Claims Bible, 2026

Running compliance on a regulated site, and correcting our own headline figure downwards

Claims Bible operates under an FCA authorisation for claims management, which means every page on the site is a financial promotion. The personal injury work sits under the referral fee ban in the legal aid legislation. The email marketing sits under the privacy regulations and UK GDPR. There’s a lot of it, it changes, and the site had five years of content written before anyone was keeping a proper record of what had been checked against what.

41 numbered entries
in the register, each separating fact, recollection and ruling.
CR-01 to CR-41
£18.3m to £11m
the headline recovery figure, corrected downwards on evidence in writing.
Leigh Day, in writing
217 edits
across 126 files to correct a regulatory rate, verified by a dry run reporting zero left.
FOS rate change, 2026
57 pages
carrying an unsupported legal citation, found and taken down.
Site audit

I co-founded Claims Bible, so I’m not writing this as someone who was brought in to tidy up. I’m writing it as the person who’d let it drift and then had to fix it, which I think is the more useful perspective.

Where it started

There were figures on the site that nobody could substantiate. The headline claim was that our partner firms had recovered over £18.3 million for the people we’d introduced, and when I went looking for where that number came from, there wasn’t anything.

There were partner names on signed consent forms that didn’t match any company that existed. 257 consents named a limited company that had never been incorporated. Another 177 named a firm of solicitors under a name it had never traded under. And one “partner” turned out, on investigation, to be a mailbox.

There were three different versions of the consent wording in use at the same time, none of which had a date on them. And there was a citation to a legal guideline on 57 pages of the site that, as far as anyone could tell, had never actually been read by whoever put it there.

None of this was anyone being dishonest. It was five years of content, written quickly, by people who’d moved on, with nobody keeping the register.

What I built

A numbered register. Every compliance question gets an entry, CR-01 upwards, and there are 41 of them at the time of writing plus a few still open. Each one records what the question is, what we know to be fact, what’s somebody’s recollection, and what the external compliance consultant has actually ruled, and those three things are labelled so you can’t mistake one for another. It runs to about 123KB and it’s the most useful document in the business.

One rule. No compliance question gets answered from reasoning. Either I can point at the rule it comes from, or it goes to the consultant, and the register records which. It slows things down and it’s stopped me being confidently wrong more than once.

A consent standard. Separate boxes for each thing being consented to, none of them ticked in advance, the exact wording stored with the record, plus a timestamp, the page it was captured on, and where the person came from. If a form can’t meet that standard it doesn’t go live. One data capture on a sister site couldn’t be made to meet it, so I switched it off rather than patch it.

Switches, not edits. A whole claim category can be taken out of solicitation in one build, so that when the answer to a question is “stop promoting this until we’re sure”, that’s a single change rather than 150 pages and a prayer. That’s covered properly in the site rebuild case study.

Tooling for corrections at scale. When the Financial Ombudsman changed its default interest rate at the start of 2026, the site still said 8% on 129 pages, including inside letter templates people were sending to their own banks. I wrote a script with a dry-run mode, checked what it was going to change, and then let it make 217 edits across 126 files and three templates, with eight pages rewritten by hand where the context needed it. The wording now says “base rate plus 1%” so it can’t go stale the same way again, and the dry run reports zero remaining.

The decision I’m most pleased with

The headline figure. The site said our partners had recovered over £18.3 million, and I couldn’t prove it, so I went to Leigh Day, the firm that had handled the biggest part of it, and asked for the real number in writing.

It came back at £11.3 million. So I changed the site, in the five places the figure appeared, to say over £11 million.

Bar chart. The headline recovery figure as published, £18.3 million, beside the figure confirmed in writing by Leigh Day, £11.3 million. The site now says over £11 million.
The headline figure before and after. The left bar had no source behind it. The right one is Leigh Day’s own number, in writing, and it’s the one the site uses now.

Nobody asked me to make our own number smaller. It would have been easy to leave it, because it’s the kind of claim nobody ever checks. But a figure I can’t evidence isn’t an asset, it’s a liability waiting for someone to ask, and I’d rather have the smaller number with a letter behind it than the bigger one with nothing.

There was a second moment like that. The external consultant signed off the whole site on the day it relaunched, and a couple of weeks later I found that the editorial policy he’d read included a claim about sources that turned out to be unsupported. The easy thing was to quietly fix the page. The right thing was to fix the page and then tell him the sign-off had rested on something that wasn’t true, so he could decide whether it changed anything. That’s what the register is for. It makes it harder to quietly fix things.

The result

  • The headline recovery figure corrected from £18.3 million to £11 million, evidenced in writing.
  • Three partner entity names that didn’t exist identified and corrected across 434 signed consents and every page that referred to them.
  • An unsupported legal citation found on 57 pages and taken down before anyone outside the business relied on it.
  • One consent standard, in use on every form, with every capture evidenced.
  • 217 edits to correct a regulatory rate across the site, verified by a script that reports zero left.
  • The whole site reviewed and signed off by the external compliance consultant on the day it relaunched.
  • 19 items open on the register at one point, each with a named owner. Which is rather the point. A register isn’t something you finish, it’s something you keep.

How it was built

The register is a document, and I wrote it. The tooling around it, the correction scripts and the consent capture, was built the same way as the rest of the site, specified and tested by me with the code written alongside Claude. What I’d say about that here is that the tooling was never the hard part. The hard part was deciding that a number on our own homepage needed checking.

What this looks like for your business

If you’re a law firm, a claims company or a financial business, the question isn’t really whether you’re compliant. Most people are, more or less, most of the time. The question is whether you could show it, this afternoon, if someone asked. A register, an adviser, a standard for consent, and the discipline of not answering from reasoning will get you most of the way there, and it costs a great deal less than finding out the other way.

Tell me what’s not working.

I’ll tell you whether I can help and roughly what it would involve.

Get in touch