You came here wanting to be told which one is best. We are not going to tell you, and the honest thing to do is start with what that costs you rather than with why we feel good about it.
What this is, and who it is for
This is the reasoning behind a rule that shapes everything else on this site: nothing here is scored, starred, tiered, ranked or ordered by merit. There are no editor's picks, no standouts, no top tens. An entry is either in the register or it is not, and if it is in, you get a description and a date.
It is written for two people. The first is a reader who finds that infuriating, because they came to make a decision and a list without an order is a list that makes them do the work. The second is an operator wondering whether there is a position here worth competing for, and who deserves a straight no.
What you lose because we don't rank, stated first, in full
The frustration is legitimate. A ranking is genuinely useful, and pretending otherwise would be a way of avoiding the argument.
Here is what you actually lose:
- Speed. A ranked list lets you stop at the first item. Ours does not. You have to read.
- A shortcut through unfamiliar territory. If you know nothing about a category, an ordering is a borrowed opinion, and borrowed opinions are cheap and often good enough.
- A tiebreaker. When two things look equivalent, a score breaks the deadlock so you do not have to.
- Permission to stop deciding. This is the real one. A rating transfers responsibility. If the thing you picked turns out badly, the list gets some of the blame. Without a rating, the decision is entirely yours, and that is heavier.
- A filter you can sort by. There is no minimum-score slider here. You get a bar that is either cleared or not.
We are not offering a substitute that quietly does the same job. There is no hidden ordering, no default sort that means anything, and nothing on this site is arranged so that the first item is the one we would choose. Where a list has an order, that order is a stated neutral key such as the date of the last check, and we say which key it is.
That is a real cost and we are asking you to bear it. The rest of this explains why we think the alternative is worse.
What a single score destroys about a small business
A score is a compression. It takes several things that are not the same kind of thing and squeezes them into one number, and the squeezing is where the information goes.
Think about what would go into a rating of a small independent business. Price. Speed. Craft. Communication. Whether they answer the phone. Whether the thing lasts. Whether the person is pleasant to deal with when something goes wrong. These do not share a unit. There is no exchange rate between "answers within the hour" and "the joinery is exceptional". To produce one number you have to invent one, and the invented rate is a set of weights.
Two things follow, and both are bad.
The weights are the actual opinion, and they are almost never published. When a list attaches a score to a business, the argument is not really about the score. It is about the decision that responsiveness counts for a fifth of it and durability for a tenth. That decision is where all the judgement lives, and it is the part that gets left out of the presentation. How a ranked list gets assembled walks through the mechanics of that in detail.
The compression flattens the thing that makes small operators worth finding. A workshop that takes eleven weeks and produces something extraordinary and a workshop that ships in three days and produces something fine are not two points on one scale. They are two different answers to two different questions, and the score cannot hold both. Averaging them produces a number that describes neither.
A description can hold both. "This is a two-person bindery; work is commissioned, lead times are long, and you talk to the person who does it" tells you more than any number could, and it does it without asserting that anyone is better than anyone else.
Why an ordering is the easiest thing in a directory to sell
There is a commercial reason as well as an editorial one, and it is blunt.
An ordering creates scarcity. There is exactly one first position, one second, one slot above the fold. Scarcity plus demand produces a price, whether or not anyone sets out to charge it. Every listing site that ranks has, sitting inside it, an inventory of positions that somebody would pay for. The temptation is structural, not moral, and it does not go away because the current editor is honest.
Once positions have a price, several other things follow on their own. Ratings become a service you can offer to improve. Review collection becomes a product. The threshold for inclusion drifts downward, because more entries means more positions and more sellable inventory. How directories actually make money sets out the five revenue models this describes, and four of the five need an ordering to work.
So the rule here is not that we are trustworthy. It is that we removed the inventory. There is no position, so there is nothing to sell, and who pays for this site has to be a very short page as a result.
The enforcement is at the field level rather than in policy. The record structure has no field for a rating, a score, a rank, a star, a tier or a "best". The build rejects those field names outright, so a future editor cannot quietly add one and ship. A policy can be revised in an afternoon by someone with good intentions and a revenue target. A validator that refuses the field name has to be deliberately dismantled, in a commit, by a person, and that is a far more visible act. What we would have to do to become the thing we criticise lists the external signals you could watch for, and a rating field appearing anywhere is the first of them.
Why we can't rate fairly at our size, even if we wanted to
Set the principle aside and the practical problem is still fatal.
