Typeface
Tool
A single coordinated design for the shapes of letters, chosen to decide what form the text on a screen takes.
Why it exists
Most of what a user actually faces on a product screen is text. Settle on nothing, and each screen ends up using whatever was typed into it, while any place left unspecified falls back to whatever default the device happens to carry.
Pagedeck's early screens worked that way. The document editor, the settings page, and the notification emails all rendered in different letterforms, and within a single screen the buttons and the body text disagreed. Two interviews produced the comment that it didn't feel like one product, though neither participant could name what was different. The sharper problem was elsewhere: in verification codes the digit 0 and the capital letter O were indistinguishable, and mistyped codes generated more than ten support tickets a month.
What it solves
Making text readable is the first job; making separate screens read as one product is the second.
- Legibility. Check first whether 0 and O, and lowercase l and the digit 1, are told apart. In a product that displays verification codes, amounts, and account IDs, that one distinction moves the support ticket count.
- Consistency. Pick one and use it across every screen and email, and the screens read as a single product. When it varies, users register the result as messy without being able to point at the cause.
- Material for hierarchy. Count how many weight steps the family carries. Separating body text, subheadings, and emphasis needs three, so a typeface with only two font weights is hard to build a product on.
- Impression. The same sentence reads stiff or soft depending on the letterforms. This comes last in the order, though — text nobody can read conveys no impression at all.
A product in a script with thousands of characters, such as Korean or Chinese, has one more thing to check. A full set is far larger than a Latin one, so it is common for a typeface to cover only the frequent characters and stop. Typing real names and addresses into a candidate reveals whether the shapes collapse on the rarer ones. What substitutes when the chosen typeface isn't on the user's device is decided by the font fallback.
How it sounds at work
"We can't use this one on the verification screen — 0 and O are identical."
It failed the legibility check, so pick another.
"There are only two weights, so headings and body text don't separate."
Not enough weight steps to build a hierarchy.
"The shapes break when a name has an uncommon character in it."
The typeface only covers part of the character set.
When Noa picked one, the same screen got built four times, once per candidate. The screen chosen was the payment notice, because numbers, a Latin account ID, and a name all appear on it together. Two candidates dropped out on the spot because 0 and O didn't separate, and of the two left, the one carrying four weights won. Taste only got to operate between those final two.
Where to go next
- Typography
see where the typeface sits within text design as a whole
- Font weight
the next value, used to build hierarchy inside the chosen family
Last updated 2026-09-16