Daniel Eden, Designer

A Design System Grammar

In my posts about Design Systems so far, I’ve theorized about the structural approach that causes existing design systems to fail, and an alternative solution that encourages the composition of properties instead of cataloguing complete components. However, a suite of style properties and simple, functional components gives an overwhelming array of options for composing them.

I’ve used the word “vocabulary” to describe a design system before—a collection of metaphorical nouns and verbs that allow designers, engineers and beyond to express their intent in building a product—but what good is an expansive vocabulary without rules to govern how words may be used together? “Colossal dredging qualm reverence” contains some relatively advanced vocabulary, but the phrase as a whole has no meaning. Put another way, what use is a pantry full of ingredients if you don’t know how to combine them in a way that makes an appetising meal?

Defining a Pattern Language

I’ve recently been thumbing through a book that had been on my wish list for some time: A Pattern Language, a comprehensive title cataloguing some 250 patterns for construction and city planning. Each pattern is defined with the following structure:

  1. The design problem
  2. Discussion of the problem
  3. Illustration of the problem
  4. The proposed solution

There are two additional pieces that make this book particularly relevant to design systems in the fashion I’ve prescribed to them in writing thus far:

  • Each pattern’s documentation is preceded with a list of other patterns that employ the upcoming pattern
  • Each pattern’s documentation is followed by a list of other patterns that are required for this pattern

What this leads to is a choose-your-own adventure book that allows you to select a particular pattern and get all the information you need about what work must be done to put the pattern to use, as well as what other features you unlock by employing the pattern.

For the purposes of demonstration, we can consider a similar approach for documenting UI components:


Composes: Login Page, Entity Creation Page, Search Form

At a minimum, a form is composed of a label, input, and button. They are required to capture user input. {Further problem definition and discussion…}

Composed of: Button, Input, Label

We can imagine coming to the definition of this Form pattern after deciding we need to have some way for the user to create some kind of data entity. If so inclined, we can dig further to find the definition of the Button pattern:


Composes: Form, Banner, Navigation

A Button is a low-level component that represents a familiar user concept. {Further problem definition and discussion…}

Composed of: Background, Border, Text, Spacing, Cursor

Aha! From our definition of the Button pattern, we start to see the primitive, atomic components surface. With further investigation, we might find the very bottom of our system hierarchy, with an illustrative example of all permutations of text size, background color, or spacing units.

One final detail I find particularly pleasing about A Pattern Language is the * asterisk in some pattern titles. Two (**) indicates almost complete certainty in the pattern’s validity to solve the given problem. One asterisk (*) indicates that the pattern is appropriate for most instances of the stated problem. Finally, the absence of an asterisk indicates to the reader that the problem is more difficult to solve with a single solution, and that the presented solution is open for interpretation or change.

What Bends Doesn’t Break

One of the most appealing things about defining a pattern language or design system in this way is the flexibility that it affords. For novice builders who just need a high-level component or concept of their project, they can find it at the top of the heirarchy and use the component as it comes. For those pioneering new experiences or experimenting beyond the confines of existing product surface, they can take those high-level components and concepts and deconstruct them into their composite parts; they can keep doing this until they have the primitive, atomic components they need to construct a bespoke pattern that still fits with harmony in the wider system.

What’s more, this approach is aligned with our expectation that product teams have deep knowledge of a particular problem set, and the system simply generalises that which is generalisable. In fact, as an exercise, you could write pattern definitions for existing design systems—like those of Salesforce, BuzzFeed, GitHub, and Google—and likely find that 99% of the patterns are shared. Nuances of visual appearance (and some very specific use cases) set aside, design systems are almost identical across platforms, products, and locales, and all but the atomic components can be accounted for without knowledge of this context.

This approach to building a grammar for our design vocabulary seems sound, and certainly makes a strong case for writing easy-to-understand documentation, but a question that remains for myself and for others building design systems is how this works in practice. How do these definitions of patterns manifest in design tools and engineering workflows? What are the technical limits and debts associated with composing and deconstructing patterns in this way? Certainly composition, even across platforms, is viable, as can be seen through AirBnB’s efforts to bridge design and engineering through concepts of UI primitives. But decomposition and re-assembly of patterns seems unrealistic, especially at the scale and complexity of platforms like Facebook or Google.

All I can say with confidence is that with the legacy of fields like architecture and traditional graphic design, the infancy of digital design systems presents us with a rich variety of avenues for experimentation—and that’s a very exciting invitation indeed.

This post is one in a loosely-connected series of posts about design systems and the teams that maintain them. Other installments include a conceptual view of how a design system ecosystem is maintained, a theoretical exploration of compositional design systems, and a collection of thoughts on how the relationship between designers and engineers should be approached.