How a logo does its work

Take any mark you think you know well. Cover the name and leave the shape. Then squint until the shape goes soft, and imagine it stitched in a single colour onto a jacket pocket, or pressed into metal, or shrunk to the size of a full stop in the corner of a browser tab.

Most designs fall apart somewhere in that sequence. The handful that survive it are, almost without exception, the ones you can draw from memory. That isn't a coincidence — it's the same property being tested twice.

A logo is a small piece of engineering with a narrow job. Nearly everything people argue about when a new one appears — whether it's beautiful, whether it looks modern, whether it captures the spirit of the organisation — has very little to do with whether it works.

A mark doesn't describe, it identifies

This is the assumption behind most bad briefs, so it's worth killing early.

People expect a logo to say what the business does. It doesn't have to, and the harder it tries, the worse it usually performs. Description pushes a mark towards the generic, and generic marks are both forgettable and legally feeble. A shield on an insurer tells you nothing, because every insurer in the world can draw a shield. A cog on an engineering firm is the same problem wearing a different hat.

Trademark law arrived at this conclusion independently, from the other direction. Marks are ranked by how distinctive they are, and the scale runs roughly: generic, descriptive, suggestive, arbitrary, fanciful. Generic terms can't be protected at all — nobody gets to own the word "bread" for bread. Descriptive ones are protected only after they've been used long enough that the public associates them with one source. Arbitrary and invented marks get the strongest protection from the moment they're used, precisely because they carry no meaning of their own.

So the legal system and the design problem agree. The emptier the mark, the more of it you own.

That's uncomfortable if you think a logo should communicate. It shouldn't, mostly. It's a handle, not a message. The company spends decades filling it.

Heraldry solved most of this by 1200

Europe ran a full-scale identity system three centuries before printing, under conditions far more demanding than a phone screen: a moving target, at distance, seen by people who couldn't read, where getting it wrong could be fatal.

Coats of arms appeared on shields and surcoats in the twelfth century and hardened into a regulated system fast. England's College of Arms was formally founded in 1484 and still operates. The rules that emerged are, in effect, a design specification, and modern designers rediscover them constantly without knowing where they came from.

The most useful is the rule of tincture. Heraldic colours are sorted into two groups — the metals (or, meaning gold or yellow, and argent, meaning silver or white) and the colours (gules red, azure blue, sable black, vert green, purpure purple). The rule says you don't put colour on colour, or metal on metal. Red on blue is forbidden. Red on white is fine.

The reason is contrast. Two dark colours side by side have similar lightness, so at distance, in poor light, or in the eye's low-resolution peripheral vision, the boundary between them dissolves and the shape stops reading. Put a light against a dark and the edge survives almost any abuse. Designers today check this by converting a logo to greyscale and looking at whether the parts still separate. That's the rule of tincture, restated in a colour picker.

Heraldry also produced blazon, which is stranger and cleverer. A coat of arms isn't defined by a drawing — it's defined by a written description in a compressed technical language, and any competent artist can reconstruct a legal-quality version from the text alone. The design is specified, not depicted. Brand guideline documents are a much clumsier attempt at the same trick.

The eye takes the outline first

Human vision is far worse than it feels. You have sharp sight across a region about the width of your thumbnail at arm's length. Everything outside that is blurred, colour-poor and low-detail, and your brain covers up the gap so smoothly that most people go a lifetime without noticing.

Which means a logo is nearly always being seen badly.

It's in peripheral vision on a shelf, or at speed on a passing lorry, or half-glanced at while you're looking for something else. Under those conditions the visual system falls back on the crudest information available: overall shape, overall lightness, and the presence or absence of a few strong edges. Fine detail contributes nothing at all.

This is why the silhouette test is the only design test that reliably predicts anything. Fill the mark solid black, remove every internal line, shrink it, and see whether it's still identifiable. Marks built from one closed shape, or from a small number of chunky elements arranged in a distinctive envelope, come through. Marks built from thin lines, fine gradients, tight lettering or interlocking detail turn into a grey blob at about forty pixels and stay there.

The Gestalt psychologists mapped the underlying grouping rules in the 1910s and 1920s, and they still hold: closure, where you complete an interrupted contour; figure–ground, where you decide what's object and what's background; proximity and similarity, where nearby or alike elements are read as a unit. Edgar Rubin's vase-or-faces figure, published in 1915, is the standard demonstration that figure–ground is a decision your brain makes rather than a fact about the image. Any designer playing with negative space is working directly on that machinery.

Colour is the fastest channel and the least reliable

Colour gets recognised before shape. It survives distance, blur, motion and terrible printing better than any other property, and it's the reason you can find a familiar package in a supermarket aisle without reading a single word.

It's also the property people are worst at remembering accurately.

Ask someone to pick their bank's exact blue out of a row of blues and they'll usually get it wrong, but show them the right one next to a near-miss and they'll spot the imposter instantly. Recognition is strong; recall is weak. That asymmetry is why colour specifications get policed so obsessively — the association only holds if it's consistent, because consistency is doing all the work.

Courts have accepted that a colour alone can function as a trademark. In 1995 the United States Supreme Court decided Qualitex v. Jacobson Products, a dispute over the green-gold colour of dry-cleaning press pads, and held that colour can be registered where it has acquired distinctiveness and isn't functional. That second condition matters: you can't monopolise a colour that does a job, like safety orange on equipment meant to be seen.

