I used to find out too late.

The UK recruitment industry turns over about £43 billion a year across roughly 30,000 agencies. Second largest market on earth, behind the United States. Almost all of it still runs on someone picking up a phone.

I know, because for a while that someone was me.

The arithmetic of a desk

A resourcer’s job is finding people, and you get good at it fast. You learn where engineers hide. You learn that three different job titles mean the same thing and a fourth means something else entirely. You learn to read a CV for the gap someone is hoping you won’t ask about.

Before that I was cold calling for a living, which teaches you a different lesson: the numbers are the numbers, and they do not care how you feel about them. Roughly 40 dials for one meeting. A 2–3% conversion on a cold list. Eighty per cent of conversions need five or more follow-ups, which means four out of five people who eventually say yes have already said nothing, four times.

You can be excellent at that job and still lose, because the thing that decides it is mostly when you called. The same script into a warm company is a conversation. Into a cold one it is a voicemail. Same words. Same person. Different week.

The asymmetry

Here is the strange thing about recruitment. On the candidate side you have an armoury: LinkedIn Recruiter at roughly £8,800 a seat a year, boolean strings, referral chains, a dozen tools for finding one specific engineer in a country of seventy million.

On the client side you have… a press release.

A company would announce it had raised, and by the time that reached my screen it had reached everyone else’s on the same morning. Every agency in the sector rang that week. Whoever got through first usually got the brief. Getting through first was mostly luck, and I do not enjoy building a career on luck.

It has not got easier since. Finding new clients is now the single biggest challenge for 23% of UK agencies, up seven points in two years, and 83% take between one and six months to close one. Activity is up. Conversion is down. Everyone is working harder on the same list.

What I did not know

The announcement is not the first public record of a raise. It is close to the last.

When a UK company issues shares it must file a form at Companies House within one month. Not eventually. Within a month. The form states how many shares were allotted and what was paid for them — and multiplied together, that is the money. In writing. Signed. On the public record. And for about three in four of these companies, that filing is the only place the money is ever recorded, because nobody writes a word about it.

I should be precise about what that means, because I got it wrong at first. Where a raise is announced, the announcement usually comes before the filing, not after: a company with a PR agency tells the press at close and files the instalments later. We tested sixty companies from one week and not one filing beat the news. The value is not in being early. It is in the majority that never become news at all.

About 750 of those forms are filed every working day. Nobody was reading them and turning them into a list, because reading 750 filings a day is not a job a person wants.

Why it ended up pointed at lenders

The timing problem is not a recruitment problem. It is the same problem on any desk that has to act on something before a competitor does, and lending has the sharper version of it: when a borrower refinances, the outgoing lender’s charge is satisfied and a new one is registered to somebody else. Both are filed. Both are public. And the lender who lost the business usually finds out at the renewal conversation, by which point the decision is months old.

Recruitment taught me the shape of the problem. Lending is where it is worth the most.

So I built the thing that reads them

Every charge registered, every charge satisfied and every change of control at Companies House gets pulled and read. The directors come from the officers register, the controlling person from the PSC register, the lender from the charge itself. On Monday morning what is left goes out as a register: company, event, date, what changed, the director, and a link straight to the filing.

The sample register holds 55 events across 45 companies in defence and advanced manufacturing, which is the first cut of a universe of 105,036 active UK manufacturing and engineering companies. Fourteen rows were checked by hand against source. Fourteen passed.

The part nobody warns you about

Reading a filing sounds simple. It is not, and the ways it goes wrong are the reason this is a product rather than a spreadsheet.

• A charge is not an amount. It names who granted it, who holds it and when. It does not reliably say how much was lent. Any register printing a facility size next to a charge is inferring it, and the moment you infer, you are guessing.
• A satisfied charge does not say why. Refinance, sale of the secured asset, or one security released while the borrowing continues all look identical on the form. Reading every satisfaction as a repayment is how you end up ringing the wrong company.
• The controller is often not a person. Control passes to another company or an overseas entity as often as to a human. Put a corporate PSC in a column meant for someone you can call and the row is worthless.
• Shells file exactly like real businesses. A holding company with no trade registers charges and changes control the same way a manufacturer does, and it passes every obvious filter.

Every one of those was found the hard way, by checking a number that looked wrong against the document it came from. Which is why every row in the register links to its filing: not as a courtesy, but because the discipline of having to be checkable is what keeps it honest.

Two things the data told me that I did not expect

Reading the sector’s history back through the same pipeline turned up two findings I have not seen written down anywhere.

Control changes hands more often than you would guess. Of 150 established companies sampled in the sector, seven changed control inside a year. That is 4.7% of that sample. It is a rate for the defence cut and not for the whole 105,036, which has not been measured. Even so it is a great deal of ownership moving quietly. None of it is announced, because a PSC filing is a duty rather than a press release.

And roughly a third of rows are not operating businesses. In one measured week, 96 of 284 rows were shells, holding vehicles or special-purpose companies. They pass every obvious filter: real company, real filing, real date. They are also completely useless to you. Taking them out by hand is most of what you are paying for.

What it is not

It is not a research platform. If you need fifteen years of history and a team verifying records by hand at scale, Beauhurst does that properly, with roughly £10m of turnover and 128 people, and this does not pretend otherwise. The difference is depth and verification, not access: the public record is open to everybody. If you want every detected round in a database with a map for £100 a month, Funding Spotter does that, and this does not pretend to be alone in reading the register either.

It is not a contact database either, which matters more than it sounds. B2B contact data decays at about 22.5% a year, and the commonest complaint about every provider in that market is the same one: bounced email, wrong number, the job title from two roles ago. A dated public filing does not rot. A filing from last Tuesday is still a filing from last Tuesday in a year’s time.

Priorfile does one thing. It tells you which companies on your patch granted a charge, had one satisfied, or changed hands, who the directors are, and where the filing is. On Monday.

The limits, before you ask

• United Kingdom only. Companies registered at Companies House. No overseas coverage and no plans for any.
• Paper filings arrive late. Between 15% and 19% come in on paper and land in the following week’s register rather than being backdated quietly.
• Shells and holding vehicles are excluded — group structures, property SPVs, dormant companies. They file the same forms and they are not businesses you can lend to.
• No legal conclusions, ever. A row says control crossed a threshold and links the filing. It never says whether a transaction was notifiable, lawful or cleared.
• Anything implausible is held back for review rather than published automatically. A wrong row that looks impressive is still a wrong row.
• Filers make mistakes. If a line looks wrong, email me. It gets checked against the source and corrected in the next edition, with a note saying so.

Where the data comes from

All of it from Companies House under the Open Government Licence v3.0, which permits commercial reuse provided the source is credited. That credit is on every page here and in every edition.

Talk to me

Email sam@priorfile.com or find me on LinkedIn. It is one person, so that address reaches me and not a queue.

And if you think the whole premise is wrong — that timing is not the constraint, that knowing early changes nothing — that is genuinely the most useful email you could send. I spent long enough on a phone to know the difference between someone being polite and someone telling you something true.