How to evaluate a typeface

Screen Shot 2016-01-10 at 17.34.56Ten years ago, there were two realistic (legal) choices for typography. Either buy font sets from Linotype, Monotype etc, or use the fonts that came with your software. Back in those days there were design applications worth buying purely for the included fonts. Of course, you could also surf the web for ‘free’ fonts, but these would either be illegal knock-offs, or amateurish affairs with disproportionate letterforms and bad kerning: maybe ok for a poster headline, but not something for serious typography.

How the world has changed. Skyfonts, Typekit, and others offer font subscriptions. FontFont and others often offer free weights of new fonts as a promotional tool. An increasing number of designers are releasing their fonts or some weights of their fonts for free to get themselves noticed. Some of these sets aren’t that useful: Ultralight with Bold Italic is hardly going to allow you to typeset much of a document. Others, most famously Museo by ExLjbris, which you now see more or less everywhere, are designed as super-families and genuinely useful weight sets are offered for free. After using Museo free weights for a branding project, I lashed out on the complete set of weights. Money well spent. Of course, you can still buy fonts the old way, but, even there, the prices have dropped in real terms and the variety is extraordinary.

To many designers, the lure of free fonts is almost irresistible. I’ve known people who have surfed the web for days to find what they want (quite possibly spending more in time than they saved in price).

One question, though. How do you know the fonts you’re buying, hiring or getting for free are any good? You can’t tell from the web-site preview. Almost anything looks good set at 72 point in one line of text.

In this article, we explore seven aspects of typeface evaluation. For QuarkXPress 2015 users, we also include a file that you can use to evaluate any font on your system. This all applies equally to paid-for fonts, subscription fonts and free fonts, though it’s harder to properly evaluate a font if you can’t print it until you’ve paid for it.

The criteria are:

  1. Legibility
  2. Colour, consistency and kerning
  3. Efficiency
  4. Completeness
  5. Compatibility
  6. Legality
  7. Hidden costs


All kinds of studies have been done on typeface legibility over the years. The results are often contradictory, over-influenced by confounding factors, or too narrow to be useful. If you’re interested, there is a magnificent book by Sofie Beier called Reading Letters: Designing for Legibility which goes through all the research and picks out the factors which are genuinely proven. Basically, though, there are three things you can do to ensure legibility.

The most important one is get a load of text and read it. The second is to check the easily confused letters pairs: ecasnuo ijltf OQDCG VYWMKX TIJL FBPETH HNM. Are they distinctive from each other? Third, you can look at the actual forms. At body text sizes, legibility is enhanced by wider letter forms, lower contrast and slightly heavier weights, along with proportionately high x-height and generous white space inside the letters. The optimal stroke weight is about one fifth of the x-height, or slightly less if the strokes are uniform. At small sizes, the eyes perceives irregularities more than at large sizes, so text meant for body copy needs to be fairly regular.

In the accompanying file (see bottom of page), we’ve created a set of nonsense text which is designed to put all the combinations of the easily confused letters together.

By the way, the evidence about whether sans-serif, serif, vestigial serif (like Optima) or even slab-serif is more legible in body text is entirely inconclusive. It has more to do with familiarity than anything. German Black Letter fonts, it turns out, are just as legible to people who are used to them, despite being almost entirely illegible to people who are not.

Colour, consistency and kerning

A page of type has its own colour. Too dark and it’s hard to read, too light and the same thing happens. Very light and very dark text is also hard to set effectively in a layout. If you’re considering a font for general use, any font that requires you to design around it is generally a poor choice. Does the type look consistent on the page? If not, this is usually a sign of poor kerning. As you know, kerning is the adjustment of individual letter pairs which, because of their forms, don’t look quite right with the standard spacing. There are more than 500 letter pairs which normally need to be kerned. This takes a great deal of time and skill. Even some commercially available fonts are poorly kerned, and most of the older, hobbyist, free fonts have little attention paid to kerning. Bad kerning will make every page look like it came out of Microsoft Word (because Word has kerning turned off by default for body text, though you can—and should—turn it on).

