7 signs a customer is about to stop paying
A customer almost never stops paying overnight. Months of signals come first, and nearly all of them sit in data already on hand: in how they pay, how they argue and how they respond. These are the 7 that best predict a default, what detects each one, and what to do about it.
Why the warning almost always arrives late
The problem is not missing information. It is that what exists is scattered: the delay is seen by whoever reconciles, the dispute by whoever handles the case, and the unreturned call by the salesperson. None of the three sees all three, and separately none of them looks serious.
The 7 signs below are ordered by how hard they are to detect, not by severity. The first four come out of data already on hand. The last three have to be looked for outside.
The 4 signs already in the data
1. Actual terms drift away from agreed terms
Not one invoice running late: the trend. A customer who used to pay at 35 days and has spent three months paying at 52 is financing their working capital, and they are doing it with a supplier who lets them.
What detects it: average days to pay per customer, month by month, against the agreed term. One customer at a time, not the ledger average, which hides everything.
What to do: ask before chasing. An internal process change at the customer and a cash problem look very similar in a report and nothing alike in a conversation.
2. Always pays right on the limit
Subtler than the first and more reliable. A customer paying on day 59 of a 60 day term, every month, is never late and is running their treasury on the edge. When a bad month comes, this is the first place they will stretch.
What detects it: deviation from the due date, not lateness. Zero deviation every time is an anomaly as informative as a delay.
What to do: nothing urgent. Flag it, and do not extend the term or the limit without looking at the rest.
3. Disputes appear where there were none
The dispute as a polite way of not paying is a classic. What gives it away is not that disputes exist, it is the change of rhythm: a customer who never questioned an invoice suddenly finding defects in three in a row.
What detects it: open disputes per customer per month, and above all how many close without the amount changing. A dispute that ends with no credit note was not a dispute.
What to do: resolve them fast and in writing. An unclosed case is the best excuse anyone could be handed.
4. Broken payment promises
The most serious of the four and the least recorded. A customer who says "Friday" and does not pay on Friday has done something other than run late: they have spent their credit.
What detects it: logging the promise with date and amount at the moment it is made, and having the system compare it with what arrived. If promises live in the head of whoever called, this sign does not exist.
What to do: escalate immediately. Two broken promises in a row justify requiring payment before the next order ships.
The 3 signs that have to be looked for outside
5. Annual accounts stop being filed
Filing with the companies register is compulsory in Spain. A company that has gone a year without filing, or files a year late, is saying something, and most often it is that the accounts are not presentable.
What detects it: an annual review of the register for the customers that concentrate risk. It does not need doing across the whole ledger.
What to do: review the credit limit before the next campaign, not after.
6. Corporate changes that do not add up
A change of director, a registered address moved to another province, a business purpose widened towards something unrelated. None is bad on its own. Several in a row, within months, is a pattern.
What detects it: register alerts or a paid commercial report. It costs money, and for most customers it does not pay for itself.
What to do: look at who signs the orders. If the director is no longer the same person, neither is the relationship.
7. Entries in default registers
The hardest signal and the latest. By the time a customer appears in a default register, the problem is months old and there are probably other creditors ahead in the queue.
What detects it: a credit information service, on alert rather than on one-off enquiry. Checking on the day credit is granted achieves little if nobody looks again.
What to do: stop the credit that same day and claim what is outstanding. At this point the priority is not the relationship, it is the position in the queue.
The table worth keeping
| # | Sign | Where it comes from | How urgent |
|---|---|---|---|
| 1 | Actual terms drift from agreed terms | Own data | Watch |
| 2 | Always pays right on the limit | Own data | Watch |
| 3 | Disputes where there were none | Own data | Tighten terms |
| 4 | Broken payment promises | Own data | Act today |
| 5 | Annual accounts not filed | Companies register | Review limit |
| 6 | Chained corporate changes | Register or report | Tighten terms |
| 7 | Appears in a default register | Credit report | Act today |
The mistake of turning this into an automatic score
The temptation, once all 7 signs exist, is to assign points and let a number decide. It works badly for one reason: the signs are not independent. A customer with new disputes and broken promises does not have two problems, they have one showing up in two ways, and adding them exaggerates.
What does work is using them to decide who a person calls this week. They are seven filters for ordering a list, not a formula for replacing the judgement of whoever knows the customer.
Frequently asked questions
What is the most reliable sign a customer will stop paying?
Broken payment promises. A delay can be a slow internal process; a broken promise is a decision. It is also the one most often lost, because almost nobody records the promise with a date and an amount.
Can this be detected without paying for commercial reports?
The first four signs, yes: they come entirely out of payment history and dispute records. The last three need external sources, and they only pay for themselves on the customers that genuinely concentrate risk.
How often should the ledger be reviewed?
The internal signs continuously, because once measured they cost nothing. The external ones annually for the bulk, and by alert for the large accounts.
Is a customer who always pays late a bad customer?
Not necessarily. It is a customer whose financing cost is not being charged for. The right question is not whether to drop them, it is whether their margin covers the days they take.
Do guarantees or promissory notes help?
They improve the position once there is already a problem, and they detect nothing. They are security instruments, not forecasting ones, and asking for them late is usually the sign that it is already late.
What about a customer showing three or four signs at once?
Reduce exposure before claiming: stop new credit, require payment of what is overdue before the next order, and put it in writing. Chasing without turning off the tap only moves the hole.
What comes next
The signs are there to avoid arriving late. When it does happen, what follows is measuring and collecting: DSO says what it is costing, the statutory late payment rate says what can be claimed, and the right collection letter says how to ask without breaking the relationship too early.
The four internal signs here are data already on hand that almost nobody cross-references. In FK they surface on the customer record on their own. Book a demo and look at them against a real ledger.