One supplier website in six on our directory no longer leads to that supplier

One supplier website in six on our directory no longer leads to that supplier

We drew 1,000 company listings at random from our own directory this morning and opened the website each one gives. For 171 of them, the address no longer leads to the business the listing describes. Eighty-nine lead nowhere useful at all: the domain has gone, is parked for sale, shows a suspended-hosting page or redirects to something unrelated. Forty-one lead to a business with a different name — the firm has been renamed, merged or bought. Another 41 lead to the same firm at a new address, which is harmless for a reader but still means the address we show is out of date.

That is our own catalogue, and the finding is about us before it is about anyone else. It is also the most practical thing we can tell a buyer about any supplier directory, including this one: the website on a listing is a claim that was true on the day someone recorded it, and roughly one in six of those claims has since stopped being true.

What we checked, and how

On 24 September 2026 our directory held 87,203 live company listings, of which 80,596 show a website. We drew 1,000 of those at random, excluding the single listing that belongs to our own parent company, and ran the same check on each.

First, the domain. We asked for its address record, and where that failed we asked a second, public resolver, so that a local hiccup would not be counted as a dead domain. Then the page. We requested the stored address, followed every redirect to wherever it finally landed, and read what came back: the status, the final address, the page title and the text. A page counted as serving the business only if it answered at the same domain we hold and the business's name appeared on it. Anything that failed on the first pass was tried again, more slowly and with a longer wait, and the second answer is the one we kept.

Two things were done by hand rather than by rule. Every one of the 93 addresses that now lands on a different domain was read and classified individually, because a machine cannot tell a rebrand from an acquisition from a hijacked domain, and those are different findings. And every page that looked like a domain marketplace or a hosting company's placeholder was checked against the page itself, not just against its wording.

The checker identified itself in every request, made one pass per site plus one retry, and did not log in to anything or submit anything.

What we found

Of the 1,000:

• 678 answered at the address we hold with a page that names the business. That is the only group we can say is simply right.
• 51 answered, but behind a bot-protection screen that an automated check cannot pass. The site exists; we could not read it. We count these as alive, not as confirmed.
• 41 redirect to the same business under the same name at a new address — a new domain, a new country suffix, a switch from one ending to another.
• 41 redirect to a business with a different name. Some are straightforward renames. Many are the visible end of an acquisition: an address that used to belong to a small consultancy now opens the home page of the firm that absorbed it, and in one case the new owner's page says so in as many words.
• 89 lead nowhere useful. Sixty-two domains no longer resolve at all. Sixteen are parked, most of them on a marketplace page offering the domain for sale. Five show a suspended-account or default hosting page. Six redirect to something unrelated or to nothing: a betting site, a lottery results page, a search engine's home page, a social-media profile, and two "not found" pages on other businesses' sites.
• 100 could not be confirmed either way. Forty-seven did not answer — timeouts, refused connections and ten certificate failures. Thirty-five answered with an error: six of those were a content-delivery network reporting that it could not find the site behind it, two were hosting subscriptions marked expired, and twelve more were "not found" pages, some of them on sites that otherwise still work, where the stored address points at an inner page that has since moved. Thirteen domains failed their lookup twice with a temporary error, and five redirects could not be judged from the page.

Add up the three groups that are certainly not the business as listed — gone, renamed or moved — and it comes to 171 of 1,000, or 17.1 per cent. On a random sample of this size that figure is accurate to about 2.3 percentage points either way, so across the 80,596 listings that show a website, somewhere between roughly 11,900 and 15,600 addresses no longer lead to the firm we name. If the 100 we could not confirm behave even partly like the rest, the true figure is higher, not lower.

The narrower number — the addresses that lead nowhere, or lead to someone else — is 130 in 1,000. Those are the ones a buyer could be misled by, rather than merely inconvenienced.

Where an address goes when the firm has gone

The 41 successor redirects are the most useful group to read, because they show what actually happens to a supplier over a few years. Some of the names are well known. The address we hold for Gerber Technology now opens Lectra's site. Cloudability's opens IBM's Apptio pages. Efecte's opens Matrix42. Vision Critical's opens Alida, and Platform.sh's opens Upsun. For a large software vendor this is familiar, and most buyers would already know.

The more interesting cases are the small ones, because nobody writes about them. Seven accountancy practices in the sample now lead to a different firm's site — most often a larger practice, in one case a national advisory firm, in another a finance staffing group. Among agencies, a studio's address opens a different agency's brand; among consultancies, an address opens the parent group's site rather than the business that was listed. None of those transitions is wrong. But a buyer who shortlists "the small practice with the specialist partner" and lands on a much larger firm's home page has learned something important about who would actually do the work, and would not learn it from the listing.

The dead end is a different problem. When a domain stops resolving, the firm may have closed, may have moved without keeping its old address, or may simply have let a registration lapse. We cannot tell which from outside, and we do not guess. What we can say is what happens to the domain next. Under the rules ICANN sets for generic domains, a deleted registration enters a 30-day redemption period during which the old owner can still recover it; after that the name is released and anyone can register it. Country domains have rules of their own, but the end is usually the same, and it is the route by which a name ends up parked, as sixteen in our sample were, most of them on a page offering the name for sale. It is also how two ended up serving something else entirely: a marketing agency's old address now opens a casino, and an accountancy practice's now opens a page of lottery results.

