How to Build a Customer Health Score Formula

Here is how I build a customer health score formula. A customer health score formula is a weighted sum: you score each component from 0 to 100, multiply it by its weight, and add the results, with weights that total 100%. My starting model uses product adoption 30%, outcome progress 25%, relationship engagement 20%, support health 15% and commercial health 10%, so an account scoring 82, 68, 55, 72 and 90 gets a customer health score of 72.4.

You can test your own weights in the free health score calculator. This is Part 1 of the Retention Analytics Playbook series. Part 2 places the score inside retention analytics and Part 3 turns it into alerts.

What is a customer health score for?

A customer health score helps a CSM answer three things: which accounts need attention now, why their condition changed, and what should happen next. The score does not replace judgment. It tells you where to look first.

I learned the limit of a usage-only score with a book of hotel and travel partners. One partner had the highest booking volume in its region and a green score. Its operations lead had stopped joining our calls months earlier. At renewal, the commercial terms were reopened from scratch. Usage was strong. The relationship was not, and our score only measured usage.

What is the customer health score formula?

Health score = Σ (component score × component weight)

component score  = the component normalized to a 0–100 scale
component weight = its share of the total, as a decimal (all weights add up to 1.00)
Σ                = add the weighted results of every component

To normalize a signal, map it to 0–100. Seat usage is easy: 41 active users out of 50 seats is 82. For signals with no natural ceiling, such as open critical tickets, write a simple ladder (0 open = 100, 1 = 70, 2 = 40, 3 or more = 10) and keep it written down.

How do I calculate a health score? A worked example

Here is one example account on my starting weights:

Component	Weight	Score	Weighted points
Product adoption	30%	82	24.6
Business outcome progress	25%	68	17.0
Relationship engagement	20%	55	11.0
Support health	15%	72	10.8
Commercial health	10%	90	9.0
Total	100%		72.4

(82 × 0.30) + (68 × 0.25) + (55 × 0.20) + (72 × 0.15) + (90 × 0.10)
= 24.6 + 17.0 + 11.0 + 10.8 + 9.0
= 72.4

72.4 looks fine. But relationship engagement at 55 is the weakest part, so the next step is to check sponsor access and who still attends meetings, not to send a "you're doing great" email. Keep the components next to the total so anyone can see why a score moved. Run your own accounts through the health score calculator to see how much each weight changes the result.

Which signals should go into a health score?

Two or three signals per component are enough.

• Adoption: active users vs purchased seats, use of the features that deliver value.
• Outcomes: milestones from the success plan that the customer agreed to.
• Relationship: sponsor attendance, number of engaged stakeholders, reply times.
• Support: severity, age of open issues, repeat problems, escalations. Ticket volume alone is ambiguous. A quiet queue can mean a happy customer or one who gave up.
• Commercial: payment status, contract changes, time to renewal.

For each signal, write down the source, the owner and the missing-data rule. If a CSM cannot explain a component to a colleague in one sentence, it is not ready.

Why does the trend matter more than the score?

A 68 that rose from 51 and a 68 that fell from 84 are different accounts. The first is recovering. The second may be entering risk. Store every score with a date instead of overwriting it. Each week, review the biggest drops and the component behind them.

Bands help route work. My illustrative split is 75–100 stable, 55–74 watch, below 55 priority. These are examples, not industry benchmarks. Your own renewal history should decide where the lines go, which is what my article on health score benchmarks covers.

I track the score itself on the same monthly dashboard as everything else:

Metric	Target	This month	Last month	Change
Average health score	70+	72.4	74.1	−1.7
Accounts in priority band (<55)	<10%	12%	9%	+3 pts
Relationship engagement (avg)	65+	55	61	−6
Accounts with a named next step	100%	88%	80%	+8 pts

How do I link each band to an action?

Two accounts with the same total can need different responses. Low adoption calls for enablement. A weak sponsor calls for stakeholder mapping. A severe unresolved issue goes to support leadership and the account owner with one shared recovery plan.

What I've seen in practice

• A dashboard nobody opened. On one team, the health dashboard refreshed daily and almost nobody looked at it. We added one column: the next step and its owner. People started using it because it told them what to do on Monday morning.
• Missing data scored as zero. A partner that had not connected its booking feed looked like a churn risk. It was an integration gap. Now I mark missing data as "unknown" and leave that component out of the total, with the weights rescaled.
• One model for every segment. A low-touch monthly plan and a strategic annual account behave differently. Once we had enough accounts, I split the weights by segment.
• Changing three weights at once. When the score moved, nobody knew why. I now change one weight per month and note the date.

Before you trust the score, run it on past data. Compare it with what actually happened: renewals, contractions, churn. Look hardest at accounts that churned while marked stable. The customer health score model guide covers setup in more depth, and the churn rate formula helps you measure the outcome you are checking against.

This week, pick five components and two signals for each, and score ten accounts you know well by hand. If a number surprises you, ask whether the account or the formula is wrong. In Part 2, I place the score inside a wider retention analytics picture.

Frequently asked questions

How do you calculate a customer health score?

Normalize each component to a 0–100 scale, multiply it by its weight, and add the results. Weights must total 100%. With weights of 30/25/20/15/10 and scores of 82, 68, 55, 72 and 90, the customer health score is 72.4.

What components should a customer health score include?

Five usually cover the relationship: product adoption, business outcome progress, relationship engagement, support health and commercial health. Use two or three clearly defined signals per component.

How do you validate a customer health score?

Run the formula on historical data and compare past scores with actual renewals, contractions, expansions and churn. Look for accounts that churned while scored as stable, and for false alarms that used up CSM time.

How should missing data be handled in a health score?

Mark it as unknown, leave that component out of the total and rescale the remaining weights. Scoring missing data as zero makes integration gaps look like churn risk.

Should a health score be shared with customers?

No. Use it internally to decide which conversation to have, then check the situation with the customer. A health score should lead to better questions, not early conclusions.

Sources

• Essential Considerations For Your First Customer Scorecard, Gainsight
• The Value of Keeping the Right Customers, Harvard Business Review