How We Confirm A Shop Has Closed Before We Say So

Of every mistake this site could make, one stands well clear of the rest: saying a trading business has closed when it has not.

For the broader context behind this business guide, read How We Verify A Physical Business Without Publishing Anyone's Home Address.

Everything else is recoverable. A wrong description is embarrassing. Stale hours waste somebody's afternoon. A missing entry costs a business nothing, because it never had anything to lose. A published statement that somebody's livelihood has ended reaches customers who were about to use them, at exactly the moment they were deciding.

So the procedure below is designed around that single error. It is deliberately slow, and the slowness is the feature.

Scope. This page does not say what the law requires of anybody, here or anywhere else. That question turns on jurisdiction and circumstance and we are not qualified to answer it. The procedure below is built to avoid the harm regardless of what any legal answer turns out to be, which is why it is written as a threshold we hold ourselves to rather than as a duty anyone is under.

What this is, and who it is for

For an operator who wants to know what would have to happen before this site said anything about their business having closed, and for a reader trying to work out how much a closure record on any site is actually worth.

The general procedure for an entry that stops responding is at the general disappearance procedure. This is the physical-business version, which is stricter, because a locked door is so much easier to misread than an unresponsive server.

Why this is the error we most want to avoid

Consider it from the operator's side rather than ours.

A business that is quietly shut for a fortnight while its owner recovers from something has not stopped trading. If a record appears during that fortnight saying it has, the harm does not wait for the record to be corrected. Somebody who would have phoned does not phone. Somebody who was going to come back assumes there is nowhere to come back to. A supplier who checks assumes the worst. And the correction, when it comes, does not travel to any of those people, because corrections never travel as far as the original.

There is a second asymmetry worth naming. Being absent from this register costs a business nothing measurable. Being wrongly recorded here as closed costs something real. Those two errors are not equivalent, so we do not treat them as equivalent. We would much rather have a missing entry than a wrong tombstone, and every threshold below reflects that preference.

There is a legal question sitting alongside this one — what a publisher exposes itself to by stating incorrectly that somebody's business has ended — and this site asserts no answer to it. It turns on jurisdiction and circumstance, we are not qualified to resolve it, and a page that guessed would be offering false comfort to the operator as much as to itself. So the threshold below is set by the harm to the operator rather than by any legal test, which is the safer of the two ways round: a procedure built to avoid the harm avoids it regardless of what any legal answer turns out to be.

What does not count as evidence of closure

Each of these is something that regularly gets treated as proof by listing sites, and none of them is.

• A closed door on one day. Could be any of four states, only one of which is permanent. See the seasonal closure problem.
• A website that stopped responding. Certificates expire. Bills go unpaid for a week. Hosting moves badly. None of it says anything about whether the business is trading.
• A phone that went unanswered. People are busy, or away, or on the shop floor.
• A quiet account. A business that has posted nothing for months has posted nothing for months.
• A third party telling us. A report is a reason to look. It is not a confirmation, and it never bypasses the sequence.
• Another site saying so. We do not know how they concluded it, and copying an unexamined claim about somebody's livelihood is precisely the practice we would criticise elsewhere.
• A change of premises, name or ownership. These are changes, not endings, and they are handled separately below.
• An absence of evidence. Not finding something is not the same as establishing it is gone.

The threshold, in full

Two failed checks, at least fourteen days apart, plus one independent human confirmation.

All three elements, in that order. There is no partial version and no discretionary shortcut. Each element exists to prevent a specific error, so it is worth taking them one at a time.

Element	The error it prevents
The first failed check	Nothing on its own. It starts the sequence and publishes nothing.
A second failed check	A transient fault being read as an ending. Outages, expired certificates, a bad afternoon on somebody's hosting, a phone unanswered once.
At least fourteen days between them	A short absence being read as a permanent one. A holiday, a refit, an illness, a stock rebuild, a week of unusually bad luck. Two failures ten minutes apart are one failure observed twice; separating them in time is what makes the second failure independent evidence.
An independent human confirmation	The specific failures a script cannot see: that the business moved, rebranded, was sold, or paused. All four of those look identical to an automated check, and all four are the wrong answer.

Nothing is published before all three are satisfied. There is no interim state visible to readers, no "may have closed" note, no greying out. During the sequence the entry sits flagged as unreachable or sits unchanged, and either way the site says nothing about an unconfirmed closure.

The same threshold is stated at our method and at gone, and it is stated identically in all three places on purpose.

The human confirmation step, and what a person actually does

This is the part that automation cannot do, and it is the reason the last step has a name attached to it.

A named person takes the entry and looks specifically for the four things the sequence could be misreading. Not for further evidence of closure — for evidence that the closure interpretation is wrong.

They look for a move: the same trade, the same people, a different place. A rebrand: the same operation under a different name. A sale: a change of hands, the business continuing. A pause: anything indicating a temporary stop, including an announcement made in a channel our check would not reach.

They also look for anything the operator themselves has said, because if the operator has published a statement, that is the record we would rather use than any inference of ours.