Rating small businesses well requires a sample. Not a sample of opinions about the business, a sample of experiences with it. A workshop that fulfils a handful of commissions a month cannot generate enough of those for any statistic to mean anything, and the smaller and more independent the operator, the worse this gets. The businesses this register exists to surface are precisely the ones for which rating is least reliable.
What you get instead, at that scale, is noise treated as signal. One unhappy customer moves a small-sample average further than a dozen happy ones move a large-sample one. A business with three data points and a business with three thousand are not measured on the same instrument, and putting them next to each other with a number attached implies they are.
There is a second problem. We check that a business is reachable and recognisably what we described. We do not buy from every entry, repeatedly, over time, across the range of things they sell. Anything we said about quality would be an extrapolation from a single encounter, dressed as a measurement. The method is deliberately narrow for exactly this reason, and the narrowness is the honest part.
And a third, which is about the operator rather than the reader. Publishing a low score about a small business is a claim about their competence that they cannot answer back on our pages. What we refuse to publish covers the general form of that problem. Rating is the version of it we would run into most often.
What we publish instead: a bar, a date, and a description
Three things, each doing a job a rating would have done badly.
A bar. Six criteria, published before the first entry, applied identically to everything, with no partial pass and no discretionary override. Our entry bar in full walks through each one. The bar is a floor, not a ceiling: clearing it means a business is real, independent, reachable and trading, and says nothing at all about whether it is good.
A date. Every entry carries the date a named person last checked it and the version of the method that produced that date. A date is falsifiable in a way a score is not. You can go and see whether the thing is still there.
A description in our words. One sentence on what it is, plus what you can actually do there: buy, hire, book, read or visit. Operator marketing copy appears only as an attributed quote, never as our description. A description is longer than a number and that is the trade we are making.
How to make the decision yourself
If we are not deciding for you, we owe you something better than good luck.
- Write down what you actually need before you look. Most bad choices come from letting the available options define the criteria. Lead time, budget ceiling, whether you need to talk to a person: decide these first.
- Use our description to eliminate, not to select. A description tells you quickly what something is not for. That is its highest-value use.
- Check the date, then check it yourself. Our date says a person found them on that day. If you are about to spend real money or travel somewhere, contact them. We say this even though it undercuts the convenience of the register.
- Ask the operator two specific questions. Not "are you any good". Something like "what is your current lead time" and "what happens if it arrives damaged". The answers, and the speed of them, tell you more than any aggregate.
- Weight your own risk. A small order to test is worth more than any amount of research.
How to choose without a rating works through this at more length, including the categories where it is genuinely hard.
What would have to be true before we'd reconsider
We do not want to claim this position is objectively correct. It is a set of constraints we chose, and a different publication could make different ones honestly.
For this site to publish anything score-shaped, all of the following would have to hold, and we cannot presently see how they would:
- A sample large enough per business that the number is not mostly noise, gathered in a way that does not depend on the operator soliciting it.
- Published weights, versioned like the method, so the opinion inside the number is visible.
- A structural guarantee that no position derived from the score can ever be sold, influenced or expedited.
- A right of reply for the operator that is at least as prominent as the score itself.
- A correction path that removes a wrong score as fast as it appeared.
If those were ever all true, it would be a different product with a different risk profile, and it would be announced loudly rather than slipped into a schema. Until then the answer is no, and the field name stays rejected at build.
Common questions
Isn't "in the register" itself a kind of rating?
It is a binary, and we try to describe it as one. In means real, independent, reachable and checked on a date. It does not mean recommended, and absence means nothing at all about a business, since the register is small and excludes whole categories for reasons unrelated to the operators in them.
Why not just publish customer reviews and stay out of it?
Because hosting them makes us responsible for them without giving us any way to verify them, and because review sections are the easiest surface on a listing site to manipulate. Reading a review section covers what volume, recency and distribution can and cannot tell you.
Do you order the register at all?
Only by neutral keys we name, such as the date of the last check. No sort on this site is an opinion, and if one ever appears to be, that is a bug worth reporting.
In Plain English
We will not tell you which business is best. That costs you speed, a tiebreaker and the comfort of having something else to blame, and we are asking you to accept that cost. The reasons are that a single score hides the weighting that produced it, that an ordering is a scarce position and scarce positions acquire a price whether or not anyone intends to charge one, and that small businesses cannot generate a sample big enough for a rating to mean anything. What you get instead is a published bar, a dated check by a named person, and a description written here. The rule is enforced by the record structure rather than by our good intentions: the build rejects rating and ranking field names outright, so adding one later would take a visible, deliberate act.
See the entry bar for the six criteria every entry has to clear, and what clearing them does and does not mean.