Customer Health Score

A customer health score tries to answer a question that no single metric can: is this account going to stick around? It pulls together signals from product usage, support interactions, survey responses, and payment history into a single score, usually 0 to 100 or a traffic-light system (green, amber, red).

The concept isn't complicated. The execution is where most teams stumble.

What goes into a health score

The best health scores combine leading indicators from several categories. Product usage is the strongest signal: how often the team logs in, how many features they use, whether usage is trending up or down. The DAU/MAU ratio works as a quick proxy for stickiness. Feature adoption rate adds another dimension, particularly whether customers are using the features that correlate with long-term retention.

Beyond usage, you're looking at support ticket frequency and sentiment, NPS or survey scores, billing history (late payments, downgrades, failed charges), and expansion signals like seat additions or plan upgrades.

A practical formula might weight these as (40% usage) + (25% support satisfaction) + (20% NPS) + (15% feature adoption). But the right weights depend on your product. A collaboration tool where daily usage is table stakes will weight usage more heavily than a quarterly reporting tool.

Building one that actually works

Start simple. Three or four inputs, equally weighted, scored on a consistent scale. Don't try to build a 150-variable model on day one. ZapScale claims 94% churn prediction accuracy with that many data points, but they've been iterating for years.

The practical approach: pick the signals you already track, normalise them to a 0 to 100 scale, and combine them. Run it for a quarter. Compare scores against actual churn outcomes. Adjust weights based on which inputs were most predictive. Repeat.

One thing we've seen across ChurnWard customers: time to value is a strong early indicator. Customers who don't reach their first meaningful outcome within the first two weeks churn at far higher rates. If you can only track one thing for new accounts, track that.

Using health scores to reduce churn

The point of a health score isn't to generate a number. It's to trigger action. Red accounts get a CS check-in call. Amber accounts get a targeted onboarding nudge or usage tips. Green accounts might get an expansion offer.

AI-enhanced models can flag at-risk accounts 47 days before cancellation on average. Even simple models give you a meaningful head start on intervention compared to waiting until the customer hits the cancel button.

This is where health scores feed into reducing voluntary churn. A customer trending from green to amber is a customer you can still save, if you notice in time.

What health scores miss

Here's the blind spot: a perfectly green account can still churn tomorrow if their credit card expires or their bank flags the next charge. Health scores are excellent at predicting voluntary churn. They tell you nothing about involuntary churn from payment failures.

That gap matters. Involuntary churn accounts for 20 to 40% of total SaaS churn. Your healthiest customer can silently disappear because of an expired card. Solving that requires a separate system entirely: automated dunning, smart retries, and pre-expiry card alerts.