Where the question can only be settled by somebody standing there, that is a physical visit and the record says so. A desk check does not get described as a visit.

If the person finds any of the four alternatives, the sequence stops and the entry is handled under that heading instead. If they find nothing that contradicts the closure reading, and only then, the entry becomes a tombstone carrying the date it was last seen alive, the date of the first failed check, and the date closure was confirmed.

One thing a tombstone is not: an official or authoritative record of anything. We hold no registry status of any kind. It is our observation, dated, with the method version attached, and its only authority is that we say exactly how we arrived at it.

Why we never publish a reason

A tombstone says what we observed. It does not say why, and it never will unless the owner published a reason themselves, in which case we quote them with attribution and a capture date.

The justification is entirely from the owner's position.

We almost never know. Every plausible reason we might infer is a guess, and a guess about somebody's livelihood is not a thing to publish.

The reasons people assume are damaging and usually wrong. A reader given no reason assumes little. A reader given a hint assembles a story, and those stories tend towards debt, dispute or failure. Most closures are none of those. Retirement, illness, a lease ending, a move, a change of direction, a decision that it had run its course.

The person it describes has to live afterwards. They will start something else, look for work, or be known in their trade. A published characterisation follows them and cannot be recalled.

A reason we cannot evidence is a claim we cannot defend. Our whole position is that everything here is checked and dated or quoted and attributed. A reason is neither.

So the wording is chosen so it cannot be read as a verdict. No speculation about difficulty, debt, dispute or conduct. No characterisation of the business or the people who ran it. Nothing that reads as an obituary or a warning. Removing an entry is not an accusation, and a tombstone is not one either. The full list of what we will never publish about a business is at what we will never publish about you, and the permanent version is on the Line.

Moved, rebranded, sold, paused: four cases we handle differently

Of every outcome the confirmation step can produce, exactly one leads to a tombstone, and that is a confirmed permanent closure. None of the four below does. They are four distinct outcomes with four distinct treatments, and they are the reason the human step exists at all.

Outcome	What actually happened	What we do	Tombstone?
Moved	Same business, same people, different place or different way of trading.	The entry is re-reviewed and rewritten with the new details, verified on a new date. The check history stays attached, so the record remains coherent.	No
Rebranded	Same operation, different name or presentation.	Re-reviewed and rewritten under the new name. We do not create a second entry, and we do not leave the old name standing as though it were a separate business that closed.	No
Sold	The business continues under new ownership.	Re-reviewed from the beginning against the entry bar, because independence may have changed and independence is a criterion. If it no longer clears the bar, the entry is removed with no note. A removal is not a closure and is never presented as one.	No
Paused	Trading has stopped temporarily, seasonally or for an unstated period.	The entry is flagged, not removed. It stays in the register and in the export. A flag is not a soft tombstone and it is not a step towards one.	No

The fifth outcome, a confirmed permanent closure, is the only one that produces a record at gone.

The reinstatement route

Free, immediate, and requiring no proof beyond your being there.

Tell us at corrections and the entry comes back. You do not need to demonstrate anything, produce documents, or explain why you were unreachable when we looked. If you are telling us you are trading, you are trading, and the burden of having been wrong is ours rather than yours.

The error then goes into the public correction log with a date and a description. That is the correct cost for this particular mistake. A site that gets closures wrong and absorbs the consequence privately has no incentive to keep the threshold high.

If you would rather the record simply came off the site entirely, request removal does that, no reason required, and it comes out of the machine-readable export as well.

If you are the owner and we got it wrong

The order of operations matters more than anything else on this page.

It comes down first. Then any discussion about whether we agree. Not the other way round.

• Tell us at corrections or request removal. Either works.
• You do not have to prove you are trading. You do not have to prove who you are beyond enough for us to be confident we are speaking to the operator, and the bar for that is proportionate.
• There is no charge. There has never been a charge for removal on this site and there will not be one.
• Nobody will try to talk you out of it, offer you an alternative, or ask you to justify the request.
• You can ask to be removed permanently and never listed again, and that is recorded and honoured.

If we missed something you sent us, that is also a correction and goes in the log.

The rest of the physical-business material is at real places, checked.

In Plain English

We do not say a business has closed unless we have gone through a slow, specific sequence, and even then we never say why.

• The threshold is two failed checks, at least fourteen days apart, plus one independent human confirmation. All three, no shortcuts.
• The second check catches transient faults. The fourteen days catch short absences. The human catches the four things a script cannot see: moved, rebranded, sold, paused.
• Nothing is published while the sequence runs. There is no "may have closed" state.
• A closed door, a dead website, an unanswered phone, a quiet account or somebody telling us are all reasons to look. None is evidence of closure.
• We never publish a reason unless the owner published one, because we almost never know and the assumptions readers make are damaging and usually wrong.
• Moved, rebranded, sold and paused are handled four different ways and none of them produces a tombstone. Only a confirmed permanent closure does.
• If we got it wrong, it comes back free and immediately, no proof needed, and our error goes in the public log.

Read the tombstone procedure → /gone/