What Makes a Good Logo? 7 Things to Check

Most logo feedback is taste dressed up as judgement. Someone says a mark feels cheap, or modern, or too corporate, and nobody can act on it. But a good deal of what makes a logo work is not taste at all. It is measurable, and it fails in predictable ways.

Here are seven things worth checking, in the order they tend to break. They are the same seven dimensions our logo review scores, five of which are calculated directly from the artwork rather than judged.

1. Legibility at small sizes

The first thing to fail is nearly always stroke weight. A mark drawn with hairline strokes looks refined at 400 pixels and disappears at 32. The test is simple: render it at 32 pixels and see whether the strokes still read as strokes rather than grey suggestion.

Contrast against the background matters just as much. A pale grey mark on white passes on a designer's screen and vanishes on a phone in daylight. If you cannot state the contrast ratio, you have not checked it.

2. Scalability

Scalability is not the same as legibility, though people use the words interchangeably. Legibility asks whether you can see it. Scalability asks whether it is still the same shape.

Complex marks lose their identity as they shrink. Three overlapping elements become one blob at favicon size. The honest test is to put the mark at 128, 64 and 32 pixels side by side and ask whether the 32 still reads as the same logo. If the answer needs a caveat, it is a no.

3. Balance and alignment

Optical balance is not mathematical centring. A mark can sit dead centre in its bounding box and still lean, because visual weight is not distributed evenly through the shape.

You can measure this: find the visual centre of mass and compare it with the true centre of the artwork. Sideways lean is worse than vertical, because horizontal misalignment is far more obvious once a logo sits next to text in a header.

4. Colour harmony

Count distinct hue families, not colours. Four shades of one blue is one colour, and that is fine. Four unrelated hues is usually a sign that nobody made a decision.

The number itself matters less than whether the colours have a relationship and enough contrast between them to survive being reproduced badly, which they will be.

5. One-colour behaviour

Every logo eventually gets used in a single colour: an embosser, an engraving, a fax from an accountant, a sponsor wall, a stamp. If the mark relies on colour to separate its parts, it collapses into a silhouette the moment colour is removed.

Convert to greyscale and see whether the dominant areas are still distinguishable. This one test catches more real-world failures than any amount of discussion about whether the palette feels premium.

6. Originality

Here we leave measurement behind. Originality is a judgement, and it is category-relative: a swoosh is unremarkable in sportswear and distinctive in accountancy.

The practical question is not whether the concept has ever been used, because everything has. It is whether someone in your category, looking at your mark for one second, would attribute it to you rather than to a competitor.

7. Accidental readings

The last check is the one people skip and later regret. Look at the negative space. Rotate it. Squint. Show it to somebody who knows nothing about the business and ask what they see first.

Unfortunate readings are almost never visible to the person who designed the mark, because they know what it is supposed to be. That knowledge is exactly what stops you seeing what everyone else sees.

What this does not tell you

Passing all seven does not make a logo good. It makes it competent, which is a lower bar and a necessary one. Whether the mark says the right thing about the business is a strategic question no measurement answers.

What the seven do is separate the two conversations. Once the craft problems are resolved, the remaining disagreement is genuinely about direction, and that is a far more productive argument to have.

If you want to see this applied to finished brands, the case studies show thirteen identities built end to end, and how to critique a logo covers giving this as feedback somebody can act on.

How to run these checks in five minutes

None of this needs specialist software. The whole set can be done with a browser and an image editor.

• Small size. Drop the logo into any editor, export at 32 pixels wide, then view that file at 100 percent. Do not zoom in. Zooming defeats the entire test.
• One colour. Apply a greyscale filter, then push the contrast hard. If two areas merge that should stay separate, you have found a real problem.
• Contrast. Sample the mark colour and the background colour and run both through any WCAG contrast checker. Below 3:1 for a large graphic is a warning.
• Balance. Place the mark inside a square, draw a line through the true centre, and squint. Optical lean is easier to see out of focus than in.
• Accidental readings. Show it to three people who know nothing about the business and ask what they see. Do not explain it first. The explanation is what stops you hearing the answer.

Not all seven matter equally

Treating these as a checklist of equals is a mistake. Legibility and scalability decide whether the logo functions at all, so they carry more weight than the rest. Balance and colour harmony are quality signals rather than pass-or-fail conditions. One-colour behaviour matters enormously for some businesses and barely at all for others, depending on whether anything is ever printed.

In our own scoring the weights run from 1.2 for legibility and scalability down to 0.8 for one-colour use and accidental readings, which is roughly the order they cause real problems in practice.

What this looks like on real brands

The Strike9 streetwear mark is a useful example of the first two criteria working together: heavy strokes and high contrast mean it survives being embroidered on a cap and reduced to an app icon without losing its shape. Weight is doing the work, not detail.

The Nabidh Arabic calligraphy wordmark is the opposite problem handled well. Calligraphic scripts carry fine strokes by nature, so the small-size question is the hard one, and the answer is a deliberately simplified form for small placements rather than a scaled-down version of the full mark.

The general lesson is that criteria trade against each other. Expressiveness costs you small-size performance nearly every time, and the fix is usually a second version rather than a compromise on one.

The failure nobody checks for

One more thing, outside the seven, that causes more real-world trouble than any of them: not having the files. A logo that passes every test but exists only as a PNG on someone's laptop will fail the first time a printer asks for vector, and it will fail again when a developer needs a favicon and a transparent version.

A finished logo is a set of files, not an image. If you are unsure what that set should contain, the brand identity checklist lists it, and logo file formats explained covers which file goes to whom.