---
title: Customer Health Score Template (and How to Build the App)
url: "https://goremy.ai/blog/articles/build-a-customer-health-score-app-instead-of-a-spreadsheet"
author: Luis Chavez-Mattos
editor: Luis Chavez-Mattos
editor_title: Director, Product
published: "2026-11-15T19:40:13.424Z"
updated: "2026-11-15T19:40:13.424Z"
topic: Spreadsheet Alternatives
description: Copy a customer health score template with weights, thresholds and a worked example, then see how to turn it into a live app instead of a spreadsheet.
---

# Customer Health Score Template (and How to Build the App)

> Copy a customer health score template with weights, thresholds and a worked example, then see how to turn it into a live app instead of a spreadsheet.

Edited by Luis Chavez-Mattos, Director, Product · Published 2026-11-15

## What is a customer health score template?

A customer health score template is a scorecard that turns several signals about an account (product usage, support history, relationship strength, sentiment, payment status) into one number or color, usually green, yellow or red. Each signal gets a weight, a rule for scoring it from 0 to 100, and a data source. The weighted total tells you which accounts need attention before renewal. The full template is below, ready to copy.

## TL;DR

- **Six signals cover most B2B accounts**: usage, adoption, support, relationship, sentiment and commercial status. Weight them so they add up to 100.
- **Score each signal on a simple 0, 50, 100 rubric** with thresholds you set from your own data, not from a blog post, including this one.
- **Add override rules** for events a weighted average hides, such as a champion leaving or a renewal inside 90 days with no executive contact.
- **A spreadsheet is fine for about 30 accounts** and starts to break when several people edit it, data goes stale, or leaders want a trend.
- **AI can fill in and draft the scores**, but a person should set the weights and check the output against accounts that actually churned.
- **The step after the spreadsheet is a small app** with a database, a scheduled score refresh, a dashboard and an action list. You can describe it in plain English and have it built.

## What goes into a customer health score?

Use these six signals as a starting set. Rename them to match how your company talks about customers, and drop any you cannot measure reliably. A score built on data nobody trusts is worse than no score.

### The template

Copy this table into a sheet, a doc or your tracker.

| # | Signal | Weight | What to measure | Data source | 100 (healthy) | 50 (watch) | 0 (at risk) |
|---|--------|--------|-----------------|-------------|---------------|------------|-------------|
| 1 | Product usage | 25% | Active users as a share of licensed seats, logins per week | Product analytics or admin export | 70%+ of seats active, stable or rising | 40 to 69% active, or falling 2 months in a row | Under 40% active, or sharp recent drop |
| 2 | Feature adoption | 20% | Number of core features in regular use, or the one outcome the customer bought you for | Product analytics, onboarding checklist | Uses the core workflow and 2+ advanced features | Uses the core workflow only | Core workflow not in use |
| 3 | Support health | 15% | Open tickets, severity, time to resolve, repeat issues | Help desk | Few tickets, none escalated | Several open or one unresolved escalation | Multiple escalations or a long-running critical issue |
| 4 | Relationship | 15% | Contact with the day-to-day champion and an executive sponsor | CRM, meeting notes | Champion and sponsor both engaged in last 90 days | Champion engaged, no sponsor contact | No contact in 90+ days, or champion left |
| 5 | Sentiment | 10% | NPS, CSAT or a CSM's gut rating | Survey tool, CSM input | Positive responses or CSM rates "confident" | Neutral, no recent data | Detractor score or CSM rates "worried" |
| 6 | Commercial | 15% | Payment timeliness, contract changes, days to renewal | Billing, CRM | Paid on time, expanding or flat | Late payments or a downgrade request | Past-due, or a stated intent to cancel |

**Formula:** Health score = sum of (signal score × weight). Scores run from 0 to 100.

**Bands:**

| Score | Status | Default action |
|-------|--------|----------------|
| 75 to 100 | Green | Look for expansion and reference opportunities |
| 50 to 74 | Yellow | Assign a named next step within two weeks |
| 0 to 49 | Red | Escalate, build a save plan, review weekly |

**Override rules (apply after the math):**

1. If the champion has left the company, the account cannot be green.
2. If renewal is within 90 days and there has been no executive sponsor contact, the account cannot be green.
3. If the customer has said in writing they intend to cancel, the account is red regardless of score.
4. If a signal has no data, mark it "unknown" and flag it. Do not score it 50 to be safe, because that hides the gap.

