Four weeks ago this masthead commissioned a straightforward piece: take the watch brands that launched during the 2020 crowdfunding boom, check who is still shipping, and report the survival rate. The brief was specific about the standard, and the sentence that mattered was this one — a survival count with a shaky denominator is worse than no count.
That piece cannot be written. Not because the work is hard, but because the record it would be built from is no longer open to inspection. What follows is what we tried and what came back, because the closing of that record is a better story than the percentage would have been, and it is one we can actually source.
What the count would have required
Three things, in order. A list of the brands that launched in a defined window — the denominator. The current status of each: shipping, dormant, gone, acquired. And a method stated plainly enough that a reader could repeat it and get the same answer.
The first is where it stops. There is no register of watch brands. There is no trade body with a membership roll. The one institution that holds a transactional record of the boom — who raised money, when, how much, and what they promised to deliver — is the crowdfunding platform itself, because that is where this cohort was born. Almost every brand in it launched by taking money from the public before it had a product.
That record exists. It is simply not available to look at.
The directory is shut
Kickstarter’s project directory is closed to inspection. So is its category search. So, and this is the part worth pausing on, is the notice on the door: the file a site publishes to set out what may and may not be gathered from it is refused along with everything else. There is no rule to follow, because the rules have been taken inside with the records. Indiegogo, the boom’s other venue, is shut in the same way and to the same extent. The dates and the exact responses are in the sources below, for anyone who wants to knock themselves.
We should be careful about what that does and does not mean. It is not evidence of bad faith, and it is almost certainly not a decision anybody took about the historical record. Blocking of this kind is bought off the shelf, to stop bulk copying and password attacks, and it does not distinguish between someone lifting a database wholesale and someone checking what happened to a company that took £400 off a reader in 2020.
That is rather the point. The record closed as a side effect of an infrastructure purchase, and nobody had to weigh what was being lost — because nobody was asked.
Kickstarter used to publish the numbers itself
It is worth being precise here, and precise includes being precise about what we could not check.
What we can show is this: there is a URL at kickstarter.com/help/stats, and asking for it today returns a 302 redirect to a post titled A Year in Review: 2025 Kickstarter Highlights. That post is well written and contains no data. It says the year was “Kickstarter’s biggest year to date”, that creators “responded not by pulling back, but by leaning in”, and that there is “a ravenous appetite for originality”.
What we cannot show is what used to be at that address. The name is suggestive and our recollection is that it once carried live aggregate figures, but we could not get to an archive to confirm it, and a recollection is not a citation. So we are not going to tell you it did.
Even taken at only what is verifiable, the redirect is the shape of the problem. A path named /stats answers with prose, in the first person plural, from the company being asked.
Even with the list, the second half is missing
Suppose the directory reopened tomorrow and we could assemble the cohort exactly. The count still would not be straightforward, because stopped trading is not a recorded event for most of these businesses.
Companies House is open, and was open the same afternoon everything else was shut, which is how we know the closed doors are closed rather than broken. But it covers UK entities only, and this cohort is international by construction: crowdfunding removed the need for a domestic distributor, which was much of its appeal. A UK register tells you about the British ones and is silent on the rest.
And a brand does not usually fail in a way that generates a filing. It stops answering email. The Instagram account posts less, then stops. The website stays up for as long as the domain is paid for, sometimes years, still taking pre-orders it will not fulfil. There is no bankruptcy, because there was frequently no company — just a founder, a supplier in Shenzhen and a Shopify account. The failure is real and the paperwork never happens.
So the honest position is not “we could not get the data”. It is that for a large part of this cohort the data was never created. The money-raising was a public event. The dying is a private one.
Why this is the actual story
We have written before that most of the current independent boom will not survive, and we stand by the argument. But we made that claim as a prediction, and a masthead that publishes predictions owes its readers a check on them later.
This is the check, and it comes back: the check cannot be run. Not by us, and — as far as we can establish — not by anybody, because the necessary record is split between a platform that will not serve it and a set of outcomes nobody ever wrote down.
That has a consequence worth stating plainly. Every survival or failure rate you have seen quoted for this cohort is a guess. We are not saying the guesses are wrong; some of them are probably close. We are saying that none of the ones we have found state a cohort, a window and a method, and a figure without those three is a number-shaped opinion. That includes any figure we might have produced by assembling a list from memory, which is precisely what the commissioning brief warned against and precisely why this piece exists instead.
What would fix it
Very little, is the frustrating part.
A platform that published a project directory as data — even annually, even in arrears — would restore the denominator at almost no cost to itself, and it would be publishing a record it already keeps. Nothing about that is technically difficult, and it does not require anyone to disclose a single backer’s details. It is a choice, available whenever either company wants to make it.
Failing that, the cohort could be reconstructed by hand from contemporaneous coverage, and we would rather that than nothing. But it would be a list of the brands that got written about, which is a different population from the brands that launched, and skewed in the direction of the ones that were doing well enough to be noticed. A survival rate built on it would be flattering by construction. If we ever run it, we will say so in the same breath as the number.
Checking one brand, which does work
None of the above stops you checking a single brand, and if you are holding a pre-order it is worth twenty minutes. The cohort is unknowable; the individual is not.
Order the checks by how hard they are to fake. Has anything shipped recently? Not posted — shipped, to somebody who is not the founder, with a photograph of the watch in a hand rather than a render. Owner forums and the brand’s own tagged photos are better evidence than its feed. Is the shop actually taking orders, and for what delivery date? A site quoting the same “shipping next quarter” it quoted last year is telling you something. Does the company exist where it says it does? For a UK brand, Companies House will show whether accounts are overdue and whether it has been proposed for strike-off, both of which are public and both of which precede the silence by months. And when did the founder last answer anybody? A support email that goes unanswered for three weeks is the most reliable signal in the list, because it is the one thing a struggling brand cannot outsource or automate.
None of that scales to a cohort — it is an afternoon per brand, and the denominator problem is exactly that nobody knows how many afternoons. But for the one brand whose money you are holding a receipt for, it is the whole answer, and it does not depend on any platform choosing to publish anything.
What we are not claiming
Two limits, because a piece about the limits of a record should be honest about its own.
We did not establish what the Internet Archive holds. We could not get a usable answer out of it on the day, so we have no basis for saying the archived record is complete, partial or useless. It may well be a route for somebody with more patience than we had that afternoon, and if it is, the reconstruction above becomes a good deal more tractable.
And we tried two platforms on one day. A closed door on one afternoon is a snapshot, not a policy. If either company opens its directory again — or tells us the block was never meant to catch a question like this one — we will say so here, dated, and get on with writing the piece we actually commissioned.
Until then, the sector’s survival rate is not a hard number that we failed to find. It is a number that does not exist, about an industry that raised its money in the most public way available and then stopped filing reports.