55% of small firms just watched a customer go bust. Your ledger saw it coming.

In April 2024, a commercial cleaning firm posted on UK Business Forums with a problem that sounds absurd until it happens to you. Their client's parent company had gone into administration. Nobody told the cleaners. They carried on mopping floors for three more weeks before a manager asked them to stop, and only then found out they were now an unsecured creditor.

That story ended well: the business that bought the client paid everything by late May, and the cleaners moved to pro-forma invoicing. Most of these stories don't end that way. And on the evidence of the last fortnight, a lot more firms are about to be in that position.

On 1 October, Bibby Financial Services published its latest SME Confidence Tracker, a survey of more than 1,000 UK SME owners and decision makers. The headline most write-ups led with was that 40% have cut their customers' payment terms to protect cash. The figure I'd pay more attention to sits further down:

The official numbers say something more nuanced, and it's worth being precise. The Insolvency Service's August statistics, released on 18 September, recorded 1,946 company insolvencies in England and Wales. That's 3% fewer than August 2025, and the 12-month rate (about 1 in 200 companies) is slightly lower than a year ago. So this isn't a wave of failures across the board.

But one line moved sharply: administrations. There were 182 in August, up 44% on July and 60% on August 2025. Administrations tend to involve larger, more established companies, the ones with staff, premises, a decent order book and, quite often, a long list of trade creditors. In other words, the kind of customer that features in your top 20.

The next release is on 20 October. Whatever it shows, the question for a £1–50m business isn't "is the economy getting worse?" It's "which of my customers is next, and would I know?"

Read the threads on this subject and one pattern stands out. Hardly anyone says "we had no idea". They say "we should have seen it".

"We kept supplying." The cleaning thread above (UKBF, April 2024) is the clearest version: three weeks of work done for a company already in administration, because nobody in accounts receivable was looking at the register. Older UKBF threads tell the same story with bigger numbers, of firms owed money and told the company is "going into liquidation" after another delivery has gone out.

"Can we set it off?" In a December 2024 UKBF thread, a business owed about £1,000 to a supplier that had gone into administration, but was also waiting for a credit note on faulty stock. A licensed insolvency practitioner replied with the rules on mutual dealings and set-off, and suggested simply talking to the administrators. It's a useful reminder that insolvency creates data problems too: what's owed in each direction, as at which date.

"The administrator came back for the late-payment interest." The Small Business Commissioner has written about a firm that received a demand for nearly £10,000 from a debt collector acting for an insolvent company's administrator. The administrator had gone back through six years of invoices, found ones paid one to three days late, and claimed statutory interest and compensation on each. Administrators have no relationship to protect. (Undated on the Commissioner's site; probably 2024.)

"Customers are just paying slower." Bibby's own commentary this month links the slower payments to tighter lending: 32% of SMEs had a credit line reduced in the last six months and 25% had a funding application declined. When your customer's bank pulls back, you become their bank, whether you agreed to or not.

An honest note on sources: the forum threads above are from 2024, not this month. I've used them because they describe the mechanics clearly, not as proof of what owners are saying right now. The current evidence is the survey data.

Take an illustrative case (not a client). A £12m distributor sells on 30-day terms, with a gross margin of 22%. Its fourth-largest customer, a fit-out contractor, buys about £60,000 a month.

If that £118,000 becomes a bad debt, the distributor needs about £536,000 of new sales at a 22% margin just to stand still (£118k ÷ 0.22). That's nine months of what this customer used to buy.

Every row from June onwards held a signal. None was hidden. The ledger showed payment behaviour getting worse every month while the balance rose, and the public register showed late accounts, new secured lending and a key officer leaving. The problem is that the first set of signals sits with credit control and the second sits on a website nobody in finance opens.

This is a weekend's build for someone comfortable with Power Query, and it costs nothing in data. Here's the design I'd use.

1. Two sources, one key.

2. A small data model.

3. Scoring that anyone can explain. Points, not a black box:

0–2 is green, 3–4 amber, 5 or more red. In the table above, the fit-out contractor scores 1 in May and 5 in June (payment slipping by 10+ days, a growing balance on falling sales, overdue accounts). That's red three months before the Gazette notice.

4. Rank by exposure, not just risk. Exposure = balance owed + open orders + the next 30 days of expected sales. A red customer owing £4,000 is a note; an amber customer owing £118,000 is a meeting. The main dashboard page is a simple scatter: score on one axis, exposure on the other, with the top-right corner shaded.

5. Alert on change, not level. A Monday email (Power Automate, or a scheduled script) listing any customer whose band worsened this week, with the reason in words: "Accounts overdue since 1 June; FD resigned 14 July." If nothing changed, one line saying so. People stop reading alerts that repeat themselves.

6. Optional: insolvency notices. The Gazette publishes insolvency notices as a feed (JSON or Atom, with no paid account required for basic use). Pulling each week's corporate insolvency notices and matching them to your customer list on name and company number catches petitions and appointments that may take days to show up on Companies House.

The manual version still beats doing nothing:

Matching. Trading names, group structures and a parent company that holds the contract but a subsidiary that buys from you will all break a naive name match. Map by hand for the customers that matter.

False alarms. A new charge is often a routine refinancing, and directors resign for ordinary reasons. That's why each signal is worth a point or two, not a verdict, and why the alert explains itself.

Lag. A private company has nine months after its year-end to file accounts, so "overdue" is a late signal. Your own ledger is usually the earliest warning; the register confirms it.

Using it badly. This is a tool for deciding terms, limits and conversations, not something to share with customers or other suppliers. For large single exposures, credit insurance or a commercial credit-reference service is still worth the cost. This is the free layer underneath, not a replacement.

More than half of small firms have had a customer go bust or stop trading in the last six months, on Bibby's numbers. Most of them probably had the evidence in two places: their own ledger and a free government register. Joining those two is cheaper than one bad debt, and it turns credit control from chasing into watching.

If you'd like this built on your own ledger, with the scoring tuned to your customers, that's a typical one-week Data Clarity Sprint. Reply and I'll send the outline.

— Aksel

Sources