What We Got Wrong: Errors in Our Own Research, and the Checks They Taught Us

Data last verified: 1 November 2026. Figures re-checked against their primary sources under the Whito rolling re-verification programme.

Last Updated on November 1, 2026

Every research publisher makes mistakes. Most fix them quietly and move on. We think the mistakes are worth more than that. They show how a figure goes wrong between a source and a page, and that is the thing a reader most needs to know before trusting any statistic, ours included.

This page lists real errors from Whito research and news since August 2026. Some we published. Some we caught just before. For each one we say what went wrong, how it was found, and the check we now run because of it. Nothing here is a hypothetical example.

The pattern in one line

When we checked our UK trades statistics hub in August 2026, we ran a separate review, set up to try to break the 20 claims the page leaned on most rather than confirm them. 15 held up. 5 were misstated. None had a wrong number. Every failure was in the base, the label or the source: who was asked, when, and where the figure really came from.

That has held across almost everything below. The numbers survive. The descriptions rot. A percentage gets copied correctly, but the date range, the country or the sample size attached to it drifts from page to page until it describes something the original study never measured.

The errors

1. A regulator’s deadline is not your deadline

What we published. In August 2026 our guide to the new vet pricing rules said practices “must comply by 23 September 2026”.

What was true. 23 September 2026 was the date by which the Competition and Markets Authority (CMA, the UK competition regulator) had to make its final order. It was the regulator’s deadline, not the practices’. The British Veterinary Association said practices would have until early 2027 to comply, and smaller practices get longer than large groups.

How it happened. A research summary was trusted without opening the CMA and gov.uk pages it was based on.

What changed. Any date attached to a law or a regulator is now read from the regulator’s own page and the relevant professional body before it goes live. Never from a summary, a news report or a tool’s answer.

2. The date a newspaper wrote it is not the date it happened

What we published. A news post said Pizza Hut announced its temporary “HUT” name on 21 August 2026.

What was true. 21 August was when a US newspaper reported it. Pizza Hut’s own post was dated 19 August.

How it was found. A second, separate fact check run after publishing. The sources were right. The date was taken from the wrong one.

What changed. When a date comes from a news article, we find the underlying announcement (the company post, press release or filing) and date it from that.

3. A US date range on a UK figure

What we published. Our UK SEO statistics page said SparkToro’s analysis found 69.5% of UK Google searches ended without a click “between January and April 2026”.

What was true. The 69.5% UK figure is correct. But SparkToro’s method section gives the January to April 2026 range for its US sample only. It does not state a period for the UK. We had attached one country’s date range to another country’s number.

How it was found. While building a separate statistics page on 29 August 2026, we checked the same study again and saw the “(US)” label next to the dates.

What we did badly. We logged it and did not correct the page until 1 November 2026. Our own policy says five working days. That was too slow, and it is the reason this page now exists: a list we can see is harder to leave unfinished than a note in a file.

What changed. When a study covers several countries, we check which sample each date, base and method note belongs to before quoting any of them.

4. The near miss: a rate that had not moved

What nearly went out. For a post on late payment interest, a search summary said the Bank of England base rate had changed in September 2026. If true, the 11.75% interest figure in the post would have been wrong.

What was true. The rate had not moved. More importantly, the legal interest rate on late invoices uses the base rate on a fixed reference date (30 June for invoices that go late from July to December), so a later change would not have altered the figure anyway.

What changed. We never quote a statutory interest rate without naming the reference date it rests on.

5. The empty box

What we found. During a Search Console review in September 2026, the page on whether fake reviews are illegal in the UK had an empty key finding box. The field that feeds it was blank, so the page showed no headline finding at all.

What changed. The finding was restored from facts already on the page. Our reviews now check that the finding box shows text, not just that the page loads.

6. Our sentence disagreed with our table

What we published. Our emergency plumber study said: “In five of these eight towns, fewer than four firms publish any figure at all.”

What was true. The table on the same page showed six towns where fewer than four firms published a price in pounds: Belfast 2, Birmingham 3, Cardiff 2, Glasgow 2, Leeds 3 and Norwich 1.

What changed. The sentence was corrected on 30 October 2026 with a dated note on the page. Every count in a sentence is now recounted from the table it summarises, not from the draft notes.

Two results that look like errors and are not

When our figure disagreed with the regulator

The CMA found that fewer than 40% of vet practices published prices. When we checked 72 practice websites in August 2026, 49 of the 68 we could assess (72%) did. We did not report that as the CMA being wrong. Our sample was found through web search, so it leaned towards practices with better websites, and our check was five months later, after many practices had started publishing. We said so on the page. A different result is only a finding once you have ruled out that you measured something different.

Why 52 claims became 49

In our October 2026 check of what AI assistants say about home care agencies, we found 52 claims about a named agency’s registration or rating. We could match 49 of them to a record on the regulator’s register. 42 of those 49 matched exactly. The 3 we could not match are not counted as wrong, because failing to find a record is a fact about our search, not about the agency. Our headline uses 49 for that reason. Both numbers are on the page.

The checks we run now

• Base, label, source. For every figure we quote from someone else: who was asked, in which country, when, and is this the original publisher or a copy.
• A separate review that tries to break the claims. It runs apart from the drafting, and its job is to find faults, not to confirm. A right number with the wrong base counts as a failure.
• Regulator dates from the regulator. Never from a summary.
• Sentences recounted from tables. Every count in the text is checked against the table it describes.
• What failed is published. Our larger statistics pages carry a “do not claim” list of figures that did not survive checking, so readers can see what we left out and why.
• Corrections are dated on the page. We do not quietly edit. Our corrections policy explains how to tell us about an error, and what happens next.

Found one we missed?

Email hello@whito.co.uk with the page and what is wrong. If we are wrong, we correct it within five working days, note the date on the page, and add it to this list.

Related: editorial policy, data and sources, the Whito Truth Index.

Jacob Whitmore

Whito is an independent UK research publisher. We publish research on how UK businesses are found, trusted and priced: what they cost to market, how their sectors work, and who actually gets recommended. Studies are free to read, every figure carries the date it was checked, and nobody pays to appear.

See Full Bio