What Is a Customer Health Score? A Practical Guide for SaaS Teams

Author: Inputly

What is a customer health score?

A customer health score is a single number, usually 0 to 100, that summarises how well each customer account is doing. It is built from weighted signals: how deeply people use the product, what they say in surveys, how often they contact support and whether billing runs clean. The point is not the number itself, it is the early warning. A well built score shows you which accounts are quietly drifting before the renewal conversation starts, and it tells the whole team where to spend attention this week. Treat it as a weather forecast, not a verdict, and always keep the signals behind the score visible.

How do you calculate a customer health score?

1. Pick four to six signals that plausibly predict staying or leaving. More than six makes the score impossible to explain.
2. Normalise each signal to 0 to 100, so active users, ticket counts and survey answers can be compared.
3. Give each signal a weight that sums to 1.0, for example usage 0.40, feedback 0.25, support 0.20, billing 0.15.
4. Multiply, add, and store the raw values next to the result.
5. Sort accounts into bands: healthy (80 and above), watch (60 to 79), at risk (below 60).

| Signal group | Example metric | Direction | Suggested weight | | --- | --- | --- | --- | | Usage | Weekly active users, key actions per account | Higher is better | 0.40 | | Feedback | Latest survey score, open comments, unanswered asks | Higher is better | 0.25 | | Support | Tickets per month, severity, time to resolution | Lower is better | 0.20 | | Commercial | Overdue invoices, seat count, days to renewal | Cleaner is better | 0.15 |

Two rules keep the score honest. First, every account is scored on the same formula, or the bands mean nothing. Second, if a signal cannot change a decision, drop it. A score that never triggers an action is a dashboard nobody opens.

Which signals actually predict churn?

Usage depth beats usage volume. An account with 40 weekly logins but no integration, no invites and no data exports is a shallow account, and shallow accounts leave first. Track adoption of the actions that sit at the centre of your product, which is exactly what feature adoption measurement is for. Then add the human signal: customers who stop answering surveys, stop attending calls or stop sending feedback are telling you something. Watch for a drop in inbound product feedback from an account that used to write every week, because the silence usually starts months before the cancellation. Support patterns matter too, but read them carefully: a spike can mean trouble, and so can a sudden drop to zero.

How is a health score different from NPS or CSAT?

NPS and CSAT are survey metrics: they measure what customers tell you in one moment. A health score is a composite that mixes what customers say with what they actually do. NPS predicts the relationship over time and works best at the portfolio level, while CSAT measures one interaction. Neither moves fast enough to catch risk in a single account. Use the survey metrics as inputs, not as replacements, and read them next to behavioural data. A customer who rates you 9 out of 10 and logs in twice a month is a different case from one who rates you 6 and uses the product daily.

How do you turn a health score into action?

A band without a playbook is decoration. Write one playbook per band and keep each playbook to three steps.

| Band | Trigger | Action within 5 days | | --- | --- | --- | | Healthy | Score 80 or above | Ask for a case study, a reference call or feedback on the roadmap | | Watch | Score 60 to 79 | Review adoption gaps, run a short onboarding refresh, answer open requests | | At risk | Below 60 | Executive check in, list every unresolved request, agree a fix with a date |

The at-risk playbook should always end with a written answer to the customer's open requests, which is the same discipline as closing the feedback loop. Score drops are also feedback, so route them into the same place as survey answers and feature requests. If you are building that layer now, feedback analytics covers how to read the qualitative side, and reducing customer churn covers the retention tactics that follow.

What are the most common mistakes?

The first mistake is scoring only what is easy to fetch, which usually means login counts, and calling it health. The second is hiding the inputs, which turns the score into a black box nobody trusts. The third is treating a single drop as a crisis: check the trend over four weeks before you escalate, and compare the account to its own history rather than to the portfolio average. The fourth is never checking accuracy. Twice a year, take the accounts that churned in the last quarter and ask how many sat in the at-risk band at least 60 days before the renewal date. That number is your real score. If it is low, change the weights. The fifth is letting the score replace human contact. The score decides who gets a call, it does not decide what you say on it.

Frequently Asked Questions

1. How many accounts do you need before a health score makes sense? Around 50 is a useful floor. Below that, a shared risk list with notes is faster and more accurate than a formula.
2. Should health scores be visible to customers? Usually not, at least not as a raw number. Share the underlying status, such as adoption progress or open requests, rather than an internal risk band.
3. What is a good score distribution? Most accounts should sit in the healthy band with a small watch group and a few at risk. If 40 percent of accounts are at risk, the thresholds are wrong or the product fit is.
4. Can you build a health score without a data team? Yes. Four signals in a spreadsheet with a weekly refresh beats a perfect model that nobody maintains. Move to automation when the refresh takes more than an hour.
5. How do you keep health scores from becoming busywork? Cap the playbooks at three actions per band, assign an owner per account, and delete any signal that has not changed a decision in two quarters.

Watch account health without spreadsheets. Create a free space on Inputly, collect the feedback signals in one place, and act on them before renewal season.