You lose an afternoon to it. Four promising names from a listing page. The first number rings out. The second is disconnected. The third answers and says they stopped doing that a few years ago. The fourth has a website that still loads because the domain renews automatically, and nothing behind it.
None of this is anybody's fault in a moral sense, and no one decided to waste your afternoon. It is an economic outcome, and once you see the arithmetic behind it, every stale listing you have ever hit explains itself.
What this is, and who it is for
This is an explanation of why dead listings accumulate, written for anyone who has hit a run of them and assumed it meant carelessness, and for operators trying to understand the ecosystem their business sits inside.
It is not an accusation. Almost nobody in this category is failing to prune out of indifference. They are failing to prune because nothing in the way listing sites are funded pays for pruning, and an activity that nothing pays for does not happen at scale. That is the whole argument, and the rest of this page is the detail.
Why adding a listing costs almost nothing
Intake is close to free, and it has three cheap sources.
Bulk import. Business details exist in many public places, and moving them into a database is a one-off engineering job. Once written, the import runs again for nothing. A category can go from empty to comprehensive in an afternoon.
Self-service submission. A form and a moderation queue. The operator does the typing. The site's marginal cost is a few seconds of someone's attention, and often not even that.
Sold inclusion. Where a listing is paid for, intake is not merely cheap. It is the thing generating revenue, so every incentive points at more of it. That is one of the five business models behind listing sites, and it is the one that shapes this problem hardest.
Whichever route, the cost of an additional listing is close to zero and the benefit is immediate: more pages, more coverage, a category that looks populated rather than thin. Everything about intake is favourable.
Why checking one costs a great deal, repeatedly
Now the other side of the ledger, and the word that matters is repeatedly.
Checking a business is trading is not a lookup. It usually needs a person. Does the site still load, and does it still describe the same business? Is the phone answered? Has the shop moved? Has the product been discontinued while the company continues? Some of that can be partly automated, and every automation produces false results in both directions: a site that returns an error for a week while the business trades happily, and a site that loads perfectly for years after the business stopped.
Worse, the check has no lasting value. It is true on the day and it starts decaying immediately. A listing checked in March needs checking again. And again. The cost of an entry is therefore not the cost of adding it. It is the cost of checking it, multiplied by every future check, for as long as the entry exists.
That gives you a recurring obligation that grows in direct proportion to the size of the index. Ten thousand entries checked twice a year is twenty thousand checks a year, forever, and none of it produces a new page or a new sale. The economics underneath this follows the money one level further down.
What happens to a listing when nobody is paying for the check
Put the two sides together and the outcome is not a mystery.
Intake produces revenue, pages and the appearance of coverage. Pruning produces a smaller index, fewer pages, and a category that now looks thin. If listings are sold, removing one removes revenue and creates a conversation nobody wants to have. If listings are free, removing one still costs staff time and returns nothing measurable.
So the rational move, for almost every business model in the category, is to keep adding and never subtract. Not through malice. Through the ordinary process of an organisation doing more of what is rewarded.
There is a second-order effect worth naming. Once an index is large and stale, checking it becomes politically difficult as well as expensive, because a serious check would visibly shrink a number the site has been quoting. Coverage claims create their own resistance to accuracy.
The reader carries the cost of all of this, and the reader is the only party in the arrangement who is not consulted.
The signals a dead listing gives off, and why they are easy to miss
Dead listings are not blank. They look exactly like live ones, which is the problem. But they usually carry traces, and the traces are subtle because they are absences rather than presences.
- No date anywhere. Not on the listing, not on the page. Nothing to say when anyone last looked. This is the largest single tell and it is invisible because you are not looking for something that is not there.
- A page-level date with no entry-level dates. The page says it was updated this year. The listings say nothing. Updating a page is cheap; checking a listing is not.
- Details that disagree internally. A phone number in one format in the header and another in the body, or an address that does not match the map, are marks of an entry assembled from sources captured at different times.
