In this article
What a score is for, and what it needs
The score exists to allocate attention. An account owner with eighty customers cannot call all of them every week; the score says which ten. It needs data the business already has in a form it can read weekly: order dates and values, usage if visible, complaints logged, contacts recorded, payment days. If contacts and complaints are not recorded anywhere, the score will be built from orders and payment alone, which is thinner but still useful.
Decide the reader: the account owner weekly, the owner monthly on the pattern. A score nobody reads is a spreadsheet.
Signals, weights, thresholds and limitations
Signals are the observable facts that, in your business, preceded customers leaving — found by looking at the last twenty who left. Typically: order recency and frequency against the customer’s own pattern; order value trend; range breadth; usage or adoption where the product allows; complaints in the last quarter and whether resolved; days since last conversation from your side; a change of contact; payment days trend. Weights: the signals that most preceded loss in your own history weigh most, and the sum is scaled to a hundred, or to red, amber and green. Thresholds: set them so that the red list is short enough to act on — ten to fifteen percent of accounts — and adjust after a quarter.
Limitations: a score sees only what is recorded; it does not see the competitor who visited last week, the customer’s new owner, or the project that ended. It produces false alarms — a seasonal dip reads as risk — and it misses the furious customer who simply went quiet, unless silence is a weighted signal. It is a torch, not a map.
Build a simple score from available evidence
Start with four signals: order recency against the customer’s usual interval; order value this quarter against last; an open or recent complaint; and days since the last recorded conversation from your side. Score each zero to three, weight them from your lost-customer pattern, add them up, and multiply by a value band so that the list is sorted by risk and worth together. Produce it every Monday from the CRM or a spreadsheet fed by exports. That is a health score, and it takes an afternoon to build.
Add signals as the data improves — usage, payment days, a contact change flag — and drop any that never move the list. Review the weights quarterly against what actually happened: did the red accounts leave, did the green ones stay? The score that is tuned to your history is worth ten that came with a software package.
Connect scores to playbooks, owners and review cycles
A score is useful only when it triggers something. Each colour has an owner and a play: red — the account owner calls this week, using the play that matches the dominant signal — the service play for a complaint, the relationship play for silence or a contact change, the commercial play for a value drop; amber — a scheduled touch this month and a watch; green — the normal rhythm, and the eligible next-best offer if the timing fits. Every intervention is recorded, so next quarter’s weight review can see which plays worked.
Review cycles: weekly, the account owner on the red and amber list, fifteen minutes; monthly, the owner on the pattern — how many red, how many recovered, how many lost despite the call; quarterly, the weights and thresholds against outcomes. The score is a routine, and the routine is what protects the customers.
Calculator · use it here or print it
Customer-health score
Score one account on five signals from 0 to 10, set the weights to what predicts churn in your business, and read the health and the play. Change the weights only from evidence — the accounts that actually left.
- Weights total
- 100%
- Should be 100.
- Health score out of 100
- 64.5
- Band (1 red, 2 amber, 3 green)
- 2
- Red: the owner calls this week. Amber: the account manager calls within the fortnight with a reason. Green: the quarterly review; ask for a referral.
- Lowest signal (1 usage, 2 service, 3 relationship, 4 payment, 5 feedback)
- 3
- The play follows the weakest signal, not the total.
Limits: a score is only as current as the signals behind it, and it cannot see what the customer has not told you. Human review of every red and a sample of greens each month keeps it honest.
What you enter stays in this browser and is not sent to us.
Mistakes, and what a working score does
The mistakes: twenty signals nobody trusts; weights copied from a template rather than from your own lost customers; a red list too long to act on; no owner, so the score is admired and not used; no plays, so the call is “just checking in”; and treating the score as a verdict rather than a prompt to look. A safeguard: for a month, note every red account and what the human review found — the score is working if most reds had something real behind them.
A working score puts the right ten customers in front of the account owner on Monday, most of them with something genuinely wrong, in time to fix it. This is the customer-success work we set up — the lost-customer pattern, the signals and weights, the score in your CRM or a sheet, the plays by signal, and the weekly and quarterly review run with you for a quarter — and the free audit starts by scoring your top twenty accounts by hand.
Questions owners ask
Do we need special software?
No. A spreadsheet fed from the CRM and the accounts, produced every Monday, is enough for a small business. The value is in choosing the signals from your own history and reading the list.
How many signals should the score have?
Four to six. Enough to catch the main ways customers drift; few enough that the account owner understands why an account is red. Twenty signals produce a number nobody trusts.
How do we set the weights?
From the last twenty customers who left: the signals that appeared most in their final six months weigh most. Review quarterly against what happened to the red and green accounts.
What should a red score trigger?
A call this week from the account owner, using the play for the dominant signal — service, relationship or commercial — with the outcome recorded. Not a discount, and not a template message.
Can the score replace talking to customers?
No. It tells you whom to talk to. The human review adds what the score cannot see, and the conversation is what changes the outcome.
What does GullySales do?
The lost-customer analysis, the signal and weight design, the score built in your CRM or a sheet, the plays by signal, the owner routine, and a quarter of weekly reviews with your account owners. Scoped in the free audit and priced in writing.