responsiveness, champion still in role.

%

Usually the most predictive signal, and usually under-weighted.

Ticket volume, severity, and tone.

%

Has the customer achieved a named business outcome they would defend in a budget review?

%

Weighted health score

71

Watch

Product usage contributes
24.0 pts
Stakeholder engagement contributes
18.0 pts
Support sentiment contributes
18.0 pts
Value realization contributes
11.0 pts
A score of 71 is a watch item. The number matters less than its direction: an account at 71 and falling is more urgent than one at 61 and stable.

Formula

How the calculation works

Health score = Σ (Signal score × Signal weight) ÷ Σ (Weights)

Weights should sum to 100% for interpretability. This calculator normalizes if they do not, so you can experiment freely without the score leaving the 0–100 range.

Getting it right

What most people get wrong

Validate the weights against accounts that actually churned

A health score is a hypothesis about what predicts churn, and most are never tested. Take the accounts you lost last year and check what they scored 90 days before leaving. If they were green, your weights are wrong. This single exercise separates scores that change behaviour from scores that decorate a dashboard.

Usage is the easiest signal and rarely the best one

Product telemetry is available, clean, and quantitative, so it dominates most models. But an account can use a product heavily and churn anyway when the champion leaves or the economic buyer never saw value. Relationship signals predict better and live in meetings and email, not event logs — which is precisely why they get under-weighted.

A score without a reason produces no action

When a score drops from 82 to 64, a CSM needs to know whether to call the champion, escalate to support, or do nothing. If the system cannot say which signal moved and what evidence sits behind it, the rational response is to ignore the number — and that is what teams learn to do.

Weight by segment, and score trajectory not just level

Login frequency may be decisive for a daily-use product and irrelevant for a quarterly reporting tool; one global weighting produces noise in both. And direction of travel usually beats absolute position — falling at 65 is more urgent than stable at 55.

FAQ

Questions, answered

How do you calculate a customer health score?+

Normalize each input signal to a common scale, typically 0–100, multiply each by a weight reflecting its importance, and sum the results. Weights should total 100%. The hard part is not the arithmetic but choosing signals and weights that actually predict churn for your product.

What should go into a customer health score?+

Product usage, stakeholder engagement breadth and recency, support ticket volume and sentiment, and value realization — whether the customer has achieved a named outcome. Weight these by segment, since what predicts churn for a daily-use tool differs from a quarterly reporting product.

Why is my health score not predicting churn?+

Usually because product usage is over-weighted and relationship signals are missing. 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 by checking what churned accounts scored 90 days before leaving, then recalibrate.

How often should health scores update?+

Continuously, or at minimum daily. A score refreshed monthly is a historical record rather than an early warning system — lead time is the entire point, and batching destroys it.

Your next account move is already in the signals

The number is the easy part.

Aartha shows you which accounts are moving it, and the cited evidence behind why.