Start with the honest position. No submissions have been assessed yet. There is no refusal rate, there are no patterns drawn from real cases, and there is nothing in the refusal file to narrate. What follows is the method by which that file will be written, published in advance so that the first pattern can be read against a procedure rather than against whatever seemed reasonable at the time.
What this is, and who it is for
An inclusion list proves nothing on its own. Anybody can publish the things they said yes to, and a list of yeses is compatible with having no bar at all. The only way to show that a bar exists is to show what it stopped.
That is what the refusal file is for, and this post explains how it gets written: what an anonymised pattern is, how one is constructed, which patterns we expect the criteria to generate, and how a refusal gets looked at again.
Two audiences. An operator deciding whether to submit, who wants to know what would sink them and whether it is fixable. And a sceptic deciding whether this register is real, who is right to want evidence of refusals rather than assurances about standards.
A note on what this post is not. It contains no case, no composite drawn from a case, no example built from something we saw, and no estimated rate. Writing one to make the page feel substantial would be the exact failure the page exists to guard against. When the assessment log has entries, this post gets rewritten from it.
How we anonymise, and why we never name a refusal
The rule comes first because everything else depends on it.
We never name a refused business. Not in a pattern, not in an example, not in passing, and not in a way a determined reader could reverse.
What gets stripped or altered:
- The name, the address, the domain and anything that resolves to them.
- Sector, location, size and backstory in combination. Any one of those is usually harmless. All four together identify a small business even when each individually does not, and that combination is how anonymisation actually fails.
- Distinctive detail. An unusual product, a memorable phrase from the site, a niche a reader could search.
- Timing that would narrow the field, such as tying a pattern to a particular week.
And a structural rule on top: every published pattern is a composite drawn from more than one case. If a pattern could only have come from a single submission, it does not publish, however instructive it would have been. That constraint is the reason this file will grow slowly, and it is not negotiable.
Why go this far. A refusal is not an accusation, and we do not want it read as one. Most refusals will be ordinary and fixable problems, of the kind almost every small site has at some point. Publishing a pattern is useful. Publishing a pattern that a reader can pin on a specific person is publishing a claim about that person's business that they cannot answer back on our pages, which is the thing the Line exists to prevent.
How a pattern is written
Five steps, in order, every time:
- Name the pattern in plain language, the way a reader would recognise it in their own site.
- State which criterion of the entry bar it fails, and why that criterion exists at all.
- Describe how we spot it, so an operator can spot it themselves without submitting first.
- Give a composite illustration with every identifying detail removed.
- Say what would have to change for the same business to pass.
The fifth step is the one that turns this from a complaints file into something worth reading. A refusal file that only catalogues failure is a list of grievances. One that says how to clear the bar is a service, and it is more useful than the individual rejection message the operator will already have received.
The patterns we expect, and where they come from
These are derived from the six criteria these submissions would fail, not from cases. Each is a prediction about what the bar will catch, written now so that later you can check whether the predictions held. None of them describes anything that has happened.
Pattern: the site that doesn't say what it sells
Fails criterion four: a named person has to be able to describe it in our own words, and nobody can describe what they cannot identify. We expect this to be the most common refusal by a wide margin, because it is the most common failure of small sites generally.
The shape of it is a homepage full of language about values, journeys and craft, with no sentence anywhere that names the thing you can buy. The operator knows what they sell so thoroughly that they stop saying it.
How it would be spotted: a reader with no prior knowledge cannot answer "what does this business do" from the first screen. What would fix it: one plain sentence, high on the page. Your homepage doesn't say what you sell is the long version, and the shopfront patterns covers the family this belongs to.
Pattern: the shop that is really an affiliate page
Fails criterion three. A business may carry advertising or affiliate links; what disqualifies it is the arrangement being the actual product. When the thing on offer is a click rather than goods or services, it is out.
Often entirely legal, properly disclosed, and run by someone perfectly straightforward. Still not what this register is for, because a visitor arriving from here should land somewhere they can buy, hire, book, read or visit, not somewhere that forwards them.
How it would be spotted: every route to acting on the page leaves for a third party, and nothing is fulfilled by the operator. What would fix it: nothing short of a different business model, which is why this refusal is usually final rather than a to-do list.
