Why your lead list is mostly wrong
We built a lead engine from the public register, ran it, and found that two thirds of what it produced was rubbish. Here is exactly how, because a bought list has the same faults and nobody shows you them.
- Our first real run produced 82,171 leads, of which 54,053 were false positives — two thirds of the list.
- The cause was SIC code matching without keyword confirmation. Code 68209 alone produced 21,489 "estate agents" that were buy-to-let holding companies.
- A second fault matched keywords against the SIC code's own description text, so every property developer became a builder on the word "building".
- A third resolved company websites by guessing the domain from the company name, then verified it by checking the domain matched the name — a circular test that proved nothing.
- A bought list has all three faults and no way for you to see them, because you did not watch it being built.
We built our own lead engine rather than buying data. It pulls the Companies House monthly snapshot — five and a half million companies — filters to Surrey postcodes, matches target sectors, checks subscriber type for PECR routing, and screens numbers before any call.
The first real run produced 82,171 leads. It looked like a triumph.
Then we looked at it properly. Two thirds of it was rubbish, and the way in which it was rubbish is worth publishing, because a list you buy has the same faults and arrives with the evidence stripped out.
82,171 leads. 54,053 of them wrong. And every one looked perfectly plausible in a spreadsheet.
Fault one: the estate agents who were not estate agents
The list claimed 38,530 estate agents in Surrey. There are not 38,530 estate agents in Surrey. There are not 38,530 estate agents in Britain.
Of those, 21,489 came from a single SIC code: 68209, other letting and operating of own or leased real estate. That is the code every buy-to-let vehicle, every property holding company and every landlord's limited company in the country files under. Almost none of them is an agency. Almost none of them has a website, a marketing budget, or any interest whatsoever in a new one.
This is the fundamental weakness of SIC-based targeting, and it applies to every list sold on that basis. The codes are self-selected at registration, never audited and rarely updated. A firm can trade as a kitchen fitter for fifteen years under a code that says something else, and there is no mechanism that corrects it.
Fault two: Companies House writing our keywords for us
This one was ours, and it is the more instructive of the two.
The classifier matched keywords — "kitchen", "roofing", "dental" — against the company's details. The bulk file writes SIC codes as a number followed by an official description: 41100 - Development of building projects.
We matched keywords against that whole string. Which meant that the word "building", sitting inside a description written by Companies House, made every property developer in Surrey a builder. Ten thousand of them. And 98000 - Residents property management matched the estate agent keyword "property management", so every residents' association in the county became a prospect.
The top result in the list was GAINSBOROUGH PLACE RESIDENTS COMPANY LIMITED. Twelve leaseholders in Leatherhead who share a freehold and a lawnmower.
Fault three: the websites that belonged to somebody else
This is the one that would have done real damage, and it is worth describing precisely because the same shortcut is everywhere in this industry.
To find a company's website, the engine generated likely domains from the company name — smith-builders.co.uk, smithbuilders.com — and fetched whichever answered. Then it verified the result by checking whether the domain resembled the company name.
Read that twice. We built the domain out of the name, and then confirmed it by asking whether the domain looked like the name. Of course it did. The test could not fail, and so it proved nothing at all.
Of 254 websites the engine reported finding, 9 had real evidence — the company number printed on the page, as a registered company is required to display. The other 245 rested on the circular check. Re-verified honestly, 34 stood up.
And this matters more than a bad column, because our outreach opens with a fact about the recipient's own website. A wrong domain means telling a builder in Woking, by name, on the phone, about a stranger's site.
Why a bought list is worse, not better
Every fault above exists in commercial lead data. The difference is that we could see ours, because we watched it being built.
A purchased list arrives as a clean spreadsheet. The bad rows look exactly like the good rows. There is no provenance column, no confidence score, no note saying this website is a guess. And the incentive runs the wrong way: the vendor is paid for volume, and removing two thirds of the file makes it worth a third as much.
There is also the compliance dimension. You are the controller for that data the moment you use it. If the ICO asks where it came from, "we bought it" is not an answer that helps you.
What to do with the list you already have
- Sample thirty rows at random. Not the first thirty — those are often curated. Look each one up properly. Whatever error rate you find in thirty is roughly the error rate in all of it.
- Check the sector claim. Does this business actually do the thing the list says? Very frequently it does not.
- Check the website. Open it. Is it that company, or a company with a similar name?
- Check it is still trading. Dissolved companies are abundant in old data and calling one is a wasted dial at best.
- Check subscriber type before you email anybody. Sole traders and ordinary partnerships are individual subscribers and generally cannot be cold emailed at all.
It is a dull afternoon. It is considerably duller than opening a cold call by describing the wrong company's website to somebody who has run their own for nine years.
Questions we get asked
Why are bought B2B lead lists so inaccurate?
Because they are usually assembled by matching industry classification codes, and those codes are self-selected by companies at registration and never corrected. Broad codes like "other letting and operating of own or leased real estate" catch tens of thousands of holding companies that share nothing with the sector you actually want. Nobody removes them, because volume is what is being sold.
What is a SIC code and why does it mislead?
A Standard Industrial Classification code is the category a company selects when it registers with Companies House. It is self-declared, rarely updated, and includes very broad catch-all codes. A company can trade as a kitchen fitter for fifteen years under a code that says something else entirely, and nothing in the system corrects it.
How do I check whether a lead list is any good?
Take a random sample of thirty rows and look each one up yourself. Not the first thirty, which are often the best. If more than a handful are the wrong sector, are dissolved, or have a website belonging to somebody else, the whole list has the same rate of error — you have just found it early.
Is it better to build a lead list or buy one?
Build it, for two reasons. You can evidence where every field came from, which matters if the ICO ever asks; and you can see the errors and fix them. A bought list hides its faults, and its faults become yours the moment you are the one contacting people.
What is the most dangerous kind of bad lead data?
A wrong website. If your outreach opens with a fact about the recipient's website and the domain belongs to a different company, you have told a business owner something confidently untrue about their own business. It is not a bad row in a spreadsheet — it is the end of that conversation.