### Tracker columns

If you keep this in a sheet, give each account one row with these columns: Account, Owner (CSM), Segment, ARR, Renewal date, one score column per signal, Total score, Status, Override applied, Last contact date, Next step, Next step due, Notes.

### Rules for adjusting it

- **Weights are a starting point.** A product used daily by everyone should lean on usage. A product bought for compliance, used twice a year, should lean on relationship and support.
- **Use different thresholds by segment** if your large and small customers behave differently. Seat activity of 50% may be normal for a big enterprise rollout and alarming for a 10-person customer.
- **Keep it to six signals.** More signals make the score harder to explain, and when a score is hard to explain, people stop acting on it.

## How do you fill it in? A worked example

Take a fictional account, Northwind Logistics. It has 200 licensed seats and renews in 75 days.

| Signal | Raw facts | Score | Weight | Points |
|--------|-----------|-------|--------|--------|
| Product usage | 62% of seats active, flat for 3 months | 80 | 25% | 20.0 |
| Feature adoption | Uses the core workflow, nothing advanced | 50 | 20% | 10.0 |
| Support health | Two open tickets, none escalated, slow resolution on one | 70 | 15% | 10.5 |
| Relationship | Champion engaged, no executive contact in 5 months | 40 | 15% | 6.0 |
| Sentiment | Last NPS response was a 8 | 75 | 10% | 7.5 |
| Commercial | Pays on time, flat contract | 100 | 15% | 15.0 |
| **Total** | | | | **69.0** |

Northwind scores 69, which is yellow. Override rule 2 applies as well (renewal in 75 days, no sponsor contact), so even if the usage score were higher, it could not turn green. The note in the tracker should say what to do: "Book an executive-to-executive call before day 45. Ask the champion for an intro. Share the usage trend and one outcome the team achieved this quarter."

That last step matters most. A health score is only worth having if each yellow and red account ends with a named next step and a date.

Notice that the biggest drag on this account is not usage, which looks fine. It is the relationship gap. A single blended number can hide that, so keep the individual signal scores visible next to the total.

## What prompt fills in the template with AI?

AI is good at the tedious part: reading a pile of notes and ticket summaries and proposing a score for each signal with its reasoning. It is not good at deciding what your weights should be, and it will happily produce confident scores from thin data. Paste this into the AI tool your company approves for customer data.

```
You are helping a customer success manager score account health.

Here is my scoring template:
[paste the table: signals, weights, and the 100 / 50 / 0 definitions]

Override rules:
[paste the four override rules]

Here is the information I have on the account:
- Account name, seats, ARR, renewal date: [...]
- Usage data for the last 3 months: [...]
- Support tickets in the last 90 days: [...]
- Recent meeting notes and emails (summaries only): [...]
- Latest NPS or CSAT: [...]
- Billing status: [...]

Task:
1. Score each signal 0, 50 or 100 (or an in-between value if the facts clearly sit between definitions). Show the one or two facts that justify each score.
2. If there is no data for a signal, say "unknown" and do not guess.
3. Calculate the weighted total and the status band.
4. Apply the override rules and say whether any apply.
5. Suggest one specific next step with a suggested due date, and name what evidence would change the score.

Return a table, then the next step. Do not invent facts that are not in the information above.
```

**What to check in the answer:**

- Does each score point to a fact you supplied? Delete any score whose reasoning cites something you never gave it.
- Is the arithmetic right? Language models make math slips. Recompute the total yourself or let a spreadsheet formula do it.
- Did it mark missing data as unknown, or quietly fill gaps?
- Is the next step something a person could do this week?

**Where not to trust it:** Do not let AI set the weights, decide a customer's sentiment from a single short email, or label an account green when the data is thin. Also check your company's policy before pasting customer information into any AI tool. If you are unsure, use anonymized summaries or ask your security team which tool is approved.

## Why does the spreadsheet stop working?

A health score spreadsheet is a good first version. It forces you to decide what "healthy" means. But CS teams tend to outgrow it in the same few ways:

- **The data goes stale.** Usage numbers are pasted in on the first Monday of the month, so by the 20th the score describes a customer who no longer exists.
- **Nobody is sure which version is current.** Two CSMs edit copies, and the leadership review uses a third.
- **There is no history.** You can see today's score but not whether it has been sliding for six weeks, which is the thing you most want to know.
- **Formulas get overwritten.** One person types a number over a formula and the total silently stops updating.
- **Nothing triggers action.** A red account is a red cell. No one is notified, and no task is created.
- **Access is all or nothing.** Executives want a summary, CSMs want their book of business, and neither should have to edit the master file to get it.