Colour meaning, meanwhile, is mostly learned rather than innate, and the associations differ across cultures. White reads as mourning in parts of Asia and as weddings in much of Europe. Red carries luck, danger, celebration or debt depending entirely on where you are and what's around it. Claims that a particular colour reliably produces a particular emotion in buyers are, in the main, marketing folklore with very thin evidence underneath.

It has to survive being made badly

Every logo that's lasted has been through a long series of hostile manufacturing processes, and its shape carries the scars.

Think about what a mark actually has to endure. Embroidery, where thread has a minimum stitch width and fine detail simply can't exist. Etching into glass. Casting in metal. Screen printing onto fabric with ink that spreads. Vinyl cut for a van, where every internal hole needs a bridge to hold it in place. A one-colour fax, historically. A monochrome engraving on a pen barrel. A favicon at sixteen pixels square.

The constraint set has changed over the centuries but never gone away, and it explains the direction of travel that people mistake for fashion. Marks get simpler over time because reproduction keeps getting cheaper and more varied, not because designers got lazier. A late Victorian trade mark could afford a great deal of engraved detail, since it was going to be printed once, well, on paper, by a specialist. A mark today has to work on a watch face and on a stadium roof and in a chat window, and the only shapes that clear all three are simple ones.

Then there's the part that sounds like superstition and isn't. Optical correction is real and measurable: a circle drawn to exactly the same height as a square next to it will look smaller, so it's drawn slightly larger — the same overshoot that type designers build into every round letter. Horizontal strokes are usually drawn thinner than vertical ones or the letterform looks top-heavy. Spacing gets set by eye rather than by number, because mathematically equal gaps don't look equal. None of this is visible in the finished mark, which is the point.

Marks can be lost by being too successful

Trademark protection isn't ownership of a word or a shape. It's a right to stop other people using something confusingly similar in the same line of business, and the central legal test is whether ordinary buyers are likely to be confused. That's why an identical name can sit peacefully on a plumbing firm and a rock band.

The failure mode is unusual and slightly cruel. If a brand name becomes the ordinary word for the product category, it stops functioning as an indicator of source and can lose protection entirely. Lawyers call it genericide.

Escalator started as a trade mark of the Otis Elevator Company and was ruled generic in 1950, after the company itself had used the word descriptively in its own advertising. Bayer lost the American rights to aspirin in 1921 for much the same reason. Thermos, trampoline, kerosene, zipper and heroin all began as brand names and ended as nouns.

Which is why brand owners write those slightly ridiculous instructions telling everyone to use the name as an adjective and never as a verb. It reads as pedantry. It's a defence.

Some marks are protected by treaty

The strongest legal protection any symbol has isn't commercial at all.

The red cross on a white ground was adopted at the 1864 Geneva Convention as the protective emblem for medical services in wartime. It's the Swiss flag with the colours reversed, a nod to the convention's origins, and it was chosen partly because it was easy to draw, easy to identify at distance, and used no existing national symbol. It's protected by international humanitarian law and by domestic statute in most countries, and misuse in wartime is a war crime.

The Ottoman Empire objected to the cross in the 1870s and used a red crescent instead, which was formally recognised in 1929. A third neutral emblem, the red crystal — an open red square standing on one corner — was added by a protocol adopted in 2005 for organisations that couldn't accept either of the others.

All three follow the same design brief the medieval heralds worked to. A single flat colour on white. No fine detail. A distinctive outline recognisable at a glance and at distance by people who can't read the language. It has to be identifiable through smoke, from the air, at speed, by someone with every reason not to look carefully.

The meaning shows up afterwards

Here's the part that irritates clients most, and it's the one that best explains why the arguments about new logos are so predictably wrong.

A mark has almost no meaning when it's launched. It's an empty container. Whatever feeling people eventually attach to it comes from years of experience of the thing behind it — the products, the service, the price, the staff, the associations — flowing backwards into the shape until the shape starts to feel like it always meant that.

The mere-exposure effect, which Robert Zajonc documented in 1968, does a lot of the quiet work. Repeated exposure to a stimulus increases liking for it, with no reasoning involved and no memory of the exposures required. A mark that felt arbitrary and slightly ugly in year one feels inevitable in year fifteen, and people will construct elaborate explanations for why it was always the right choice.

Which means the day-one reaction to a redesign is almost worthless as a prediction. New marks are routinely mocked. Some of the mockery is deserved and some isn't, and the noise mostly measures how attached people were to the previous one rather than anything about the new one.

The genuine risk in a redesign isn't ugliness. It's discarding recognition that took decades to accumulate — throwing away a silhouette people already know in exchange for one they'll have to learn from scratch. That's a real cost, and it's rarely on the invoice.

What a logo can't do

It can't fix a bad product. It can't make people trust an organisation they've been given reasons not to trust. It can't explain a complicated business, and if it's trying to, someone has misunderstood the brief.

It can do three things. It can be found quickly in a cluttered visual field. It can be told apart from its neighbours. And it can hold, without leaking, whatever reputation gets poured into it over time.

Judge one on that and the assessment gets much simpler. Shrink it. Print it in black. Show it to someone for half a second and ask what they saw. Stitch it on a shirt. If it comes through all of that still looking like itself, the design is doing its job, and whether you personally like it is a separate question that nobody needs answered.