# How many icons does a product actually need?

> Most products use fewer than 120 icons. Here is the coverage checklist that finds the gaps, and why a large set still helps even when you ship a small one.

URL: https://semanticicons.com/blog/how-many-icons-does-a-product-need
Published: 2026-11-04
Updated: 2026-11-04

Icon sets advertise totals. Two thousand icons. Five thousand icons. The number tells you almost nothing, because you will use about a hundred of them, and the question that decides your day is whether the hundred you need are all there.

## Count your own screens first

Open the product and list every icon in use. Most teams are surprised by how short the list is.

| Product | Typical count |
| --- | --- |
| A single purpose tool | 30 to 50 |
| A standard web app | 60 to 120 |
| A large product with many surfaces | 150 to 300 |
| An editor or a design tool | 300 to 600 |

If you are above 300 and you are not building an editor, you probably have duplicates: three arrows that differ by two degrees, four checks, two houses.

## The coverage checklist

These are the groups every interface needs. Walk them and mark the gaps.

**Navigation.** Home, back, forward, up, down, menu, close, expand, collapse, external link, breadcrumb separator.

**Actions.** Add, edit, delete, copy, duplicate, move, share, download, upload, print, refresh, undo, redo, search, filter, sort.

**State.** Success, error, warning, information, pending, locked, unlocked, visible, hidden, favourite, pinned.

**Objects.** File, folder, document, image, video, audio, archive, link, calendar, clock, tag, user, group, location.

**Communication.** Mail, message, comment, notification, phone, send.

**Commerce, if relevant.** Cart, card, receipt, tag, truck, return.

That is about sixty, and it covers most of what a normal product does. The rest are specific to your domain, and they are the ones a general set is most likely to miss.

## The gaps that cost you

A general set is nearly always complete on navigation and actions. It fails on domain nouns.

Building for a veterinary clinic? You need animals, and most sets carry a dog, a cat and nothing else. Building for a kitchen product? You need a colander, a whisk, a chopping board. Building for construction? You need a wheelbarrow, a hard hat, scaffolding.

This is why a large set still matters even though you will ship a hundred icons. Not because you want two thousand, but because the long tail is where your specific need lives. We built [3000](/) for exactly this reason: the interesting coverage is not the sixtieth icon, it is the two thousandth.

## How to check a set before you adopt it

Do not browse the grid. Browsing rewards sets with attractive common icons.

Instead, write down the twenty most specific things your product talks about, in your own words, and search for each one. Count the misses. Twenty specific searches tells you more than an hour of scrolling.

While you are there, search in sentences rather than nouns. If the set cannot answer *the meeting is about to start*, its index is a file name list, and you will be guessing names for as long as you use it. See [finding an icon without knowing its name](/blog/find-an-icon-without-knowing-its-name).

## Keep your own list short

Adopting a large set does not mean shipping one. Import what you use, and let tree shaking drop the rest. A hundred icons at about 300 bytes each is 30 KB before compression, which is nothing.

What you should avoid is drawing your own alongside a set. A drawn-in-house icon beside a set icon almost never matches on weight, corner radius or optical size, and the mismatch is visible even to people who cannot name it.

## Questions people ask

### Is a bigger icon set slower?

Not if it tree shakes. The set size affects your bundle only through what you import. Check the built output rather than trusting the package size.

### Should we draw our own icon set?

Only if you have a designer who will own it for years. A set is a contract: one grid, one stroke, one hand. Fifty icons drawn by four people over two years do not form a set, they form a pile.

### What if a set is missing one icon we need?

Ask whether the set can generate one to the same contract, or whether you can commission it. A single mismatched icon in an otherwise consistent set is more visible than you expect.

### How do we stop the count from growing?

Review additions against the existing set. Most requests for a new icon are answered by one already there under a different name, which is a search problem rather than a coverage problem.