If you run customer success at a SaaS company, you know the drill: too many accounts, too little time, and not enough insight into who’s healthy and who’s about to churn. This guide is for anyone tired of flying blind and ready to put real numbers behind “customer health”—without getting lost in the weeds or chasing vanity metrics. We’ll walk through setting up automated health scoring models in Endgame, highlight what actually matters, and point out where you can skip the nonsense.
Why Bother With Automated Health Scoring?
Let’s get real: “customer health score” is one of those terms that gets thrown around in meetings but rarely means the same thing to any two people. At its best, a health score gives you a heads-up before a customer churns—or a nudge when it’s time to pitch an upgrade. At its worst, it’s a dashboard widget that everyone ignores.
The trick is to build scores that actually predict something useful (renewal, expansion, churn), not just look pretty. That’s where automation comes in. Tools like Endgame can crunch the data 24/7, flag issues before they blow up, and let your team focus on real conversations, not detective work.
Step 1: Get Your Data in Order
Automated health scoring is only as good as the data you feed it. Garbage in, garbage out. Here’s what to focus on:
- Product usage: Logins, feature adoption, usage frequency, and depth.
- Account milestones: Onboarding completed, support tickets, training attendance.
- Financial signals: Payment status, contract renewal dates, recent upsells or downgrades.
- Engagement: Email opens, meeting attendance, support interactions.
Pro tip: Don’t obsess over tracking everything. Start with the data you already have. You can always add more later once you’re confident the basics are solid.
What to skip: Vanity metrics (like “number of users invited” if nobody logs in). Stick to indicators tied to real outcomes—renewal, expansion, or churn.
Connecting Data to Endgame
Endgame integrates with a bunch of tools out of the box—CRMs, analytics, support platforms. For anything else, you may need to set up a webhook or use their API. It’s usually straightforward, but if your data’s a mess, block off some time to clean things up first. Nothing kills a scoring project faster than inconsistent or missing data.
Step 2: Define “Health” for Your Business
Not every company’s health score looks the same. What signals that an account is thriving or at risk? Before you start clicking around in Endgame, get clear on what “healthy” actually means for your product.
- Expansion potential: Are your best-fit customers using features that lead to higher plans?
- Risk indicators: What do churning customers have in common? (e.g., stopped logging in, downgraded plans)
- Milestones: Are there key actions (like onboarding completion) that separate renewals from churns?
Quick exercise: Grab 5 accounts that renewed, and 5 that churned. List out what they did (or didn’t do) in the months before. Patterns will jump out. Build your score around those, not what’s easy to measure.
Step 3: Build Your Health Score Model in Endgame
Once you’ve got your data and a sense of what matters, it’s time to build. Here’s how to do it in Endgame:
1. Create a New Health Score
- Go to the Health Scores section and click “Create New Score.”
- Give it a clear name (e.g., “Renewal Prediction,” not “Score v5 Final FINAL”).
2. Pick Your Inputs (Signals)
For each signal, you’ll set a rule. Examples:
- Product usage: “Has logged in at least 4 times in the last 30 days.”
- Feature adoption: “Used Feature X at least twice this month.”
- Support load: “Fewer than 3 support tickets in 60 days.”
- Engagement: “Opened last 2 customer success emails.”
You can combine signals with AND/OR logic. Keep it simple at first—three to five signals is usually enough to start.
What works: Focus on signals tied to outcomes (renewals, expansions, or churn). If it doesn’t move the needle, leave it out.
What doesn’t: Don’t get lured into adding every available metric. More signals = more noise, not more insight.
3. Weight and Score Each Signal
Not every signal matters equally. In Endgame, you can assign weights (e.g., product usage = 50%, support tickets = 20%, payment status = 30%). Base this on your earlier analysis, or just your gut to start. You’ll refine as you go.
Pro tip: Don’t overthink the math. It’s better to get a rough score that mostly works and improve it, than to get stuck in spreadsheet hell.
4. Set Thresholds
Decide what counts as “healthy,” “at risk,” or “unhealthy.” For example:
- Healthy: 80-100
- At Risk: 50-79
- Unhealthy: 0-49
You can also trigger specific actions or alerts when an account crosses a threshold—like flagging for a CSM follow-up.
Step 4: Automate Alerts and Workflows
A health score is only useful if it drives action. Endgame lets you set up automated workflows:
- Notify a CSM when an account drops below a certain score.
- Create tasks for outreach to at-risk customers.
- Trigger renewal playbooks for healthy accounts due for renewal.
Don’t go overboard—start with just one or two key automations. Too many alerts and your team will tune them out.
Step 5: Test, Tune, and Ignore the Hype
The first version of your health score will be wrong. That’s normal. The magic is in tweaking it over time:
- Check the “misses”: Did any “healthy” accounts churn? Did any “at risk” ones renew? Dig in and adjust your weights or signals.
- Talk to your team: CSMs know when a score feels off. Use their feedback.
- Drop what doesn’t work: If a signal never helps predict anything, cut it.
What to ignore: Fancy machine learning health scores that promise magic predictions. Until your team really trusts the basics, don’t waste time chasing AI models or “predictive” black boxes. Most teams get more value from clear, simple rules.
Step 6: Share the Score (But Don’t Worship It)
Once your model is decent, make it visible—put it in your account dashboards, share with sales and support, and use it in customer reviews. But remember: it’s a tool, not gospel. There will always be exceptions.
Pro tip: Encourage your team to challenge the score. If they think an account is at risk (but the score says “healthy”), dig in. The goal is to make your model better, not to shut down human judgment.
Skip These Traps
A few things that sound good on paper, but usually waste time:
- Over-customizing for every segment: Start with one model for your main customer type. Only split later if there’s a clear need.
- Scoring based on NPS alone: NPS is lagging, not leading. Use it as one input, not the whole story.
- Building for perfect data: You’ll never have perfect data. Good enough beats nothing.
Wrapping Up: Keep It Simple, Iterate Often
Automated health scoring in Endgame can save your team a ton of time and headaches—but only if you keep it grounded in reality. Start with the data you have, focus on signals that actually predict real outcomes, and don’t get distracted by flashy features or over-complicated models.
Treat your health score like a living thing: adjust it, test it, and don’t be afraid to throw out what isn’t working. Simple, honest, and actionable beats perfect and ignored every time.
Now go build something your team will actually use.