Health score

A health score turns customer success from reactive to proactive. It can show a wobble weeks before an account decides to leave, when an intervention still works. By the time someone answers a satisfaction survey, they are often already gone, so a good score weights usage over surveys. It also ranks accounts: intense effort on the weak end, a light touch on thriving ones. A high, stable score often points to expansion revenue, since the customer has fully deployed the product.

• Build the score from metrics that predict churn in your own business, not from a borrowed template.
• Weight usage and engagement above survey answers, which report a problem only after it exists.
• Check the weighting against history: trace an account's score before it left, and fix what does not match.
• Watch for signals that cut both ways. Low ticket volume can mean a happy account or one getting no value.
• Set a threshold that triggers a specific action, and route it to the account owner.
• Revisit the model every quarter or two, since a formula built a year ago may score a business that no longer exists.

A health score condenses many signals about one customer account into a single number or a red, amber and green rating. Typical inputs are how often people log in, how many of the paid seats are used, whether key features are adopted, how many support tickets are open, how recently the main contact replied, and whether invoices are paid on time. Each input gets a weight, and the weighted sum is the score.

It exists because a team managing hundreds of accounts cannot watch each one on instinct.

• Copying a template. A score built for another business weights signals that may not predict churn in yours. Test every input against your own history.
• Building it from surveys. Satisfaction scores arrive late and many customers do not answer. Usage and engagement show a problem earlier.
• One score for every customer. A ten-seat team and a 500-seat enterprise use the product in different ways. Score by segment, or at least check that the thresholds fit each.
• A score nobody acts on. A red account with no named owner and no playbook is a number on a dashboard. Tie each threshold to a specific action.
• Trusting it without checking. If accounts that score green still churn, the model is wrong. Compare scores with actual renewals every quarter and adjust the weights.
• Hiding the inputs. A single number is useful for sorting accounts, but the owner needs to see why it fell: fewer logins, an unpaid invoice, a champion who left.