If you have fewer than about 30 accounts and one person maintaining the sheet, keep the sheet. If you are seeing three or more of the problems above, an app is worth building.

## What should a customer health score app do?

Before building anything, write down what the app needs. This list also works as a brief for a developer, an internal tools team, or an AI builder.

1. **Store accounts.** Name, owner, segment, ARR, renewal date, seats.
2. **Store signals over time.** Each refresh saves the raw facts and the per-signal scores with a date, so you get trend lines instead of a single snapshot.
3. **Calculate the score automatically.** Weights and thresholds live in one settings page that an admin can change. The app applies the override rules and shows which one fired.
4. **Refresh on a schedule.** Data comes in from a CSV upload or from the systems that already hold it (CRM, help desk, billing, product analytics), on a daily or weekly cycle.
5. **Show the right view to each role.** Executives see portfolio totals: ARR by status, accounts that changed band this month, upcoming renewals at risk. CSMs see their own accounts sorted by urgency. Managers see both.
6. **Make every red and yellow account carry an action.** A next step, an owner and a due date. Overdue actions surface at the top.
7. **Notify people.** A message to the account owner and manager when an account drops a band.
8. **Keep notes and an audit trail.** Who changed a score, an override or a weight, and when.
9. **Handle sign-in and permissions.** Only the right people see customer data.

A useful first version does points 1 to 3 and 6. Add the data connections and notifications once people use it weekly.

## How do you check that the score is right?

A health score that nobody trusts is decoration. Before you roll it out, backtest it.

1. **Pick 15 to 30 accounts from the last year or two**, including some that churned, some that expanded, and some that renewed flat.
2. **Reconstruct what you would have known 90 days before renewal** and score them with the template.
3. **Compare.** Did the churned accounts land yellow or red? Did the expansions land green? Where the score was wrong, find out which signal misled you.
4. **Adjust weights and thresholds**, then test again on accounts you did not use the first time, so you are not just fitting the past.
5. **Review quarterly.** Products change, customers change, and a signal that predicted churn last year may not this year.

Be honest about what a sample this small can tell you. It shows whether the score is obviously broken. It does not prove precision. If your company has a data team, ask them to run a proper analysis of which signals correlate with renewal.

Also watch for the common failure: a score that is green for almost everyone. If 80% of accounts are green and you still lose customers, your thresholds are too generous or you are missing a signal, usually relationship or outcome.

## Can you build a health score app without waiting on engineering?

Yes, and this is a good fit for [Remy](https://goremy.ai), which builds full-stack business apps from a plain-English description. Paste the template and the "what the app should do" list above into Remy as your description. It writes a spec, builds the app with its own database and sign-in, deploys it to a live URL, and a typical first build takes about an hour. Remy costs $99 a month ($79 billed annually), and AI usage is billed at cost; getting an idea to a deployed app typically runs $100 to $500 in tokens. You own the code. Check with IT about where customer data may live before you load real accounts.

## Frequently Asked Questions

### What is a good customer health score?

There is no universal number. In the template above, 75 and up is green, but those cut-offs are a starting point. A good score is one that has predicted renewals and churn in your own history. Backtest it on past accounts and move the thresholds until green accounts renew and red accounts do not.

### How many metrics should a health score have?

Most teams do well with six. Fewer than that and the score ignores something important, such as relationship strength. More than that and nobody can explain why an account is yellow, which means nobody acts on it.

### Should health scores be automated or set by the CSM?

Use both. Let data drive signals like usage, tickets and payments, because it is consistent and current. Let the CSM supply judgment on relationship and sentiment, and allow a documented override. If the CSM's gut and the number disagree repeatedly, treat that as a clue about a missing signal.

### What is the difference between a health score and a churn prediction?

A health score is a transparent rule-based summary that a person can read and challenge. A churn prediction is usually a statistical model trained on past outcomes that estimates the probability of cancelling. Many teams start with a health score, learn which signals matter, and add a model later if they have enough history.

### Can I use AI to calculate the health score?

AI can read notes and tickets and propose per-signal scores with reasons. Have the app or spreadsheet do the weighted math, and have a person review any score where the evidence is thin. Never rely on AI output you cannot trace back to a fact.
