A category describes. It does not decide.
Sans, serif, display, script, grotesque, neoclassical or monospaced are useful terms. They place a typeface within a tradition, describe aspects of its construction and help organise a catalogue. None of them, on its own, explains why that typeface belongs in a project.
“We chose a sans because the brand needs to feel contemporary” is not yet a decision. It is a broad association that could fit thousands of families and still fail in the first paragraph, a narrow interface, a name with diacritics or a table full of figures. The professional question is not which category sounds appropriate, but what the typeface must do and how it behaves when asked to do it.
Selection is a question of fit: the relationship between voice, function, behaviour, system and context. These are not isolated stages. They correct one another. An expressive voice may need discipline at small sizes; an extensive family may be unnecessary for a short campaign; a typeface that commands a poster may lack the language coverage, web performance or licensing required by an international product.
Start with voice, but write it as a relationship
Voice is not a row of adjectives attached to a moodboard. “Elegant”, “friendly” and “powerful” remain too open unless they are translated into observable tensions. Frame them as relationships: institutional without becoming rigid; direct without aggression; editorial without nostalgia; technical without anonymity; expressive without losing control.
That wording forces comparison. Do open forms create clarity or distance? Does contrast bring articulation or fragility? Does width give the identity room or scatter the text? Does the italic introduce direction, a secondary voice or more emphasis than the content can support? These answers are design interpretations, not universal psychological properties. They need to be tested with the people who write, read, approve and maintain the system.
Research into familiarity and reading argues for restraint. Exposure can influence performance, and unfamiliar letterforms can affect preference, but familiar is not automatically better and new is not automatically difficult. A voice needs enough distinction to be recognised without demanding effort that the function cannot justify.
Define the job before comparing forms
A typeface does not work “well” in the abstract. It works for something. Before opening a catalogue, a project should be able to complete this sentence: we need the typeface to…
- organise an editorial hierarchy of titles, text, notes and data;
- keep one brand voice across screen, print and motion;
- remain clear in navigation, controls and short messages;
- concentrate information within narrow measures;
- sustain extended reading;
- command headlines or covers;
- distinguish levels without constant reliance on colour;
- cover the languages and signs used by the actual content.
The function changes the test. Signage involves distance, viewing conditions, proper names and speed of recognition. A report depends on figures, tables, notes, hierarchy and reading continuity. A responsive identity needs transitions across scales, channels and teams. A campaign may only need a sharply defined display voice, provided there is a compatible solution for informational text.
Separate letter recognition from reading in use
As a working distinction, legibility concerns how readily letters, signs and words can be distinguished; readability concerns continuous composition: measure, size, leading, spacing, contrast, structure and content. The distinction prevents a common mistake — attributing to the typeface what the composition has caused, or the reverse.
The evidence does not support categorical shortcuts. Arditi and Cho found small, conditional effects in a controlled study of serifs; this does not show that every serif is more legible, or that every sans is. Studies involving readers with dyslexia also show interactions between family, spacing, style and preference. A “dyslexia-friendly” label cannot replace evaluation with users.
Tests should therefore keep the content stable and change one variable at a time. Comparing families at different apparent sizes, with unequal leading or different line lengths creates a false winner. First equalise the perceived task: apparent height, measure, optical weight and contrast. Then observe confusions, texture, return sweeps, fatigue, hierarchy and error.
Digital accessibility widens the test. WCAG does not prescribe a particular family, but it does require content to tolerate text resizing and altered spacing without breaking, and it defines contrast requirements for text against its background. An accessible choice is not a font name. It is a composition that can adapt without losing information or function.
Test rhythm, density and scale with real content
A typeface is not experienced one letter at a time. Repetition builds texture: the frequency of verticals, the opening of counters, the weight of punctuation, the sequence of capitals and lowercase, and the pressure between words and lines. That rhythm can calm, accelerate, compress or fragment a page.
Density is not width alone. Character proportions, counter size, weight, spacing and measure all contribute. A condensed design may solve navigation or narrow headlines, but it must be tested with the longest words, not only short samples. An extended design may stabilise a wordmark while multiplying line breaks in a column.
Nominal size is not enough either. Two typefaces at the same size can have different apparent heights, contrast and leading requirements. Where optical sizes or an opsz axis exist, inspect what actually changes: detail, contrast, spacing, proportions or robustness. The axis describes a capability; it does not guarantee better reading by itself.
Test content should include real headlines, extended paragraphs, navigation, buttons, figures, dates, prices where relevant, abbreviations, place names, punctuation, small capitals when needed and every intended language. Hamburgefonts may expose construction. It cannot reproduce the semantic and orthographic pressure of a product.
Evaluate hierarchy as a system, not a collection of weights
A family with many styles does not automatically provide a resolved hierarchy. The issue is whether its differences are visible, useful and repeatable. Can Regular and Medium be distinguished on the actual medium? Does Bold create emphasis without closing counters too far? Does the italic perform a semantic role while preserving continuity? Do widths alter density without appearing unrelated?
