How to build a customer health score before you can afford CS software
A customer health score is a simple, weighted view of which accounts are about to renew, expand, or quietly walk away, built from four signals you almost certainly already have in a spreadsheet: usage, growth, friction, and contact. You do not need Gainsight, Vitally, or a customer success hire to build one. You need forty-five minutes and the willingness to actually look at the numbers instead of trusting your gut.
Most guides to health scoring are written by customer success software vendors, for customer success teams. They assume you have a CSM, a support queue, and a budget line for "customer platform." At 20 to 150 customers and no CS hire, that advice does not apply to you, and the version that does apply barely exists.
Why you need a customer health score before customer 20
Founders can hold customer health in their head when there are 8 accounts. You know who is happy because you talk to them. Somewhere between 15 and 30 customers, that stops being true, and it stops quietly.
The failure mode is specific: a customer stops logging in three weeks before they email asking to cancel. Nobody noticed the drop because nobody was looking at usage data between renewal conversations. The cancellation email feels sudden. It was not sudden. It was invisible, because nothing was tracking it.
The stakes are higher than founders assume. ChartMogul's benchmark data puts the median monthly customer churn rate for early-stage SaaS companies under $300K ARR at 6.5%, roughly double the rate of companies at $1-3M ARR. Churn is heaviest exactly when founders have the least infrastructure to catch it.
A health score is not a customer success luxury. It is the mechanism that replaces "I think they're fine" with "their usage dropped 40% two weeks ago, I should call them today." That difference alone recovers deals that founder intuition misses every time.
The four-signal customer health score model that fits in a spreadsheet
A working customer health score needs four signals, not the 8 to 12 weighted inputs most enterprise CS guides recommend. Vitally's four-metric framework makes a similar case for simplicity, though it's still built for a CS team with a dedicated platform. The founder version below strips it down further, to four signals you can score by hand, each rated 0 to 10 per account, in a single spreadsheet tab:
- Usage trend (weight 40%): is weekly active usage flat, up, or down over the last 30 days versus the prior 30? This is the single most predictive signal for SaaS churn: if usage is declining, everything else is a lagging indicator.
- Seat or workflow growth (weight 25%): has the account added users, connected more integrations, or expanded into a second use case? Expansion signals are the earliest positive counterpart to churn risk.
- Support friction (weight 20%): how many tickets or "quick questions" came in this month, and were they resolved or did they stall? Repeated unresolved friction on the same feature is a stronger churn predictor than ticket volume alone.
- Direct contact recency (weight 15%): when did a founder or teammate last have a real conversation with someone at this account, not an automated email? Accounts that go 60+ days with zero human contact churn at a materially higher rate, independent of usage data, because nobody catches the reason behind the drop.
Multiply each score by its weight, sum them, and you have a single number per account, updated weekly, that took less time to build than a single sales call.
The mistake almost every founder makes with this model
The mistake is not skipping health scoring. It is building it once and never updating it. A health score calculated in January and never touched again is worse than no health score, because it creates false confidence.
The second mistake is scoring every account on the same curve. A 3-person startup customer and a 200-person account behave completely differently: what counts as "healthy" usage for one looks alarming for the other if you use one flat threshold. Segment by account size or plan tier before you set what "good" looks like, even if the segmentation is just two buckets.
The third mistake is treating the score as a report instead of a trigger. A health score that nobody acts on is a spreadsheet, not a system. The number only earns its keep when a drop below a threshold automatically means someone reaches out that week, not "eventually."
How to actually implement this in a week
Here is the sequence that gets a working health score live without a data team:
- List every active customer in a spreadsheet, one row each, with columns for the four signals above.
- Pull the usage numbers from whatever analytics tool already logs activity (even basic login timestamps work if usage tracking is thin). This is usually the slowest step and the only one that matters for the first version.
- Score seat growth and support friction from memory or your support inbox for accounts under 50. This does not need automation yet.
- Fill in contact recency from your CRM or, honestly, your memory and email search for "last time I talked to [account]."
- Weight and sum the four columns, sort by total score ascending, and look at the bottom 10. Those are your this-week priorities, not the accounts on your renewal calendar.
- Repeat weekly. The value compounds because you start seeing trend lines, not snapshots: a score that dropped from 8 to 5 in two weeks matters more than a static 5.
Founders who run this weekly for a quarter consistently report the same result: two or three "surprise" cancellations they would have missed show up as declining scores three to four weeks before the cancellation email arrives. That lead time is the entire point, and it's the same lead time that separates founders who reduce SaaS churn proactively from founders reacting to cancellation emails.
What to do first
Build the customer health score spreadsheet today, not the perfect version, the four-column version. Score every current customer once this week, sort by total ascending, and call the bottom three before you do anything else on your task list. The first pass will be rough. It will still catch something your gut missed.
Frequently asked questions
Do I need customer success software to build a health score?
No. A spreadsheet with four weighted columns, updated weekly, catches the vast majority of at-risk accounts for founders under roughly 150 customers. Dedicated CS software becomes worth the cost once manual updates take more than an hour a week.
What's the minimum number of customers where this matters?
Most founders can track health informally up to 10-15 customers. Past that, invisible churn risk starts accumulating because no single person has full visibility into every account's usage pattern.
How often should I update the health score?
Weekly. Monthly is too slow to catch a usage decline before it becomes a cancellation email; daily is more effort than the signal usually justifies at early stage.
What's the single best signal if I can only track one?
Usage trend. A declining usage pattern over 30 days is the most reliable early warning signal SaaS companies have, more predictive than support tickets or even direct feedback. Baremetrics' churn research reaches the same conclusion: usage data catches risk earlier than billing data ever will.
Should the weights be the same for every customer segment?
No. Enterprise accounts with more stakeholders tolerate lower per-user usage than a 3-person startup customer using the same product. Segment your thresholds by account size before treating any single score as comparable across your whole base.
What do I do when a score drops sharply?
Reach out personally within the week, ask a specific question about what changed (not "just checking in"), and treat the conversation as diagnostic, not a save attempt. Most recoverable churn is recoverable specifically because someone asked early enough.
A health score built this week, however rough, replaces guessing with evidence. The founders who keep their best customers past year one are rarely the ones with the best product. They are the ones who noticed the drop three weeks before the cancellation email, because they were the only ones actually looking. Once the score is live, pair it with a real net revenue retention target so declining accounts have somewhere to point besides a spreadsheet cell.