Pattern: the launch that hasn't launched
Fails criterion one. A waitlist is not a thing you can buy, use, hire, read or visit today. Neither is a landing page with a mailing list, a pre-order for something unfinished, or a launch announcement.
This is the least painful refusal, because it comes with a date attached. It is a "not yet" rather than a no.
How it would be spotted: the primary action on the site is joining something rather than doing something. What would fix it: shipping. Submit again when a stranger can complete a transaction.
Pattern: the business whose independence can't be established
Fails criterion two. Independent ownership is recorded as a checked fact with the basis stated on the record, not inferred from the size of the website. Where the ownership question has no answer we can establish, the answer is no rather than a guess.
This one refuses businesses that are, in fact, independent, simply because we could not show it. That is uncomfortable and we would rather say so than pretend the criterion is cleanly applied.
How it would be spotted: no ownership information anywhere on the site, and nothing checkable elsewhere. What would fix it: a trading-details line on the site naming who is behind it. Putting your trading details on your site covers what to include and what to leave off.
Pattern: the site where no human can be reached
Fails criterion four in practice. If nobody can be reached, nothing can be confirmed at review, nothing can be re-checked later, and nothing can be corrected if we get it wrong. An entry we cannot correct is an entry we should not publish.
How it would be spotted: no working contact route, or a form that silently discards messages. We test contact routes from an address we do not control, which catches more broken forms than anyone expects. What would fix it: a monitored address, tested from outside.
Pattern: the good business in an excluded category
Fails nothing about itself. This is the most important pattern in the file and the reason the file needs to exist.
Whole categories are excluded outright at the entry bar: adult commerce; gambling, betting and prediction markets; lending, credit and debt services; supplements, health products and anything making a health claim; multi-level marketing and recruitment-driven income schemes; crypto tokens, exchanges and trading services; and anything requiring a licence or authorisation we cannot verify.
The exclusions reflect regulatory, safety and verification exposure we are not equipped to carry. They are not a view about the businesses in them, and a refusal on this ground says nothing whatsoever about the operator. There is no editorial appeal, because reopening a category is a legal decision taken with advice rather than an editorial one taken because a good business asked well.
What would fix it: nothing the operator can do, and we would rather say that plainly than leave someone resubmitting into a wall.
The refusal rate, and what it should tell you
There is no refusal rate. Nothing has been assessed.
We are not going to estimate one, extrapolate from the criteria, or quote a plausible-sounding figure with a hedge attached. An estimated rate on a page about honesty would be a strange thing to publish.
A conforming refusal rate can come only from an assessment log, never from this description of the method. This article makes no rate claim.
Two things worth saying now about how to read it later.
A high rate is not automatically a good sign. It could mean the bar is doing its job, or it could mean the submission form is attracting the wrong things, or that our guidance before the form is poor. A low rate is not automatically bad either; it could mean the pre-submission advice works. The number on its own carries less information than people expect, which is why the patterns matter more than the percentage.
And the rate has to be published even when it is embarrassing. A rate we only mention in good quarters is not a measurement, it is marketing.
How to get a refusal reconsidered
Free, unlimited, and with no queue to jump, because there is no paid queue to jump.
- You will have been told the specific criterion that failed.
- Fix that thing.
- Submit again. How to submit a business sets out what happens next.
There is no blacklist, no strike count, and nobody keeping track of how many times you have asked. There is also no way to pay for a second look, an expedited review, or a more sympathetic reviewer. Buying anything from this site, now or ever, has no effect on register consideration, and that rule is written into the entry bar rather than only stated here.
If you disagree with the refusal rather than wanting to fix it, say so and a second person reads it. That costs nothing either.
In Plain English
This edition supplies no assessment log, observed pattern, or refusal rate. The six scenarios above are criteria-derived examples, not accounts of events. The complete method is: never name a refused business; strip identifying combinations of sector, location, size, and backstory; publish a pattern only as a multi-case composite; allow reconsideration without charge; and never let payment affect a decision.
Read the full refusal file for the current state of it, and the shopfront basics if you suspect your own site is the first pattern.