The short version
- Line length decides whether a long page gets finished, and it is set by the layout container rather than the font.
- Most hierarchies carry more levels than a page needs. A level readers cannot tell apart without comparing is one everybody chooses differently.
- The body family deserves more attention than the display family. It carries almost every minute a reader spends on the site.
- A type rule written into a style guide gets relitigated. Put the same rule in a token and it holds in files nobody reviews.
Three decisions in a type system change how a page reads. The length of a line of body text comes first, then the number of hierarchy levels, then which family carries body copy. Everything after that is preference, where most type reviews spend their time.
Which typography decisions actually change how a page reads?
The ones acting on a reader’s eye rather than on their impression of the brand. A line of body text long enough that the eye loses its place on the return costs you readers on every long page. That is a container decision, made in layout, and a type review never touches it.
The other two are how many hierarchy levels exist and which family carries body copy. Neither is a taste question, and both get settled once somebody decides rather than collects opinions.
What each choice changes, and where it is actually set:
| Choice | Decision or preference | What it changes | Where it is set |
|---|---|---|---|
| Line length of body text | Decision | Whether a long page gets finished | The layout container, not the font |
| Number of hierarchy levels | Decision | Whether a page can be scanned | The component set |
| Family carrying body copy | Decision | Every minute a reader spends reading | One token |
| Which grotesque for display | Preference | The mood of the first screen | A variable somebody will change |
| Letter-spacing on capitals | Preference | Nothing a reader could name | CSS, endlessly |
Why is line length the decision people skip?
Because it does not look like a typography decision. It is set by the width of the column the text sits in, so it belongs to whoever built the grid and never reaches a font review.
The symptom is that long pages feel like work, and nobody reports it that way. They report that the article was too long, and it was the same length it had always been. The fix is unpopular because it costs nothing: narrow the reading container and leave the space empty.
How many hierarchy levels does a page need?
Fewer than most systems ship. A level earns its place when a reader can tell it apart without comparing it to its neighbour, and when it appears often enough for that distinction to be worth learning. A level failing both tests still costs, because everyone has to choose between it and the one beside it.
The gaps matter more than the count. Two levels a hair apart in size read as an inconsistency, and a reader who cannot see the difference concludes the page is careless.
Which family should carry body text?
Whichever one holds up at the size and weight your longest page actually uses, tested on that page rather than on a specimen. A family chosen from a specimen sheet was chosen at display size, and type behaves differently inside a paragraph.
The display family attracts more argument than it deserves. It appears on headlines and on the front of a deck, a small share of the reading and most of the opinions.
A type system survives in components and tokens. Written into a style guide, it lasts until the first person who disagrees opens a file.
What makes a type decision stick?
Something that fails when the rule is broken. A style guide states a rule and relies on a person remembering it, which holds until the week somebody is in a hurry.
An audit of 200 published essays on this site found 2,801 em-dashes against a house rule of zero. The rule had been written down and restated in briefs, and nobody had decided to ignore it. The sweep that fixed it had to classify each dash rather than replace them all, because roughly a third joined two independent clauses and a blanket comma would have produced comma splices.
Type rules have the same shape. Put the reading width into the container component and the scale into tokens, and the decision holds in files nobody reviews.
The neighbouring arguments are why a written rule needs something mechanical behind it and designing for readers rather than users.
Frequently asked questions
What are the most important typography decisions in a design system?
Line length of body text, the number of hierarchy levels, and which family carries body copy. Line length is set by the layout container rather than the font, and the body family carries almost every minute of reading.
Does line length affect readability?
Yes, and it is a container decision rather than a font decision. A line long enough that the eye loses its place on the return makes a long page feel like work. Readers rarely name it, and report that the article was too long.
How many heading levels should a website have?
As few as a reader can distinguish without comparing them. A rarely used level still costs, because everyone has to choose between it and its neighbour. The gaps matter more than the count.
Should you choose a body font or a display font first?
Body first. The display family shows up on headlines and the front of a deck, a small share of the reading and most of the opinions. The body family carries every long page, and belongs tested at the size the site ships.
Why do typography rules stop being followed?
Because a written rule relies on memory. An audit of 200 published essays on this site found 2,801 em-dashes against a house rule of zero, written down and restated in briefs with nothing mechanical checking it.
What belongs in a design token rather than a style guide?
Anything you do not want argued a second time: the type scale, the width of the reading container, and the family assigned to body copy. A token applies a decision in files nobody reviews.
Sources
- Chua Network delivery data across 8 client accounts (internal fact bank)
- Chua Network engagement records, anonymized (internal experience bank)
