Wide and Shallow
Five to nine top-level items, no more than three levels deep on the main path. Go deeper and no one reaches the bottom.
Scanning ten options on one screen is far easier than tapping through four levels in a row. The former is one act of recognition; the latter is four decisions. The five-to-nine range for a top-level menu comes from how much information human short-term memory can handle effectively at once. magic-seven Fewer than five usually means the level shouldn't exist; more than nine means it's time to group.
But the real point is depth. Every extra level loses a batch of people. Not because the click is effort, but because each level demands the user guess "which category does this belong to"—guess wrong once, and they have to back out and start over.
"One guess" sounds light, but it compounds. If each level's guess is 80% right, tapping through four levels succeeds only about 40% of the time—get one wrong and you're back up a level. Users don't blame their own guesses; they decide "this product is hard to use." So if nine items fit on one screen at the top level, don't split them into three layers. Better a fuller-looking screen than making users pay for your categorization logic.
What you'll run into:
- A feature genuinely exists, but users say they've never found it
- Pages below the third level get almost no traffic
- To keep the nav looking clean, you keep tucking things deeper and deeper
What to do when you exceed nine
- Group without adding a level. Split a dozen-plus items into three groups on one screen with dividers or subheadings. Users still scan everything at a glance—no extra tap.
- Cut the low-frequency ones. Check the data first. An item nobody touched in a month belongs in settings rather than the nav.
- Add search. When there genuinely are many items and users need all of them (like settings), a searchable entrance beats careful categorization.
- Switch to a matrix. If content has multiple attributes and users approach it from different angles, replace categorization with filtering.
- Add one more level. Consider it last. It looks the tidiest and costs the most.
The first option is the most underrated. Visual grouping and structural layering are two different things—the former adds no steps, the latter does. Confuse them and the moment you pass nine items, you'll reach for another level by reflex.
Why is "one more level" the most expensive fix? It isn't just one extra tap—it changes the user's mental model of the whole product. What could be seen in a single glance now requires first deciding "which big category does this belong to." And once the layer exists, every future piece of content has to answer "which level does it go in"—a cost paid on every visit. The other four options keep the choice in the user's hands; adding a level hands the judgment to them.
How to fix it when it's already too deep
- Pull the high-frequency pages up. Find the third-level pages with the most traffic in your data, and give each a direct entrance at the first or second level.
- Flatten depth with search. Good search makes hierarchy irrelevant—users jump straight to the target. It's also the fallback when depth genuinely can't be compressed further.
- Add a "recently used" section. Put where the user went last time on the first screen, bypassing the levels entirely.
- Merge levels that hold a single item. A category with only one thing under it is pure waste—fold it up.
Breadcrumbs can't solve depth; they only answer "where am I." A user not getting lost on the fourth level doesn't mean they're willing to tap four times every visit. Wide and shallow describes navigation structure; the priority inside a single page is a separate concern—handle both and the result is neither wide nor messy.
These four fixes usually combine: pull high-frequency pages up as direct entries, tuck low-frequency ones behind search or settings, and put recently visited places on the first screen. Depth isn't a decorative choice in the design—it's a toll every visit pays. Charge it as rarely as you can.