Template
Customer health score template — free CSV
One row per customer, ten columns, and a health score out the other side. The file below has the exact column names the scorer on this page expects, so you can download it, fill it from the tools you already use, and upload it back without renaming anything.
customer-health-score-template.csv
Header row only — 10 columns, no example customers. Add one row per account.
company_name, plan, mrr, signup_date, last_login_date, logins_30d, support_tickets_90d, payment_failures_90d, seats_used, seats_total
The download is headers only, so uploading it untouched will just tell you there are no rows to score. Fill it in first — or jump to the scorer.
What each column is for
Only the account name is required. Everything else can be left blank, and a blank column is scored as neutral rather than as a failure — but each column you do fill in makes the ranking more specific to your business.
| Column | Example | What goes in it |
|---|---|---|
| company_nameRequired | Northwind Logistics | The account name. Every row needs one — a blank name is skipped rather than scored. |
| planOptional | Scale | Plan or tier label. Shown next to the account so you can see whether risk clusters on one plan. |
| mrrOptional | 890 | Monthly recurring revenue in your own currency, digits only. Used to size the risk, never to compute the score itself. |
| signup_dateOptional | 2024-02-11 | When the account started paying. ISO dates (YYYY-MM-DD) are safest; MM/DD/YYYY also parses. |
| last_login_dateOptional | 2025-11-14 | The last time a user from that account was active in the product. This is the heaviest signal, so it is worth getting right. |
| logins_30dOptional | 22 | Number of logins (or sessions) in the 30 days before your export. Ten or more counts as fully healthy on this signal. |
| support_tickets_90dOptional | 1 | Tickets opened or currently open in the last 90 days. Count what you can defend — open tickets and tickets opened are both usable, just be consistent month to month. |
| payment_failures_90dOptional | 0 | Failed charges or dunning attempts in the last 90 days. Two or more is full severity on this signal. |
| seats_usedOptional | 18 | Seats actually active in the last 30 days, not seats assigned. Assigned seats hide the account that pays for ten and uses three. |
| seats_totalOptional | 20 | Seats the account pays for. Leave both seat columns blank if you don't sell by seat — the signal just goes neutral. |
Already have a different export?
You don’t have to rename anything. The scorer recognises common variants — account_name, company, last_seen, sessions_30d, tickets_90d, failed_payments, active_seats, licensed_seats and similar — and it ignores columns it doesn’t know, so extra fields like an NPS score or an owner name are harmless. Keep the header row at the top and one row per account.
How the health score is calculated
The scorer computes a churn risk score from 0 to 100, where higher means more risk. The health score is its mirror image: health = 100 − risk. So a health score of 100 is an account with nothing wrong on any signal, and 0 is an account failing every one of them.
The risk score is a weighted sum of six signals, each converted to a severity between 0 and 1 and multiplied by a fixed weight: recency 30, login frequency 18, payment health 18, support load 14, seat usage 12, tenure 8. The weights add to 100, so the total lands on the same 0–100 scale. The churn risk score guide has the per-signal formula, the missing-data defaults and a worked example you can check on a calculator.
One row, worked through
An account last seen 42 days ago, two logins in 30 days, three tickets and one failed payment in 90 days, 12 of 20 seats used, signed up 300 days ago, scores 65 risk — which is a health score of 35, in the at-risk band. The number is made up to show the arithmetic; the point is that recency alone contributes 30 of the 65, which is why the save play it generates is a re-engagement email and not a billing call.
What the health bands mean
Quarterly check-in, and ask for the referral or case study while they're happy.
One signal is drifting. Monthly review, and watch the trend rather than the number.
A human touch before the next renewal date, built around the worst signal.
This week, highest MRR first. Call, don't email.
Where to get each column
Four exports, twenty minutes, one file. Nothing here needs a data warehouse — a founder with Stripe, a helpdesk and an admin screen has every column.
Billing — MRR, payment failures
Stripe, Paddle or Chargebee: the subscription amount and the invoice history. Export the amount as a plain number and count failed charges from the invoice list. If you bill annually, divide by twelve so one number means the same thing across the file.
Product — last login, logins in 30 days
Your auth provider (Clerk, Auth0, Supabase, Cognito) or your own sessions table. Last login is the single most useful column in the template: it's the one signal that moves before the customer says anything.
Helpdesk — support tickets
Intercom, Zendesk, Help Scout, or a Linear/Jira queue. Filter to the last 90 days and group by account, not by requester, so one noisy user doesn't look like a whole account in trouble.
Admin — seats
Whatever your plan screen reads from. If you don't sell per seat, skip both seat columns; a missing column is scored as neutral, not as a failure.
Using it as a scorecard, not a one-off
A health score is worth most when the same file comes back every month. Keep the column names and the export window fixed — the last 30 days of logins, the last 90 days of tickets and failures — and the score for each account becomes a trend instead of a snapshot. Two things to hold constant:
- The same export window. If this month’s file covers 30 days of logins and last month’s covered 90, every account will look like it got worse.
- The same definition per column. “Open tickets” and “tickets opened” are both fine; switching between them between cycles is not, because the movement you see will be the definition changing rather than the customer.
To hand it to a manager, export the ranked CSV from the scorer below — it carries the score, the band and the driving signals per account — or paste the bands and MRR at risk into whatever one-page scorecard your team already reads. Opening the template in Excel or Google Sheets works too: the CSV imports straight into columns, and saving it back out as .csv keeps it uploadable here.