Read from the filing. Checked, named, nothing added.

The legal starting point

Two duties do the work.

Charges. When a company creates a charge over its assets, the Companies Act 2006 requires it to be registered at Companies House within 21 days of creation. When the debt is discharged, a statement of satisfaction is filed. Both are public, both are dated, and both name the lender holding the security.

Control. A company must enter a change of person with significant control on its own register within 14 days, and tell Companies House within a further 14. Significant control is defined at three thresholds: more than 25%, more than 50%, and 75% or more.

What Priorfile does with it

1. Reads the filings as they arrive. The Companies House stream runs continuously rather than in a nightly batch, which was measured rather than assumed: roughly 11 filings a minute late at night and around 207 a minute on a weekday morning, every one dated the same day it appeared. The position is replayed from the last known point if a run is missed.
2. Narrows to the sector. 105,036 active UK manufacturing and engineering companies, or the defence and advanced advanced manufacturing SIC codes, or the watchlist you supply, and nothing else.
3. Separates the events. A charge registered, a charge satisfied, a controller arriving, a controller leaving. A satisfied charge and a new charge on the same company within weeks of each other is a refinance, and it is read as one.
4. Names the people. The officers register gives the active directors, the PSC register gives the controlling person, and the charge itself gives the holder. All three are public and current by law, so each row carries a name rather than a switchboard.
5. Removes what is not an operating business. Shell companies, dormant companies, holding vehicles and special-purpose vehicles all file the same forms. In one measured week, 96 of 284 rows were not operating businesses. They come out by hand.
6. Publishes on Monday with a link to the filing behind every line.

The traps, and how each is handled

These are the reasons a naive read of the same data produces wrong rows.

A charge is not an amount

A registered charge names who granted it, who holds it and when. It does not reliably state how much was lent. Any register that prints a facility size next to a charge is inferring it, and this one does not.

A satisfied charge is not always a repayment

It can be a refinance, a sale of the secured asset, or the release of one security among several while the borrowing continues. The row reports that the charge was satisfied on a date and links the filing. It does not tell you why, because the filing does not say.

A controlling person is often not a person

Control frequently passes to another company or to an overseas entity rather than to an individual. Reading a corporate PSC as a named human puts a company name in a column meant for someone you can call. The row says which it is.

Shells and holding vehicles

The commonest cause of a useless row. A holding company with no trade files charges and control changes exactly like a manufacturer. Filing frequency, accounts type, employee count and name patterns all feed the exclusion, and the last pass is done by eye.

Thresholds are not legal conclusions

The statutory control thresholds line up with the ones used elsewhere in UK law, which makes it tempting to draw a conclusion from a crossing. Priorfile never does. A row reports that control crossed a threshold on a date and links the filing. It never states whether a transaction was notifiable, lawful or cleared. That is a question for the reader and their advisers.

Paper filings

A minority of filings arrive on paper and appear days later than electronic ones. They are added to the following week’s register rather than backdated silently.

What you get

One email, Monday morning, for your watchlist or your sector. Company name and number, the event, the date it was filed, what changed, the director, and the filing link. That is the whole product.

Accuracy, with its denominator

Fourteen rows from the sample register, across control changes and newly registered charges, were checked by hand against the source filings. Fourteen of fourteen matched. That is a small denominator and it is quoted with it: it is not a claim of perfection, and it does not put a useful bound on the error rate on its own. Any row that looks implausible is held back for review rather than published.

Source and licence

All company data comes from Companies House under the Open Government Licence v3.0, which permits commercial reuse with attribution. Priorfile is not affiliated with or endorsed by Companies House.