In the accompanying file, we’ve created a set of test pages which put all of the difficult pair combinations together, with a number of ways of looking at them, including blurred, which is part of how the eye processes text. Look at the page from a distance, look at it from reading distance. If there are odd gaps in the text, or if it just doesn’t seem to flow, then the kerning is bad, or the letter spacing or letter forms are bad, or all three. In that case, discard the font: it will not look any better when you see your page for the hundredth time. Note that you should actually print out the test file: type displays on your screen differently from how it displays on the page. It may well be fine on paper even when odd on screen (and vice versa). Don’t take the risk.


Is your font efficient? You have no doubt seen the article about the two school students who worked out that the US government could save $millions by changing their font, as it would use less ink. What they hadn’t understood was that fonts of the same point size have differing optical size. The font they recommended was the same at 11 point as another at 10 point in terms of legibility.

‘Official’ font size doesn’t exist. The ‘point size’ may mean the height of the capitals plus the depth of the descenders plus a bit, or it may mean that plus a lot, or it may (for example in Zapfino) be substantially less than the Capital+Descender. In movable metal type, the point size was the height of the ‘body’, which was the lead block on which the glyph was cast or cut. Clearly, this body had to be big enough to accommodate the full height of every single glyph, so it generally would be the descender to cap height, plus a bit. What that ‘bit’ was would vary from foundry to foundry. If the font had full-sized capitals with accents (and traditionally, accents are not printed on capitals), the body would have to be larger, or, rather, for the same point size, the letters would have to be smaller.

Digital fonts have no ‘body’, so the point size is arbitrary at best. However, the optical size of the font is not just the ‘real’ height from the bottom of the descender to the top of the capital. Fonts having a larger x-height proportional to t-height appear larger, which is to say they are optically larger. ITC fonts, for example, whatever you think of their interpretation of Garamond, always have a high x-height, which makes them ‘larger’ at the same point size than those with a smaller x-height. This is linked with the better legibility noted earlier. The trend over the last hundred years has been to proportionately larger x-heights, so a lower x-height also makes a font look more old fashioned, without having to resort to decorative forms, rough types or pseudo-medieval fonts. If you’re setting for an old-fashioned look, great—but remember that your font is going to have to be set larger, possibly with more leading as well, to have the same apparent size.

In choosing a typeface, also consider what it’s going to be printed on. We won ‘brand of the year’ for a re-brand in the UK stationery and office products industry a couple of years ago. Part of the decision on typefaces involved carefully checking every piece of packaging and how text was reproduced on it. You may think you are designing only for black on bright-white 150 gsm matt art paper through offset printing via a 2400 dpi Linotronic. However, if that font then gets deployed onto flexo, thermal, or, heaven-help-you, Risograph, your wonderfully elegant font which seemed perfect down to 6 point may just be a blur of dots. This is especially the case when printing on a coloured background. Black on red is almost illegible at the best of times. I struggled the other day with a product where the designers had decided to reverse white onto red on flexo-printed plastic at (I think) four point. The general rule is six point is the smallest legible size. White on red with a bit of blurring from flexo print would have needed at least eight point.

There’s a page in the evaluation file (below) which will help you check x-height and optical size.


Years ago, I bought the short-lived Deneba Canvas, which came with thousands of URW fonts. The thing is, this was in 1996. The € Euro symbol had not yet been invented. So, on my system, I have several thousand fonts which don’t have that glyph. In QuarkXPress, it’s fairly easy to overcome this by using a Conditional Style to swap all the instances, especially as the € is not really part of any typeface and so won’t particularly jar. However, what if a font doesn’t contain accents? What about the full set of punctuation, maths symbols or international letters that you need. Many older fonts, from the early days of TrueType, have limited letter sets. It’s really only since OpenType and Unicode that we are used to finding everything we want.