Build one minimal page or screen with every required level: title, standfirst, body, quotation, note, label, datum and action. If the hierarchy only works after colour, rules and boxes are added, the type may be contributing less than expected. If every level needs a different family, the system may gain variety while losing maintainability.
Variable fonts extend control. OpenType permits continuous variation along defined axes and can adjust outlines, metrics and positioning. This can answer a measure or hierarchy with precision, but it introduces more decisions too. A mature system does not hand over sliders without rules: it defines useful positions, names, applications, limits and fallbacks.
Check languages, signs and OpenType features
“Extended Latin” or a language count is not enough evidence. Coverage must be checked against content, orthographies and required signs: diacritics, localised punctuation, currencies, symbols, numerals and case pairs. Complex scripts also depend on shaping, anchors, mark order and the composition engine.
OpenType features should be evaluated by need. Tabular figures may be essential for data; true small capitals for publishing; fractions, superscripts or localised forms for particular content. A long feature list adds no value if the application does not expose it, the team cannot govern it or the languages have not been tested.
The test should record what exists, where it works and how it is activated. If a feature or coverage claim has not been verified in the build, mark it N/D. A catalogue cannot stand in for the actual font resource.
Include implementation, licensing and maintenance
The selected typeface has to reach the medium. On the web, this includes authorised formats, file weight, safe subsetting, @font-face, face selection, axis declarations, loading strategy, fallback and what happens while the font arrives. WOFF2 offers standardised compression for web font delivery, but it does not remove the need to decide how many files, ranges and characters should be served.
Applications and documents raise questions of real compatibility, embedding, rendering, export and delivery environments. Print introduces size, screening or resolution, stock, ink, gain and proofing. There is no single “works”. There is a production chain that must be tested end to end.
Licensing is also a design requirement. It must cover the beneficiary, channels, embedding, web or app use, supplier workflows and expected duration. Permission cannot be inferred from a file format or from what software technically allows. The current EULA is the authority; an excluded or ambiguous use remains pending.
Longevity depends on governance. Can the team access the files and documentation? Are there enough styles for foreseeable growth without buying unnecessary complexity? Does fallback preserve hierarchy? Can an update be tested without breaking historical documents? Is there a route to add languages or adapt forms if the project expands?
Pairing: assign responsibilities before seeking contrast
Pairing is not a search for two personalities that “get along”. It is an allocation of responsibilities. One typeface may sustain reading while another provides display contrast; one may organise an interface while another articulates editorial content. A single family with sufficient widths, italics or axes may be the stronger answer.
A pairing should add clarity or range, not decorate uncertainty. Test both families within the same hierarchy and content, then ask: what job does each one perform? Is the difference clear without becoming competitive? Do proportions or tensions allow continuity? Are roles duplicated? Does the extra loading, licensing and maintenance earn its place?
Three decision scenarios
Editorial system. A publication needs extended text, titles, quotations, notes and figures. Selection starts with paragraphs and full pages, not a cover word. Compare apparent size, rhythm, italics, figures, signs and languages. The cover may use the same family or a secondary display face if responsibilities remain clear.
Digital product. An interface needs short labels, messages, tables and states across different screens. Test small sizes, zoom, customised spacing, fallback, loading and localisation. A variable family may reduce fragmentation and tune hierarchy, but only when axes are correctly implemented and documented.
Cultural identity. A system needs recognition across posters, signage, publications and web. Voice carries more weight, but it does not cancel function or rights. Check whether character survives at reduced scale, whether informational text has a solution and whether files and licensing cover the institution, its suppliers and its channels.
These are test models, not client cases or measured outcomes.
A ten-step working process
- Write a decision sentence: “We need a ___ voice that can ___ in ___ for ___.”
- Inventory the real content, languages, signs, levels and media.
- Separate mandatory requirements, preferences and risks.
- Shortlist by documented capability, not appearance alone.
- Set the same content under comparable conditions.
- Test the edges: smallest size, highest density, longest word, most complex figure and weakest intended medium.
- Build hierarchy, pairing and fallback as one system.
- Validate with users and owners where the consequence justifies it.
- Verify files, features, implementation, licensing and maintenance.
- Record why the choice was made and which conditions would invalidate it.
The final question is not “Do I like this typeface?” Taste belongs in visual direction and does not need to be denied. A professional decision must answer more: does the typeface express the voice, do the work, behave with the content, sustain the system and live within its context?
Category describes. Behaviour decides.
CTA: Explore the catalogue and test each family with the words, hierarchies and constraints of your project.
Sources and further reading
- W3C WAI, WCAG 2.2: text spacing, resize text and minimum contrast.
- Microsoft Typography, OpenType Font Variations Overview.
- Unicode CLDR, Language Support Levels.
- Aries Arditi and Jianna Cho, “Serifs and font legibility”, 2005.
- Sofie Beier and Kevin Larson, “How does typeface familiarity affect reading performance and reader preference?”, 2013.
- Tanja Medved, Anja Podlesek and Klementina Možina, “Influence of letter shape…”, 2023.