Can you predict company failure from filed accounts? We tested 100.

On 10 June 2026, The Gazette, the UK’s official public record, published corporate insolvency notices for 110 companies. Liquidations, administrations, winding-up orders. Businesses that had run out of road, each one almost certainly leaving unpaid suppliers behind.

We took every single one of them, and we asked a simple question: if you had checked these companies before doing business with them, using nothing but the accounts they had filed at Companies House, would you have seen it coming?

We scored the last set of accounts each company filed while it was still trading, exactly as CompanyIQ would have seen them at the time. No knowledge of what happened next, just the public record, as it stood.

How we kept ourselves honest

Before running anything, we set rules. We took every notice from the day, not a curated sample. We set no targets and no thresholds in advance, and committed to publishing whatever the results showed, including the failures our analysis missed. We scored each company as of the day its last accounts were filed, so the analysis could not see anything that happened afterwards. And we have published the full company list, with Companies House links, so you can check every claim in this post yourself. The detailed methodology is at the end.

Of the 110 notices, 2 could not be matched to a company record, 6 had never filed any trading accounts at all (more on those later), and 2 were removed for data errors. That left exactly 100 companies.

The headline: 81 of 100 were already in the red zone

The average company in our study scored 42 out of 100 on its final accounts. For comparison, the average trading company we analyse scores 55. And 81 of the 100 scored below 50, putting them in our Poor or Critical bands, the territory the CIQ Score marks as serious concern. Among trading companies, only around a third score that low.

Not a single one of the 100 scored Excellent. Only 3 scored Good. On the public record, most of them already looked like exactly what they turned out to be.

What the failed companies had in common

74% of the companies that failed were showing two or more serious warning signs at the same time (among trading companies, only 20% look like that).

One warning sign is common, plenty of decent businesses have a tight year. But warning signs arriving together is nearly four times rarer among companies that survive, and that was the failure profile.

The two signs that mattered most

Two warning signs stood out, each roughly five times more common in the companies that went on to fail.

The first: they owed more than they owned. 45% of the failed companies had liabilities bigger than everything they had, against 10% of trading companies. If one of these businesses had sold every asset it possessed, it still could not have paid everyone back.

The second: they could not cover their bills. In 48% of the failed companies, the money due to go out over the next year was more than the money they had or were owed, against 10% of trading companies.

About a third of the failed companies had both at once, it is a business struggling but still trading and still taking orders. Possibly from your sales team.

What did NOT predict failure

Stable directors were no protection.

The failed companies actually scored slightly better on director quality than trading companies do. Most were steady, founder-run firms with long-serving boards that looked fine right up to the end. A familiar face at the top tells you almost nothing about the balance sheet underneath.

Official warnings barely existed, 88% of the failed companies had no auditor warning of any kind attached to their accounts; small companies are not required to be audited, so nobody was ever going to sound the alarm, the warning was in the numbers, and only in the numbers.

78% of the failed companies had filed accounts late at some point, but so had 59% of trading companies. A repeated pattern, sitting alongside money problems, is when it becomes part of the pile-up.

Most of these companies never published a profit figure

Here is a detail that should change how you think about checking small companies. 91 of the 100 filed under the small companies exemption, which means no profit and loss account on the public record. This is completely legal and extremely common.

You might expect that to make the warning signs invisible. It did not. The exempt companies were flagged at almost exactly the same rate as the ones that published full accounts, because the warning signs live in the balance sheet, which every company must file. What a company owns, what it owes, and what is left over turned out to be enough.

The warning was visible up to two years out

Accounts are filed up to nine months after the year they cover, so the last accounts before a failure can be old. In our study, the gap between the final accounts and the insolvency ranged from under a month to several years.

Whether the accounts were filed six months before the failure or two years before, roughly four in five of the companies were already flagged below 50. The distress was not a sudden event that appeared at the end; for most of these companies, it had been sitting on the public record for a long time.

The ones we missed

Three of the 100 scored Good on their final accounts: scores of 76, 69 and 65. One of them entered compulsory liquidation just three and a half months after filing accounts that looked genuinely healthy.

Seven companies showed none of the major warning signs at all and no method based on filed accounts will ever see those coming. What the accounts can tell you is whether a company isalready fragile. For 81 of these 100, the answer was yes.

Six more companies from that day’s notices never made it into the study, because they went from incorporation to insolvency without ever filing a set of trading accounts. For those, there were no numbers to check at all. That is a different problem with a different answer, and it gets its own post later in this series.

What this means when money is on the line

If you sell on credit terms, the practical lessons from these 100 companies are short. Before you extend terms, ask two questions of the filed accounts: does this company owe more than it owns, and can it cover what is due this year? Those two facts are sitting in the public filings of every UK company, and they were visible in most of these failures long before the end.

This is now built into every CompanyIQ report. Alongside the CIQ Score, each full report shows a count of five warning signs, the same checks this study found stacking up in the companies that failed, each one a plain yes or no. You can read exactly what they mean on the How It Works page.

The warning signs of company failure are usually already on the public record. The hard part is not the data, it is knowing what to look at, and we have just shown you.

This is part one of an ongoing series. We will run the same test again on another day’s insolvencies and publish what we find, whatever it shows. If you extend credit to UK companies, run your first analysis at company-iq.co.uk.

Methodology, for the sceptics

Selection: every corporate insolvency notice published in The Gazette (London edition) on 10 June 2026, across creditors’ voluntary liquidation resolutions, winding-up orders, and administrator appointments. Members’ voluntary liquidations, which are solvent closures, were excluded by category. The mix was 75 voluntary liquidations, 21 court-ordered liquidations and 4 administrations, which mirrors the real shape of UK insolvency. Two companies appear in the sample that belong to the same group and failed together.

Scoring: for each company we identified the last non-dormant accounts filed before the insolvency notice, and analysed them through the identical scoring process used for every live CompanyIQ report. The analysis saw the companies as they stood on the day those accounts were filed: filing history cut off at that date, the company presented as active (which it was), and the board reconstructed as it existed at the time. Director histories across other companies were excluded, because that data is only available as of today and could leak knowledge of later failures. Comparison figures for trading companies come from the latest analysis of each active company in the CompanyIQ database, which is a base of real user-requested analyses rather than a randomised control sample.

Verification: the full list of all 100 companies, with their insolvency type and date, the exact accounts scored, the score, the key warning signs, and a direct Companies House link for each, is available to download here: the evidence pack. Every company in it is a matter of public record.