For English, the easiest way to check, before even downloading a font, is to paste this text in to the preview:

ABCDEFGHIJKLMNOPQRSTUVWXYZ abcdefghijklmnopqrstuvwxyz àèéïöùüçûæÆœŒ .,;:?!/[]{}()*-– —…“”‘’_ 0123456789 ≤≥÷+=≈≠±-·√°@€£$%&*|«»\<>/~”‘§¶©®™

If you’re working in another language, you’ll need to adjust the letters to include specifics that you need. If a font is short on glyphs, there may be a more recent version that does have a full set, or a ‘Pro’ version which is more complete. However, it’s unlikely that this is going to be a free font, though it may well be money well spent.

While you’re checking that, also ask yourself if you need small caps and old style figures. Small caps as created by QuarkXPress (and competing software) are created on the fly by reducing the point-size. That never looks as good as true small caps, where the letter forms have the same weight, but are shorter. If you are going to need small caps, make sure they are in the font—even if that does mean having to pay for the ‘Pro’ version, or a small caps compliant version. The same is true for old style figures. They look better in text than tabular figures. Some fonts have a ‘maths’ set which includes fractions. Again, although QuarkXPress does a great job in creating fractions, these are never as good as true fraction glyphs. The same goes for ordinals and others. If your work includes a lot of mathematical notation, or a lot of figures in text, then it may be worth building the design around a set of fonts that come complete with everything you need.

Cambria math includes a complete set of fractions
Cambria math includes a complete set of fractions


A little known fact about OpenType fonts is that they have to be ‘signed’ to work on Windows with the full set of OpenType features. The ordinary features will work, and all the features including discretionary ligatures, stylistic sets and so on will work on a Mac, but to work on Windows, a bit of code has to be inserted. The process for doing this is quite involved, and, since most font designers are working on Macs, it may not have occurred to them. Therefore, if you are going to use a font which will also be used on a Windows PC, check that the OpenType features actually work, if you’re going to rely on them. Of course, this only really applies to ‘free’ fonts, though, for commercial fonts, it’s still worth checking. Not everything that is for sale is worth the price…

Equally, it’s worth asking the question: are there webfonts available? If the font has been released as Creative Commons, Public Domain or SIL (see the next section) you can easily convert to webfonts using any one of many websites or utilities. If it’s a copyright font, though, you most likely do not have the right to change it. On the web, it isn’t difficult for the copyright holder to track you down and sue you.


Is the typeface you are planning to set legal? ‘Of course it is’, you say, ‘they wouldn’t be allowed to have it on a website if it wasn’t’. Well, yes. Copyright on fonts is tricky. In the USA, the fonts themselves can’t be copyrighted. In the UK they can, with an industrial copyright lasting 25 years, though this may (or may not) be extended to 70 years. Fonts can be trademarked, and almost all commercial fonts are. There was a famous court case between two high-end foundries which was won by the plaintiff. It turns out that keeping the same first three letters of the name and changing the rest does not get round trademarking. Twenty years ago, especially before Corel licensed the ‘official’ fonts which it shipped with Corel Draw, the world was awash with ‘Aardvark’ (for ‘Aachen’) and others. There have been numerous cases of manufacturers of bumper CDs of fonts being successfully sued.

Where does that leave you? Well—and this isn’t legal advice, so get advice if you intend to sail close to the wind—every font on a website should download with a license file. Check the license before using the font! Also check the embedded copyright. Any font utility can do this. If the two aren’t the same, ditch the font. If the license is ‘private and non-commercial use only’, ditch the font. Not only is it inconvenient to have to separate your non-commercial from your commercial work, but many designers who have scanned in and digitised fonts without any consideration for copyright release them as ‘private and non-commercial’ in the (mistaken) belief that this gets them round copyright restrictions. It doesn’t. They may be on the right side of copyright, but they may not be. Equally, fonts with odd names that you find on an old floppy disk in the office may themselves have been subject to lawsuits. The onus is on you to check, so check the license file, and discard anything which looks dodgy.

Also check the license for terms of use. Are you allowed to modify the font and distribute it? If so, this can be very good news if you need to issue it to hundreds of dealers for their catalogues, especially if you want to include a couple of otherwise hard-to-find symbols, assuming you have the tools to make the changes. On the other hand, if the license is restrictive, you have to stick to that.