The age of a listing told us little

We expected older listings to be worse. They were not, or not in a way we could rely on. Of the listings first added to our directory in 2019, 12 per cent certainly no longer lead to the business as listed. For those added in 2020 the figure was 19 per cent; for 2022, 16 per cent; for 2023, 18 per cent. The 2023 group is the largest, because most of the catalogue was loaded that year, and it is the one we would have expected to be freshest.

The reason is that the date a listing was added is not the date its details were checked. A large import can carry information that was already years old when it arrived. That is true of our catalogue, and we would expect it to be true of any directory built the same way: collected in bulk, then displayed indefinitely. "Listed since" on a supplier page tells you how long the page has existed. It does not tell you when anyone last looked at the address on it.

The category did seem to matter, although the numbers in each category are small enough that we would not publish them as rates. In our sample, app development companies had the most dead domains: 19 of 92. Digital marketing agencies had the fewest: 3 of 68. Design agencies were closer to the app developers, accountancy firms to the marketers. The pattern is plausible — small software studios form and dissolve quickly, and marketing agencies depend on being findable — but it is a pattern in a sample of a few dozen per category, not a measurement.

Why a stale address matters more than it looks

A directory entry with a dead website looks like a small embarrassment. For a buyer it can be three different problems.

It can hide the real counterparty. When an address redirects to a successor, the business you would be contracting with is not the one described on the listing. Its size, its terms, its data-protection arrangements and the people who would do the work may all have changed. A redirect is the cheapest possible evidence of that change, and it is easy to miss because the browser does it silently.

It can hand an impersonator a ready-made identity. A domain that lapses and is re-registered by someone else comes with the old firm's reputation attached. Anyone who still has that firm in their supplier records, email allow-lists or accounts-payable system is now trusting a domain that belongs to a stranger. Nothing in our sample suggests any of these domains is being used that way. But sixteen of them were parked on the morning we checked, most with a price on them, and the price of one is the price of an identity a buyer's systems may already trust.

It weakens every other check. Most supplier due diligence starts from the website: the legal entity in the footer, the address on the contact page, the certificates and client logos. If the website is someone else's, every one of those checks is checking the wrong business.

A ten-minute check that catches it

None of this needs a tool. Before you shortlist a supplier from any directory, ours included, it is worth ten minutes.

1. Type the address yourself and watch where you land. Do not follow a link in an email or on a listing. When the page loads, read the address bar. If the domain is different from the one you typed, you have been redirected, and the next question is why.
2. Find the business's own name on the page, and then its legal name. The trading name should appear on the home page; the legal entity should appear in the footer, the terms or the privacy notice. It should match the company you are about to contract with. A mismatch is not proof of anything, but it is a question to ask before signing, not after.
3. Look up the domain's registration. ICANN's free lookup service shows when a generic domain was created and, often, when it was last renewed. A domain registered last spring for a firm that describes fifteen years of history is worth a phone call.
4. Check the company register. In the UK, Companies House will tell you whether the company is active, what its registered name and address are, and whether it has recently changed its name. Compare that with what the website says. Most countries have an equivalent, and most are free.
5. When an address leads to a successor, find the successor's own statement. Search the new firm's site for the old name. A firm that has bought another usually says so — "now part of", "has joined", "has acquired". If it does not, ask it directly who would hold your contract.
6. Check that email and website agree. Proposals and invoices should come from the same domain the website uses. If a supplier you already pay writes from a new domain, or tells you its bank details have changed, confirm it by phone on a number you already hold, not one in the message.
7. Repeat it for suppliers you already have. A supplier review that checks certificates and insurance once a year can check the website in the same sitting. Remove dead domains from allow-lists and supplier records when you find them, rather than leaving a lapsed domain trusted by default.

None of these steps depends on us, and together they would have caught every kind of problem described above.

What this says about our own listings

This is a measurement of our catalogue, published before we have fixed it, because the number is useful to a buyer now and would be less honest if we cleaned up first and reported the result.

Some context the reader is owed. Earlier this month we found that an old import had attached thousands of contact records, websites included, to the wrong companies, and we removed or corrected those links on 18 September. The listings in this sample are the ones that still show a website after that repair, so the 171 are not that problem again: they are addresses that were right when recorded and have gone stale since.

One listing in the directory belongs to Rutba, our parent company. It was excluded from the draw, its site answered normally when we checked, and leaving it out changes none of the figures above.

Nothing in the sample has been changed yet. The addresses that now lead nowhere, and the ones that lead to a successor, are the obvious next step, and how we treat a listing whose business may no longer exist in the form described is a decision we will make deliberately rather than by script. When it is done we will say what we did.

The limits of this count

A thousand listings is enough to put the headline figure within a couple of points, but not enough to rank categories or countries, and we have not tried to.

The check ran from one place, on one morning. A site that was briefly down, or that blocks traffic from some regions, will have landed in the group we could not confirm, which is why that group is reported separately rather than folded into the failures.

"The page names the business" is a deliberately loose test. It catches a site that has become someone else's, but it would pass a site that mentions the business without being its own, and it cannot tell a live firm from a website that has been left running after the firm stopped trading. Those would need a human to read each page, and we read by hand only the 93 addresses that redirect elsewhere.

The classification of those 93 is ours. Where the new page did not make the relationship clear, we said so and left it unclassified rather than calling it an acquisition.

And a website is only one field. A listing whose address still works can be out of date in a dozen other ways — a service it no longer offers, a team that has halved, a price from two years ago. The website is simply the field that can be checked from outside in a second, which is why it is the right place to start, and the wrong place to stop.