How to identify customers at risk of leaving

Behaviour: order frequency and value against the customer’s own run rate — a customer who ordered monthly and has not in six weeks, or whose average order has halved; usage where you can see it; a narrowing of the range they buy. Service: a complaint in the last quarter, especially an unresolved or repeated one; a delivery failure; a warranty claim. Relationship: a change of the person who buys or decides; a meeting declined; calls or messages unanswered; no contact from your side in the last quarter, which is a risk you created. Commercial: payment days lengthening; a request for a price review; a competitor mentioned; a tender or quote request that went elsewhere; a contract or renewal date approaching with no conversation.

None of these alone means much. Two or three together, in an account that matters, is a customer deciding.

Pick four to six signals you can actually see in your CRM and accounts, give each a weight — the ones from your lost-customer analysis weigh most — and produce a score per active customer, weekly, sorted from worst to best. Red, amber and green thresholds. It can be a spreadsheet; the sophistication is in the signals chosen, not the tool. Weight the score by the customer’s value, so that the list the account owner reads on Monday morning is the accounts that matter most and are at most risk.

Then the human review: the account owner looks at the red and amber accounts, adds what the score cannot see — a conversation last week, a known change at the customer, a project ending — and decides which need a call this week. The score finds; the person judges. A score without the review generates false alarms and misses the customer who is silent because they are furious.

Service risk — a complaint, a failure: the level-three complaint process, a senior call, the fix, and a follow-up a week later; the intervention is to be seen to take it seriously. Relationship risk — a new contact, silence: a visit or a call to introduce yourself to the new person and re-establish why they buy from you, with the history and the proof; the intervention is to start the relationship again before the competitor does. Behaviour risk — shrinking orders: a call to ask, plainly, what has changed, and an offer that fits the answer — a range review, a delivery change, a modest incentive on the products they stopped buying. Commercial risk — slow payment, a price challenge, a competitor named: a conversation with the decision-maker about value, a review of terms within policy, and, if it is price alone, a decision about whether the account is worth matching.

Each play has an owner, a script, a timeframe and a record. The plays are written once and used every week; the judgement is in choosing which one.

The mistakes: watching revenue, which reports the loss after it happened; a score with twenty signals nobody trusts; no human review; the same play for every risk; the salesperson as the only relationship, so their departure takes the account; and no record of interventions, so nobody knows what worked. A safeguard: for your top twenty accounts, write the date of the last contact that was not an order or an invoice — any older than a quarter is a risk you own.

Measure by the share of at-risk accounts identified before they stopped buying, interventions run, accounts recovered, and churn by segment over time. Early warning turns most churn into a conversation you had in time. This is the churn-analysis and health-scoring work we do — the lost-customer signature, the score from your data, the weekly review and the plays by risk type, set up in your CRM and run with your account owners for a quarter — and the free audit starts with the twenty customers who left last year.