Hidden Costs

There is such a thing as a free lunch (just ask the Salvation Army), but most free lunches come with strings attached. Why is this font available for free, or at low cost? If it’s a promotional weight, what will the cost be of buying the other weights you need? Without Bold, Italic and Bold Italic, many typesetting jobs will be a nightmare. You can actually pair an italic from a different font if you need to (though not an oblique), but, then, why not just use that other font? If it’s a subscription font, how long do you want to pay the subscription for? Typekit and Skyfonts may be great for previewing fonts, but over ten years, it may just be cheaper to buy the font outright. Also, is it really the right font for the job? It’s terribly easy to convince yourself that this amazing font you’ve found with a full set of weights, perfect kerning and beautifully legible letters is the one, but is this because it really is, or because it’s free? An inappropriate font will work against the rest of the design, up to the point that you either replace it with a paid-for font, or allow it to be printed with subsequent damage to you and your client’s reputation.

Using the Font Evaluation file

For QuarkXPress 2015 users, here is the Font Evaluator file: Font Evaluator.qxp. For everyone else (and those just curious) here is the Font Evaluator PDF file with Futura Medium examined.

Font Evaluator.qxp Instructions: Change the “Normal” Character Style sheet to the font you wish to preview. All other style sheets will follow this. Change the name of the font below to the name of your font. All the headers will follow this. [This relies on the Content Variables function in QuarkXPress 2015.]

The way this works is that you change the Normal Character Style to the font you are interested in, and change the name of the font on the front page, making sure that it is still tagged with the Paragraph Style ‘Font’. From this, QuarkXPress will change all the headers and footers to suit.

Page 2 gives you the character set, for English, at 24, 18, 11 and 6 point. David Ogilvy’s investigations, described in Ogilvy on Advertising (really, everyone should read this), indicate that 11 point is the most legible size for most fonts. The point of the character set is to check that everything you need is there.

Pages 3 to 5 give you the kerning—once straight up, once blurred, and once upside down. If you see a problem with any of the versions, discard the font! The problem will show up in extensive body text. These pages should also give you an impression of the overall evenness and colour of the font. The frequent sets of capitals make this a challenging test for most fonts. Futura, as you see, comes out pretty well.

Page 6 gives you some columns at different sizes. If the text looks cramped between the lines, chances are your typeface will require extra leading. Not a problem, but it does make it less efficient. The same goes for if ‘rivers’ of space are opening up between the words. Resist the temptation to reduce the inter-word spacing. Studies show that increasing inter-word and decreasing inter-letter actually improves legibility, whereas decreasing inter-word and increasing inter-letter makes it worse. Increase the leading instead.

Page 7 gives you the sets of easily confused letters. Futura does great on M versus W, and the very open c contrasts nicely with the more closed e. a and o can be a problem, because Futura uses the closed a. For capitals we are doing fine, except for I (capital I) and l (lower case l). For the word ‘Ill’ Futura is a disaster: it’s just three lines. In fact, the whole group ijltf is problematic in Futura.

Pages 8-9 give you easily confused letters as blurred text—the file uses QuarkXPress’s shadow feature to achieve this. As the eye scans a line, it picks up information from the words in peripheral vision to help interpretation. You can assume that any body text font will be legible when you concentrate on individual words, but checking whether you can distinguish the blurred words can show up problems which will slow down reading and make it tiring.

Page 10 gives you a box for measuring the x-height versus the t-height. There are instructions (as notes) in the Quark file, though you can’t see them in the PDF. Each small square is 5% of the box, so if you scale the text until it occupies the entire box, you can get the proportions. Classical typography suggests proportions for descender:x-height:ascender of 4:6:4. Modern large-x-height fonts such as Helvetica are in the proportion 4:9:4. Futura has proportions of 4:9:7, which is why the descenders always look stubby compared to the elongated ascenders. Unfortunately, this means that Futura is optically smaller than its font size suggests. Despite being (generally) very legible, notwithstanding the I and l problem, in body text you will have to set it larger to get the same legibility as Helvetica.

Finally, page 11 gives you a checklist of what to look for.