Not a bought list. The engine reads the ARB statutory register, works out which practices are genuinely one-person, then reads each practice's own website to see what work they actually sell. Every number on this page came out of a real run.
In the UK it is a criminal offence to call yourself an architect unless you are on the ARB register. So the register is not a list of architects, it is the list. The engine sweeps all of it, then narrows in four passes.
"Under £10k a month" is not a field you can buy anywhere. But there is a proxy hiding in plain sight, and it is exact.
Corroborated against Companies House: micro-entity accounts (turnover ceiling £632k, in practice far lower), single active director, and years trading.
The engine crawls each practice's own website and scores the language. A practice only qualifies on evidence, and the evidence travels with the lead, so every row can be checked in one glance.
UK PECR treats sole traders as individual subscribers, and cold email to them needs consent, not just legitimate interest. Limited companies are corporate subscribers, where B2B email is lawful with a clear opt-out. Since the whole point of this list is one-person practices, that distinction decides the campaign. Every lead is checked against Companies House and gated before it reaches you.
Contact details are taken preferentially from each practice's own published website rather than the register itself, because the register's terms restrict commercial reuse. The email_source field records which route every address came from.
A 600-row slice of the 0 qualified practices, sampled across all four send tiers so the mix you see is the mix that exists. Ranked by sendability, then fit. Every row is real.
| Practice | Contact | Where | Advertises | Size | Send? |
|---|
Full export carries 36 columns including the evidence sentence, PECR class, mailbox type, accounts category and a personalisation-safety flag.
Drag the assumptions. The engine has already done the expensive part, these are the only numbers left to argue about.
The list regenerates. The register updates continuously, so the engine can be re-run on a schedule and only surface practices that are new since last time.
A qualified lead is a name and an email. What turns it into a conversation is knowing what this specific practice is probably struggling with, and having a story that matches. The agent reads their site and their socials, then writes a profile and a full nurture sequence off it. Below is a real one, from their real website.
The system reads a stranger's website and writes to them in your name. So every profile and every sequence lands in an approvals queue, and approving one only moves the lead to queued. Sending stays a separate, deliberate action.
The compliance tier from section 04 is enforced here too, not just displayed. A tier C practice is a sole trader, so the server refuses to record an email to them at all and tells the rep to use LinkedIn instead.
This is not a mock-up. The pipeline is six Python modules that swept 28,296 registrants, crawled 5,529 websites and cross-checked 2,595 practices against Companies House. Every page fetched is cached, so re-scoring the whole list against new rules takes seconds rather than another crawl.