Build a simple customer health score

By the time a customer clicks cancel, the decision was usually made weeks earlier. The warning signs were there: fewer logins, a key user gone, a pile of support tickets. A health score pulls those signs together so your team can see which accounts need attention this week instead of finding out at renewal.

Why it works

Churn is usually the end of a slow decline that shows up in data long before the cancellation. Usage drops, a key person leaves, tickets pile up. Each signal on its own is easy to miss. Combining them into one score makes the decline visible.

A single score also helps a small team spend its limited time well. Instead of checking every account, they look at a short red list each week and act while the customer is still open to help.

How to do it

1. List accounts that churned recently. Look at what they did in the months before they left. Note patterns in usage, support contact, billing and team changes.
2. Pick three to five signals. Good starting points are recent logins or active users, use of your core feature, open or repeated support tickets, failed payments, and whether the main contact still works with you.
3. Score each signal simply. Use green, yellow and red, or 0, 1 and 2. Define the thresholds in plain words, such as “red if no admin has logged in for 30 days”.
4. Combine into one score. Add the points up, or weight the signal that best matches your churn history more heavily. Keep the formula short enough to fit in one sentence.
5. Review red accounts every week. Put the list in front of the person who owns customer relationships, even if that person is the founder. Each red account gets an owner and a next step.
6. Act on the cause. A drop in usage might call for a training session. A string of tickets might need an engineer to look. A new contact might need a fresh onboarding call.
7. Check the score against reality. Every quarter, compare scores with who actually churned or renewed. Adjust signals and thresholds where the score got it wrong.

Example

A project management tool for agencies notices that several accounts churned soon after their original champion left. The team builds a score from three signals: weekly active users, projects created in the last month, and whether the account admin has logged in recently. Each Monday the founder reviews the red list. When an agency’s admin goes quiet, someone emails the remaining users to offer a short call and help naming a new admin. The team now hears about staff changes before renewal instead of after.

Metrics to track

• Churned accounts flagged beforehand. The share of churned accounts that were red or yellow in the months before they left. Should rise as you tune the signals.
• Red accounts with an action. The share of red accounts that get an owner and a next step within a week. Aim to keep it high.
• Red-to-green recovery rate. The share of red accounts that return to green within a set period.
• Gross revenue retention. Revenue kept from existing customers, before expansion. Should improve over the following quarters.

When it backfires

• Too many signals. A score built from a dozen inputs becomes a black box. People stop trusting it and stop checking it.
• No one acts on it. A dashboard that nobody reviews changes nothing. The weekly review and named owners matter more than the formula.
• Measuring the wrong usage. Logins can look healthy while customers ignore the feature that delivers value. Score the action that matches why they bought.
• Treating the score as the truth. A green account can still churn over price or a merger. Use the score to direct attention, and keep talking to customers.

Questions people ask

Do I need a customer success platform to do this?

No. A spreadsheet or a saved query in your database is enough to start. Move to dedicated software only once you know which signals matter and need the score updated automatically.

How many signals should the score include?

Start with three to five. More signals make the score harder to explain and harder to trust. You can add signals later if you find churned accounts the score missed.

Should customers see their own health score?

Usually not as a number, since a low score shown to a customer can feel like a judgement. Share the facts behind it instead, such as features their team has not set up yet.