Building a Brand Icon Set That Reads at Any Size
On the second game I worked on, the inventory icons were drawn at 128 pixels and looked excellent in the art review.
On the second game I worked on, the inventory icons were drawn at 128 pixels and looked excellent in the art review. They shipped at 32 pixels on a console UI that most people saw from across a living room, and in the first week of player feedback three separate posts asked why the health kit and the ammo box were the same picture. They were not. They just had the same silhouette once the detail collapsed.
I now work on brand identities for businesses with fourteen employees rather than studios, and the same failure shows up every month. A designer delivers a set of twelve lovely icons, they go into the site navigation at 18 pixels and onto a storefront window at two feet across, and half of them stop meaning anything at one end or the other.
Why most small brand icon sets fall apart
Four causes account for nearly all of it, and none of them are about drawing talent.
The set was drawn at one size. An icon designed at 200 pixels gets detail it cannot carry down. A game HUD team solves this by designing at the smallest intended size first and only then scaling up, which is backwards from how most logo work goes and is the single most useful habit to steal.
The stroke weights drifted. Two icons at 1.5 pixels, three at 2, one at 2.5 because the shape felt thin. Side by side in a navigation bar, the inconsistency reads as sloppiness long before anyone can name what is wrong.
The optical volume is uneven. A circle, a square and a triangle drawn to the same bounding box look like three different sizes, because the eye measures area rather than extent. The circle needs to overshoot its box by a couple of percent and the triangle needs more.
And the metaphors are inconsistent in kind. A set that mixes objects, actions and abstractions, so a camera next to an arrow next to a swirl meaning creativity, gives the viewer no pattern to learn. HUD icons are almost always nouns for that reason.
The constraints worth setting before you draw
Pick the grid first. I use 24 by 24 with a 1 pixel keyline inset all round, which leaves a 22 pixel live area and keeps every icon optically centered in a 24 pixel slot. If your smallest real use is 16, use a 16 grid and accept that each icon gets one idea and nothing else.
Pick one stroke weight and hold it everywhere, including the inside of counters and the ends of lines. At 24 pixels, 2 pixels is the weight that survives both a retina screen and a cheap laser printer. Pick one corner radius for the outer shapes, usually 2, and one for the inner details, usually 1 or none.
Decide whether the set is stroked or filled before anything else, because mixing them in a single set never looks deliberate. A stroked set is lighter and better for utility. A filled set reads at distance and is what you want on signage and on a vehicle. Plenty of brands need both, in which case draw the stroked set and derive the filled one, not the other way round.
Finally, cap the vocabulary. Three geometric moves, say a circle, a rounded rectangle and a 45 degree diagonal, and every icon in the set built from those. This is the part that makes twelve drawings look like one family, and it is the same thinking as the constraints in a brand color palette that has rules rather than swatches.
The advice I argue with in client meetings
Two pieces of standard guidance cost small businesses real money, and the second one is the expensive one.
The first is that custom icons are a brand necessity. For most of a site, they are not. Nobody has ever chosen a plumber because of the shape of their hamburger menu. Search, cart, menu, close, phone, chevron: these are conventions, and inventing your own version makes your interface slightly harder to use in exchange for a distinction nobody notices. Use a well built open library for the utility layer, and spend the budget on the six to eight icons that carry something only your business does, which is usually your services or your process steps.
The second is that icons should be derived from the construction of the logo. This sounds disciplined and it is usually a trap. A logo is drawn to work at one commanding size and often carries a signature detail: a thin inline, a notch, a custom curve. Scale that detail into a 20 pixel icon and it turns into mush or, worse, into a dark blob that pulls the eye away from the content. The honest relationship is shared personality and shared geometry, not shared detail. Inherit the corner radius, the stroke feel and the angle vocabulary. Leave the signature flourish in the logo, where it has room.
I will push back on myself here too. There is one case where deriving directly works, which is when the logo is itself a simple geometric mark with no fine detail. If your mark is three stacked bars, by all means let the icons breathe the same air.
Testing, in the order that catches the most problems
- Export the whole set as a single PNG at 16 pixels and look at it on your phone. Anything you cannot name in under a second is not finished. This one test kills about a third of a first draft.
- Fill every icon solid black and squint. If two icons share a silhouette, as my health kit and ammo box did, one of them needs a different outer shape rather than different interior detail.
- Print the sheet at actual size and photocopy it twice. Our brand strategist tests logos this way and it works identically on icons: the second generation copy shows you which hairlines are structural and which were decoration.
- Convert to grayscale. Any icon that only works because of color is relying on something your print vendor, your fax-era client and roughly eight percent of men will not give you.
- Show the set to three people who have not seen the brief and ask them to name each icon with no prompt. Write down the words they use. If two people say document for the icon you drew as an invoice, the problem is the drawing, not the viewers.
- Place them where they will live: a real navigation bar, a real price list, a real van door mockup. Icons judged on an artboard in a neat grid pass tests they fail in context.
Budget an afternoon for this. On the last set I shipped, twelve icons, the tests sent four back for redrawing and two of those were ones I had been pleased with.
Delivery and the file hygiene that saves the client money
Hand over SVGs with strokes expanded to paths, because a developer who scales a stroked SVG without the right attributes gets a hairline at one size and a slab at another. Name files in a flat, predictable scheme: icon-delivery.svg rather than Icon_Delivery_FINAL_v3.svg. Include a 24 and a 48 pixel PNG of each for people pasting into documents, and one sheet image of the full set for the style guide.
Write down the rules in two or three lines each: grid size, stroke weight, radius, the three geometric moves, and what to do when a new icon is needed. Six months from now somebody else will add the seventeenth icon, and those lines are the difference between a family and a pile. That page belongs next to the type and color pages in a brand style guide a small business will actually use, and it should sit beside the same decisions you made about brand fonts that work everywhere, since icon weight and body text weight are read together.
FAQ
How many icons does a small business brand need?
Six to ten custom ones is plenty, covering the services or process steps that are specific to you. Everything else, the interface conventions, is better taken from an open library that already has hundreds of consistent glyphs.
Should brand icons match the logo exactly?
They should share geometry and personality, not detail. Inherit the corner radius, stroke feel and angles. A signature flourish that makes the logo work at large sizes usually turns into noise below about 32 pixels.
Stroked or filled icons for a small business?
Stroked for screen and documents, filled for signage, vehicles and anything read from more than a few feet away. Draw the stroked set first and derive the filled version, and never mix the two styles inside one set.
Is it worth paying for a custom icon set?
For the handful that describe your actual offer, yes, because those are the pictures that end up on your pricing page and in your proposals. For search, menu and close, no. Spending a custom budget on interface conventions is the most common way I see icon money wasted.
More in Brand Identity
Brand Identity
Designing a Brand Mascot That Does Not Age Badly
A bakery owner came to us last spring with a mascot she had paid 1,400 dollars for: a plump, cheerful badger in a baker's hat,...
Brand Identity
Brand Photography for a Small Business Without a Big Budget
Your logo is consistent, your colours are set, and then the website, the social profiles and the brochure are filled with a mix...
Brand Identity
Social Media Branding for a Small Business on Every Profile
A customer finds your bakery on Instagram, then searches for you on Facebook and lands on a page with an old logo, a different...