Customer Health Score

A composite indicator combining usage, engagement and support signals to estimate the likelihood an account renews.

What it attempts

To summarise many signals into one indicator so limited attention reaches the accounts most at risk. The ambition is reasonable; the usual execution is not, because the model is rarely tested.

The unvalidated-score problem

Most health scores are assembled from components that sound predictive — logins, tickets, feature adoption — with weights chosen in a meeting. Whether the resulting number predicts churn is an empirical question that is almost never asked, and the answer is frequently that it does not.

Fitting it properly

  1. Assemble accounts with known outcomes over at least a year, with their signal history as it stood before the outcome.

  2. For each candidate signal, compare its distribution among churned and retained accounts. Keep the ones that separate.

  3. Weight by observed separation rather than by stakeholder preference. This is the step that turns a score into a model.

  4. Validate on a holdout: rank accounts by score at a past date and check churn rate by band. A score that does not separate the bands has failed.

  5. Re-fit on a schedule. Product and segment changes move what every signal means.

Signals worth considering

  • Depth of use — breadth of features and users, not login frequency, which mostly measures habit.

  • Trajectory — direction of change over recent months, often more predictive than level.

  • Champion presence — whether your known advocate is still there. A departure is among the strongest single signals available.

  • Support pattern — repeated unresolved issues rather than volume, which can indicate engagement as easily as trouble.

  • Commercial position — time to renewal, discount level, and contract terms.

Where it misleads

  • A single composite hides which component moved, which is what determines the response.

  • Usage-based scores miss accounts where the economic buyer never uses the product and the decision is made elsewhere.

  • Scores used as team targets stop measuring risk, since the components are directly influenceable.

RELATED TERMS

COMMON QUESTIONS

What is a customer health score?
A composite indicator combining usage, engagement, support and commercial signals into a single risk view. Its value depends entirely on whether the components were fitted to actual churn outcomes.
Why are health scores usually inaccurate?
Because the components and weights are chosen by intuition and never validated. A score that has never been tested against who actually churned is a plausible-looking number with unknown predictive value.
How do you validate a health score?
Take accounts as scored six months ago and compare against who churned since. If low-scoring accounts did not churn at a higher rate, the score does not work regardless of how sensible its inputs read.
Should health scores be shown to customers?
Generally not. A score is an internal risk instrument built from partial signals, and exposing it invites a discussion about the model rather than about the relationship.

Learn how to apply this: RevOps 101: Revenue Operations Foundations

Definitions are the vocabulary. The courses are where you learn to operate it, with the interactive audit tools.

See the course