adhouse / How we work
Evidence first. Then the work.
Advertising platforms optimise towards whatever they’re told. When the tracking is wrong, they get better at the wrong thing. So before we recommend anything, we find out what is really being counted, and show you how we know.
From the outside in.
The same order every time, because each step decides what the next one needs to check.
Read the public record.
Google’s and Meta’s ad libraries show which adverts are live, for how long, and where each one sends people. Anyone can read them. Almost nobody does.
Load the pages in a real browser.
We record every request your pages make: which tags load, which fire and which never appear. Reading the page code misses tags that load later, so we watch the page run.
Grade every fact.
Seen, inferred or unknown, and dated. Old readings are checked again before use, and anything contradicted is retired for good.
Go inside the account, with permission.
Read-only access answers what the outside can’t: what counts as a result, what each costs, and which enquiries became customers.
A person decides.
The founder reads the evidence and signs off every recommendation. The tools are AI. The judgement is human.
Observed, open, or to test.
Every finding we write carries one of three marks, so you always know how sure we are.
Observed
We saw it happen, or saw it not happen, in a real browser or in your account. We say when, and where.
Open
Something the outside can’t show, like what happens after an enquiry is sent. We name the question and where the answer lives.
To test
Anything that needs a test enquiry or a change to confirm. We agree it with you first.
Absence is never assumed
A tag missing on one visit doesn’t prove there’s no tracking. We say what loaded and what didn’t, and never stretch it further.
The engine behind it.
We built our own research engine, so the slow, careful part of the work is done before anyone opens an account.
It reads the public ad libraries, finds the company behind each advertiser on the Companies House register, loads their pages in a headless browser, and files what it sees as graded, dated facts. Findings are chosen from those facts by fixed rules, then read by a person. Anything that fails a check doesn’t go out.
By September 2026 it had researched 414 UK advertisers and recorded 9,678 facts across 26 kinds of check. It is also where our published research comes from.
Start with a conversation.
Twenty minutes, free, about your business and your advertising. You’ll know by the end whether we can help.