How Intelligent Health Scoring Helped Founders Protect $380K in Revenue

Most SaaS companies discover churn after the fact. A customer cancels, and suddenly your CS team is scrambling to understand what went wrong. By that point, the relationship has usually been deteriorating for weeks or months. The signals were there — you just weren't measuring them.

Customer health scoring changes this dynamic entirely. Instead of reacting to cancellations, you build a systematic, data-driven model that continuously evaluates every account's likelihood of renewing. Think of it as a credit score for your customer relationships: a single number that aggregates dozens of behavioral signals into an actionable risk assessment.

When we first implemented health scoring across our portfolio of 2,400 SMB accounts, our monthly churn rate was hovering around 4.2%. Twelve months later, it had dropped to 2.9% — a 30% reduction that translated to nearly $380K in preserved ARR.

After testing over 20 different inputs, we found that five signal categories consistently predict churn with the highest accuracy:

Product usage depth is the strongest predictor. Not just logins — you need to track whether customers are using your core features. We measure daily active users as a percentage of licensed seats, feature breadth (how many of the top 10 features they use weekly), and workflow completion rates. An account where only 2 of 15 seats are active is waving a red flag, even if those 2 users are logging in daily.

Support ticket sentiment carries more weight than volume. A customer who files frequent tickets but uses positive language ("How do I...") is actually healthier than one who files rarely but with frustrated tone ("This still doesn't work"). We run basic sentiment classification on every ticket and weight negative-sentiment tickets 3x in the score.

NPS and survey responses give you direct signal, but the absence of response is itself a signal. Customers who stop responding to surveys have a 2.3x higher churn rate in our data than those who respond, even with low scores. At least a detractor is still engaged enough to tell you they're unhappy.

Payment patterns reveal risk early. Failed payments, downgrades, removal of seats, and delayed renewals all contribute. We found that customers who remove even one seat within the first 90 days churn at 4x the baseline rate.

Engagement with your team matters. Customers who attend QBRs, respond to CSM emails, and join webinars churn at half the rate of those who go dark. We track response time to CSM outreach as a proxy for relationship health.

You don't need a PhD in data science to build an effective health score. Start with a simple weighted model on a 0-100 scale.

We allocate weights as follows: product usage gets 35% of the total score, support sentiment gets 20%, NPS/survey data gets 15%, payment health gets 15%, and engagement gets 15%. Within each category, define 3-4 specific metrics and normalize them to a 0-100 range.

For thresholds, we use three tiers. Accounts scoring 70-100 are "healthy" — they get standard touchpoints and expansion plays. Accounts at 40-69 are "at risk" — they trigger a proactive outreach sequence from the CSM within 48 hours. Accounts below 40 are "critical" — they get an immediate executive-sponsored save attempt.

The key insight is that your initial weights will be wrong, and that's fine. Run the model for 60 days, then compare predictions against actual churn. Adjust the weights based on which signals were most predictive for your specific customer base. We recalibrate quarterly and see meaningful improvements each time.

Here's a concrete example of health scoring in action. One of our mid-market accounts — a 50-seat deployment paying $24K ARR — had always been considered a happy customer. Their NPS was 8, they renewed on time the previous year, and their primary admin was responsive.

But when we implemented health scoring, their score came in at 52. The reason: seat utilization had dropped from 78% to 31% over three months, and their usage of our reporting module (a sticky feature) had gone to zero. Something had changed internally.

Our CSM reached out within 24 hours. It turned out the customer had hired a new VP of Operations who was evaluating competitors. They were already in a pilot with another vendor. Because we caught it early, we were able to schedule an executive meeting, address their concerns about our reporting capabilities (we had actually shipped improvements they hadn't seen), and ultimately retained the account with a 12-month renewal.

Without health scoring, we would have found out when they sent the cancellation notice — probably too late to save it.

After 12 months of running our health scoring program, the numbers speak for themselves. Monthly gross churn dropped from 4.2% to 2.9%. Net revenue retention improved from 98% to 107% as the CS team shifted time from reactive saves to proactive expansion. Average save rate on at-risk accounts went from 15% to 41%.

The hardest part wasn't building the model — it was operationalizing it. Making sure scores updated daily, alerts fired reliably, and CSMs actually acted on the data. That's the problem SaaSy was built to solve.

SaaSy continuously computes health scores across all your accounts by pulling data from your product analytics, support desk, billing system, and CRM. When an account's score drops below your configured threshold, it automatically triggers the right playbook: alerting the assigned CSM, drafting a personalized outreach email, and scheduling a check-in. No spreadsheets, no manual score calculations, no accounts falling through the cracks.

If you're still tracking customer health in spreadsheets — or worse, not tracking it at all — you're leaving revenue on the table every single month.