Account health score

Account health score is a composite measure of how likely a customer account is to renew, expand, or churn, expressed as a single number or a band such as green, amber, and red. It blends several underlying signals, usually product usage, engagement breadth, support experience, and commercial standing, into one figure that a customer success team can scan in seconds. If an account scores 82 out of 100 and green, it is on a healthy trajectory. If it drops to 41 and red, it needs intervention.

The score matters because renewal risk and expansion potential are rarely obvious from any single data point. An account can pay every invoice on time and still be quietly disengaging because only one stakeholder ever logs in. A health score forces these scattered signals into one comparable number, so a team managing 200 accounts can decide where to spend the next hour.

Health scores are most useful when they are predictive rather than descriptive. A good score does not just summarise the past, it correlates with the outcome you care about, which is renewal and growth. That is why the weights behind the score should be tuned against actual retention outcomes, not set by intuition alone.

Most health scores are a weighted sum of normalised sub-scores. Each input is scaled to a common range, typically 0 to 100, then multiplied by a weight that reflects how strongly it predicts renewal. The weights should sum to 1 so the final score stays on the same scale. The inputs below are the most common building blocks.

1. 1

   Usage score

   How much the account uses the product relative to what a healthy account of its size does. This is the strongest single predictor for most products. Measure active features, frequency, and depth rather than raw login counts.

2. 2

   Engagement score

   How many distinct users and stakeholders are active. An account driven by a single champion is fragile. Breadth of adoption across roles and teams reduces single-point-of-failure risk.

3. 3

   Support score

   The quality of the support experience, derived from ticket volume, severity, first response time, and sentiment. A spike in high-severity tickets or negative sentiment pulls the score down.

4. 4

   Commercial score

   Payment behaviour, contract value trend, and time to renewal. Late payments, downgrades, and a renewal date inside 90 days all sharpen the signal and adjust the weighting.

A worked example. Suppose usage scores 70, engagement 40, support 90, and commercial 80, with weights of 0.4, 0.25, 0.2, and 0.15. The score is (70 x 0.4) + (40 x 0.25) + (90 x 0.2) + (80 x 0.15), which is 28 + 10 + 18 + 12, giving 68. That single number hides the real story, which is that engagement is the weak link. The value of the score is not the 68, it is what the decomposition tells you to do next.

A metric tree decomposes the health score into the sub-scores that feed it, and then traces each sub-score down to the operational signals a team can act on. This turns a single amber number into a precise instruction about which lever to pull.

The first level splits the score into its four contributing dimensions. Each dimension then decomposes further. Usage breaks into active features, frequency, and depth. Engagement breaks into active user count and stakeholder breadth. Support breaks into ticket severity and sentiment. Commercial breaks into payment health and renewal proximity. When the headline score falls, the tree tells you whether the cause is shallow usage, a shrinking set of active users, a support fire, or a commercial warning.

This is where the gap between a dashboard and a decision closes. A dashboard shows the score dropped. A tree shows that it dropped because the only two active users went quiet, which is a different problem with a different owner than a billing issue.

Health score benchmarks depend on how the score is constructed, so absolute numbers are less useful than the renewal behaviour each band predicts. The most reliable benchmark is the renewal rate observed within each band over the past year. The ranges below are a common starting point for a 0 to 100 score.

The number that matters most is the spread in renewal rate between green and red. If green accounts renew at 92 per cent and red accounts at 48 per cent, the score is genuinely predictive and worth acting on. If both bands renew at similar rates, the weighting needs retuning before the score earns any trust.

Improving the score is really about improving the underlying account, not gaming the number. The metric tree points you at the weakest branch, and each branch has a different owner and a different play.

The decomposition decides where the effort goes. If usage is the weak branch, enablement beats discounting. If engagement is thin, widening adoption beats a new feature. Spending energy on a strong branch while a weak one drags the score is the most common waste of customer success time.

KPI Tree lets you model this by connecting each branch of the health score to the team and action that influences it. Customer success owns engagement and usage depth, support owns the ticket and sentiment branch, and finance owns the commercial signals. Assign RACI ownership on every node so each branch has an accountable owner, and the score pushes to that owner the moment it moves, rather than surfacing only when someone opens a dashboard. The verified impact loop then checks whether the save play actually lifted the score, so you learn which interventions work.

1. 1

   Setting weights by intuition

   Weights chosen by gut feeling rather than tuned against actual churn outcomes produce a score that feels right but does not predict anything. Validate the weights against historical renewals before trusting the number.

2. 2

   Treating the score as the goal

   The score is a proxy for renewal and growth, not the outcome itself. Optimising the number without improving the underlying account, for example by removing a harsh input, makes the score look better while the account quietly drifts to churn.

3. 3

   Using a single input dressed up as a score

   A health score that is really just login frequency under a new name misses support fires and commercial warnings. A composite needs genuinely independent signals to be useful.

4. 4

   Ignoring the trend

   A static green score can hide a sharp decline that has not yet crossed a band boundary. The direction of travel often matters more than the current value, so track the slope, not just the level.

5. 5

   Never recalibrating

   As the product and customer base evolve, the signals that predicted churn last year may not this year. A score that is never retuned slowly loses its predictive power and starts misdirecting attention.