- A frozen social link. The listing links to a profile whose last post is years old. The listing does not know that, because nothing follows the link.
- Generic description, specific detail. A vague summary that could describe any business in the category, attached to a very specific address, is characteristic of bulk import.
- The category is suspiciously complete. A long, evenly populated list in a small area is usually a sign of import rather than of a thriving local sector.
- Nothing has ever been removed. If a site has a way of marking closures, look for one. If it has none, closures are not being recorded, which means they are not being found.
The reason these slip past is that all of them look like ordinary listing pages. Nothing is broken. There is no error state for out of date.
What a register that prunes has to give up in exchange
We do prune, and it is only honest to say what that costs, because the costs are large and they fall on you as much as on us.
We are much smaller, and always will be. Every entry carries a recurring obligation, so the size of the register is capped by the amount of checking that can actually be done. There is no version of this that ends with comprehensive coverage. A register that prunes is a register that grows slowly by construction.
Gaps are permanent, not temporary. Whole categories and whole places will be missing, and the honest answer to "why is there nothing here for my area" is that nobody has checked anything there yet. We would rather say that than fill the space.
Nothing here is comprehensive enough to be your only source. If you want every option, this is the wrong kind of site.
And we will still have dead entries. This is the part that matters most. Pruning does not mean an index is clean. It means the errors are bounded by a cadence instead of being unbounded. Between two checks, a business can close and the entry will be wrong, and there is no procedure that prevents it. What we are claiming is narrower and duller than accuracy: a published check date on every entry, a published cadence, and a public record when something is confirmed gone.
That last part is the tombstone. When a business is confirmed to have stopped trading, the entry is not deleted. It becomes a marked record showing that it existed, that it was here, and the date of the final check. The reasoning is that a deletion tells you nothing, whereas a tombstone answers the question you actually had, which was what happened to them. The procedure for those records is at Gone, and the sequence that leads to one is described in what happens when a listed business disappears. Getting it wrong in that direction is serious, so the bar for declaring a closure is deliberately high and no reason for a closure is ever stated unless the operator published one themselves.
The procedure behind all of this, including where it fails, is at our method, and the register schema it applies to is at the register. How we check a business is still trading is the long version.
How to check a listing yourself before you act on it
You do not need a register to do this. Three checks, in order of speed, before you commit any time to a name you found on a list.
1. Look for a date, then decide how much it can carry. Find any date attached to the individual listing. No date means the listing is making no claim about currency, and you should treat it as an unverified name rather than a live business. A date that exists still needs weighing against how fast the thing decays: a company's continued existence changes slowly, a small shop's opening hours change constantly.
2. Go to the business's own site and look for recent life. Not the listing's copy of it. On the real site, look for anything that has to be maintained by a person: a dated post, current prices, a seasonal notice, a change to opening hours. A copyright year is weak evidence because it is often generated automatically. A stock notice or a holiday message is strong evidence, because nobody writes one of those in advance by accident.
3. Use a channel that fails loudly. Email and contact forms fail silently. A dead mailbox usually just swallows the message. A phone number tells you within ten seconds whether anything is at the other end, and so does turning up. When it matters, use the channel that cannot pretend.
If all three come back positive, you have done more checking than the listing did.
In Plain English
What it means. Adding a listing is a one-off cost, keeping it true is a recurring one, and almost no listing site is funded to pay the second. Dead entries accumulate as a consequence of that arithmetic, not as a consequence of carelessness.
Who it helps. Anyone using listing pages to find a supplier, and any operator wondering why an old listing of theirs never disappears.
When to use it. Before you spend an afternoon phoning names off a list.
What to do next. Run the three checks above on the next listing you plan to act on. Start with the date, because if there is not one, the other two are the only evidence you have.
Where closures are recorded
A register that never records a closure is a register that is not looking. The tombstone procedure, with the final check date each record must carry, is at Gone; no tombstone has been published yet.