Screening for acquisition targets from public filings
A practical method for building an acquisition target list from registry data and company websites, and the filters that remove the noise before it costs you a week.
The companies most worth acquiring are frequently the hardest to find. They have never raised, never sold, and never appeared in a deal database, because nothing about them has ever been a deal. They exist in a statutory register and on their own website, and that is the whole of their public footprint.
Which means a target screen has to be built from the register up. This is the order that wastes the least time, and the order matters because each step is cheaper than the one after it.
1. Write the thesis down as filters, not adjectives
Before touching data, convert the thesis into criteria a spreadsheet can apply.
“Founder-owned industrial services businesses in the North of England, big enough to matter and small enough to be off the radar” is a sentence. The filters are: jurisdiction, active trading status, an activity definition, a geography defined by trading address, and a size proxy. Vagueness at this stage does not stay vague — it reappears later as a list nobody can defend.
Be explicit about what would disqualify a company too. A recent change of control, an insolvency event, a parent that is itself a fund. Exclusions are as much of the thesis as inclusions and they are much easier to apply early.
2. Filter on status before anything else
Take the registry population and remove everything that is not a going concern: dissolved, struck off, in liquidation, and — the important one — dormant.
This is the single highest-yield filter in company screening and it costs almost nothing, because status is a reliable field on any decent register. It routinely removes a large share of the raw population, and it removes it on a factual basis rather than a judgement call.
Do it first for exactly that reason. Every filter after this one is more expensive per row.
3. Treat the industry code as a starting point, never as the answer
The filed industry code is the obvious next filter and the one most likely to mislead you. It is chosen at incorporation and almost never updated, so it describes what a company was, filtered through whatever code seemed closest at the time.
Use it to cast a wide net, then correct it against what the company actually says it does. In practice that means accepting a broad set of codes, and expecting to remove companies from inside that set rather than to have caught them all.
The companies you will miss are the ones whose business has moved on. Those are often the most interesting names on the list, which is why a code-only screen quietly selects against the targets you most wanted.
4. Screen geography on the trading address
Two addresses, not one. The registered office is a legal address for service and for smaller companies it is very often their accountant’s, which means a screen run on registered office will produce a target list clustered around professional services firms rather than around the actual industry.
The trading address comes from the company’s own website, or from an operating site named in its filings. Where you only have the registered office, record it as the registered office and treat the geography as unconfirmed.
5. Use size proxies, and label them as proxies
Where full accounts are public, use them. Where they are not — which is the normal case for private companies, because small-company exemptions apply nearly everywhere — you are working with proxies:
- Balance sheet size from abridged filings, where they exist.
- Whether the company files under a small or medium regime, which is itself a size band.
- Employee numbers, where the jurisdiction requires or the company publishes them.
- The scale the company’s own website implies: sites, fleet, named locations, client roster.
Any of these is defensible. What is not defensible is converting them into an estimated revenue figure and then screening on it as though it were filed. Keep the proxy visible in the sheet so that when someone asks why a company made the cut, the answer is a fact rather than a model output.
6. Confirm the company is actually trading
Active status on a register plus one independent signal. A live website with current content, a recent non-dormant filing, a licence, a recent contract award, a job posting. One is enough; none is not.
This step is where a hand-built list stops resembling a registry export, and it is the step that is skipped in every bulk file you will ever be sold.
7. Resolve the group before you contact anyone
Where a business sits under a holding company, an operating company and a property entity, all three appear as separate rows and only one of them is the business. Decide which entity trades, which one owns, and which one you would actually approach — then record all three and mark which is which.
Skipping this produces the two most visible failures in outreach: contacting the same business three times under different names, and addressing a letter to a dormant holding entity.
8. Keep the evidence with the row
For every company, one line: the filter it satisfied and where that came from. “Active, non-dormant, filed under industrial cleaning, website describes contract cleaning for food manufacturers, trading address Greater Manchester, filed as a small company.”
It takes seconds per row and it does three things. It survives a challenge from an investment committee. It survives the person who built the list moving on. And it makes the screen re-runnable next quarter instead of disposable.
What good looks like
Not length. The test of a target screen is whether you can say, for any company on the list, which criterion it met and what the source was — and for any company excluded, which filter removed it.
A few hundred confirmed trading companies you can defend is a working longlist. Ten thousand registry rows is a research task you have not started yet.
Frequently asked questions
How long should an acquisition target longlist be?
Long enough to survive attrition and short enough that every name can be justified. For a focused buy-and-build thesis in one country, a longlist of a few hundred trading companies narrowing to a few dozen approaches is normal. A list of ten thousand registry rows is not a longlist, it is an unfiltered export.
Can you screen on revenue if most companies do not file it?
Not directly, and pretending otherwise is where most screens go wrong. Where full accounts are not public, use the proxies that are — filed balance sheet size, whether the company files as small or medium, employee bands where disclosed, and the scale implied by the company's own website. Record each as the proxy it is.
Is it worth screening companies that have never done a deal?
That is usually the entire point. Companies that have transacted are already in every deal database and every competing buyer's list. The quiet, never-sold, owner-managed business is harder to find and is where the return on the research actually is.