Marcus Andrey Study
Ler em português Library

Design Systems · Step 1 · 6 min read

What a design system solves

A design system solves one problem that grows with every team: the same decision made again and again, slightly differently. What that costs, what a system gives back, what it doesn't solve, and the fictional ferry company whose system, Harbour, runs through this course.

Twelve buttons scattered on a board, each a little wider, rounder or more turned than the last, the near-duplicates an interface inventory turns up. Pass the pointer over them: the ones near it snap to one size, one radius and one grid, the nearest first, while the far ones stay as they were. That's this step: a design system starts by seeing how many versions of one thing a product has grown, and making them one. This step has four lessons: what a design system solves, when you don't need one yet, the anatomy of a design system, and the interface inventory.Illustration: Marcus Andrey, drawn with Hairline (MIT, Lucas Marques).

A design system exists because good people, working apart, make the same decision in slightly different ways. Nobody sets out to build seven primary buttons. One designer picks a blue on Monday, a developer copies it from a screenshot on Thursday, another team starts from a different file a year later. Each choice was fine. Together they make a product that feels like several, costs more to change and fails in places nobody checked. A design system is how a team makes that decision once and shares it.

What you'll be able to do after reading:

  • Name the problem a design system solves, in one sentence
  • List what drift costs: time, quality, accessibility, trust
  • Tell what a system gives back and what it can't fix
  • Introduce Harbour, the system we'll build in this course

1. Seven buttons that do the same job

Before: seven "primary" buttons, two products

  • Book sailingApp, home · 6.4:1
  • Book sailingApp, results · 5.0:1
  • ContinueApp, checkout · 3.1:1 fails
  • Book nowApp, offers · 6.0:1
  • Check inConsole, list · 8.4:1
  • Check inConsole, detail · 4.2:1 fails
  • Save sailingConsole, admin · 5.0:1

Five blues and a teal, seven corner radii, three text sizes. Each one was a reasonable choice on the day. Two fail the 4.5:1 text contrast WCAG asks for.

After: one Harbour button, one decision

Book sailing button.primary.background → color.action.primary → color.blue.600

One colour, one radius, one height, measured once: white text on it is 6.4:1. When the brand blue changes, it changes in one place, in both products.

The problem a design system solves, in one picture: the same decision made seven times, slightly differently, by people who never met. Contrast ratios measured for each label on its own fill (WCAG 2.2, 1.4.3).Tidewater and Harbour are a fictional company and system. Demo: Marcus Andrey, CC BY 4.0.

This course follows one company the whole way through. Tidewater is a fictional ferry company with two products: a booking app, where people buy tickets on their phones, and a staff console, where port staff check passengers in and manage sailings on desktop. Two teams built them, three years apart.

Look at the left panel. Every button does the same job: "this is the main action on this screen". There are seven of them, and two fail the 4.5 to 1 contrast WCAG asks for normal text (W3C, Understanding 1.4.3). Nobody decided that. It happened, one reasonable choice at a time.

This is the problem. It's not ugliness, and it's not lazy people. It's drift: the same decision made again by people who don't share a source.

If you took UX Foundations, you met the four parts of a system in What a design system is. This course goes into each one, in practice, with Tidewater.


2. What drift costs

  • Time. Every screen starts with choices someone already made. Reviews spend time on paddings instead of on the problem.
  • Quality. Each copy of a part has its own bugs. Fix the focus ring on one button and the other six still have the old one.
  • Accessibility. Contrast, focus, labels and touch size are checked once per copy, or not at all. The two failing buttons in the figure are what "not at all" looks like.
  • Change. A rebrand or a dark mode means finding every copy by hand. With fourteen blues, nobody is sure they found them all.
  • Trust. People notice when a product doesn't hang together, even if they can't say why. Staff who use the console all day notice most.

NN/g puts the benefits of a system the other way round: work "can be created and replicated quickly and at scale", teams get "a unified language", and products get "visual consistency across products, channels, and (potentially siloed) departments" (NN/g, Fessenden, Design Systems 101, 2021).


3. What a system gives back

One decision, many uses. The right panel of the figure is one button whose colour comes from a chain of named decisions: button.primary.background points to color.action.primary, which points to color.blue.600. Change the end of the chain and both products follow. That chain is a set of tokens, and Step 2 is all about them.

A shared language. When the designer says "primary button" and the developer types Button variant="primary", they mean the same thing. That saves more meetings than any component.

Checked once, right everywhere. Contrast, keyboard focus, screen reader names: done once in the component, inherited by every screen that uses it.

Time for the hard problems. The console's check-in flow is hard. The colour of its button shouldn't be.


4. What it doesn't solve

A design system is often sold as the answer to everything. It isn't.

  • It doesn't make design decisions for you. It makes the small, repeated ones. What the check-in flow should be is still a design problem.
  • It doesn't fix a bad process. If teams never talk, a library won't make them. It helps, because it gives them something to talk about.
  • It isn't free. NN/g is clear that building and keeping one "is a time-intensive activity which requires a dedicated team", and that systems "are not a one-and-done solution". The next lesson is about when that cost is worth it.
  • It doesn't make the product good. Seven matching buttons on a confusing screen are still a confusing screen.

5. Meet Harbour

Tidewater has decided to build a system. They call it Harbour. Over this course, Harbour will get:

  • Tokens for colour, type, space and radius, with a light mode, a dark mode and a compact density for the console (Step 2)
  • Components with clear properties and states (Step 3), and patterns like "show a delay" (Step 4)
  • A Figma library, code, and the bridge between them (Step 5)
  • Docs people read and a way to measure adoption (Step 6)
  • Governance, and later a second brand, the budget line Skiff (Step 7)

Every token name you see in this course is Harbour's. We'll also compare Harbour with real public systems, like Material Design, IBM Carbon and GOV.UK, so you see how others solved the same thing.


6. Common mistakes

  • