We generated twenty brand kits and measured what industry actually changed

There is a genre of article called "brand kit ideas for [your industry]" and almost all of it is written the same way: somebody picks a colour palette that feels right for dentists, writes two paragraphs about trust and cleanliness, and moves on to the next industry. It is unfalsifiable. Nobody can check whether a sage-green palette is actually what dentistry calls for, because there is nothing behind the claim except the writer's taste.

We are in a position to do something slightly better, so this is the article we could write and they could not.

Over the last month we generated a brand kit for twenty different industries, using the same engine that runs behind the free brand kit generator, each one from a written brief describing a real business in that trade. Every kit is published, along with the brief it came from, at brand kit ideas. The engine is deterministic: given the same brief it produces the same kit, so every result below is reproducible rather than remembered.

Then we measured the twenty outputs against each other. The result was more interesting than we expected, and it is not what the genre assumes.

Colour individuated. Typography did not.

Across twenty industries:

• Signature colour: 20 distinct values out of 20. No two industries received the same one. Not similar-but-different — twenty different hex values.
• Heading typeface: 5 distinct faces out of 20. One face, Fraunces, was selected for eleven of the twenty. (What a typeface pairing is doing in an identity.)
• Body typeface: 3 distinct faces out of 20.
• Complete type pairings: 10 out of 20. Half the industries share a pairing with at least one other.

So the same input variation that moved colour every single time moved typography less than half the time. That asymmetry is the finding, and it is worth sitting with, because it contradicts how the "brand kit for your industry" genre is written. Those articles spend most of their words on fonts.

Why this happens, and why it is not a flaw

The temptation is to read that as the engine being lazy about type. It is not, and understanding why says something useful about identity generally.

Colour has an enormous continuous space and very weak constraints. Any hue works at any weight; a saturated coral is as legible as a deep green. So a small difference in the brief — a different audience, a different tone word, a different offer — moves the result, and there is somewhere for it to move to.

Typography has a small discrete space and very strong constraints. A typeface has to remain readable at body size, has to have the weights an identity system needs, has to not be so distinctive that it dates in eighteen months. The set of faces that clear those bars for a small business is genuinely small — smaller than the set of colours by orders of magnitude. When twenty briefs land in a space that narrow, collisions are not a failure of imagination. They are arithmetic.

There is a second reason, and it is the more important one. Typography encodes register, and register does not vary by industry nearly as much as people assume. Fraunces — the face selected eleven times — is a serif with a slightly editorial, slightly warm register. The eleven that received it are the coach, the real estate agent, the restaurant, the photographer, the consultant, the barber, the dentist, the law firm, the accountant, the florist and the yoga studio. A dentist and a florist have nothing in common as businesses, and that is the point: what they share is a buyer choosing a person, in a transaction where the writing wants to feel considered rather than efficient. The industries differ enormously. The register does not.

The trades that got something else got it for a structural reason, not an aesthetic one. The SaaS startup and the therapist both received Archivo, a grotesque; the boutique and the nutritionist received Space Grotesk. What those pairs have in common is not category but density of information at small sizes — one is a product interface, one is clinical detail, one is a product catalogue, one is nutritional fact. A serif with editorial contrast is worse at that job regardless of how the industry feels.

What the colours actually did

The hue distribution across twenty industries:

| Hue family | Industries |
| --- | --- |
| Green | 8 |
| Red | 4 |
| Magenta | 3 |
| Blue | 2 |
| Yellow | 2 |
| Violet | 1 |

Two things here are worth naming honestly.

Green being the largest group is not a claim that green is right for small business. It is a fact about the briefs we wrote and the engine's selection behaviour, and eight out of twenty is not a majority — it is a plurality in a six-way split. If you take one thing from that table, take the spread, not the winner.

Blue at two is the genuinely surprising number. Blue is the default corporate identity colour, and in a set that includes an accountant, a law firm, a dentist and a consultant — four trades where a stock-photo brand guide would reach for navy without pausing — it came up twice. The engine reads the brief for what the business is actually competing on. The law firm's brief, published in full on its page, describes the audience as "business owners who need a straight answer" and the tone as "exact, reserved, sourced". Navy is not the answer to that brief; navy is the answer to "we want to look like a law firm", which is the one thing that brief does not ask for. The result was a deep green. Whether you agree with that call is a legitimate argument, and it is a more interesting one than "here are ten palettes for lawyers".

The part your industry does not determine

Here is the practical version of all of the above, and it is the reason we published the twenty pages with their briefs attached rather than as a gallery.

Your industry determines less of your identity than the genre implies, and a different part of it than you would guess. It sets constraints — what has to be legible in what conditions, what conventions your buyer will read as competent, what a competitor's identity already occupies. It does not determine your colour, and the twenty distinct signature values are the evidence: the thing that moved was not the trade, it was the brief. Audience, offer, tone and position moved the colour. Industry mostly moved the constraints.

Which means the sentence "a brand kit for dentists" is doing less work than it appears to. There is no such thing. There is a brand kit for this dental practice, whose buyer is anxious, whose differentiator is being the practice that explains things — and that kit will look nothing like one for a cosmetic clinic in the same trade whose buyer is choosing on aspiration.

That is why every one of the twenty pages prints the brief beside the kit. The kit without the brief is decoration. The brief is the thing that produced it.

If you want to run this yourself

The engine is free and ungated, and nothing below requires an account:

1. Write the brief first, in prose. Who you serve, what you sell, how you want to sound, and what you are competing against. This is the input that does the work, and it is the step people skip.
2. Run it through the brand kit generator.
3. Compare the output against the closest of the twenty published examples. If yours looks similar to one whose brief looks different, that is the signal that your brief was too thin to individuate anything.

The last step is the one worth taking seriously. A generic output is almost never the generator's fault. It is a brief that did not say anything a model could act on, and the fastest way to see that is to hold it next to somebody else's.

If you want the vocabulary before the procedure, our glossary defines the terms this post uses — brand kit, brand voice, brand drift — each in a paragraph you could quote without the page around it. And if you want the build order rather than the analysis, how to build a brand kit walks the sequence: positioning first, colours nearly last.

One caveat on all of the above: this is a vendor writing about its own engine, using data only it can see. That is a real limitation and not a small one. How to read a comparison page, including ours is the method we would apply to a post like this if somebody else had written it.

If you would rather run it than read about it, what you can actually finish without signing up walks the kit generator and five other tools end to end, and says where each one stops being enough.

Two things a kit does not settle, both of which come up immediately afterwards. Layout is one: a template solves layout, not the thing that is wrong is about the decisions a template leaves open, and the four packs on our carousel template gallery are each rendered from a real kit's palette by the exporter itself rather than mocked up. The other is which of these trades has a page here at all, which is a judgement about search demand rather than about the trades: several are covered under use cases instead, because the product problem there is bigger than the identity one.

The convergence finding here is also the evidence in when a brand becomes a system: colour individuates and typography does not, so the room to differ is in the description rather than in the look.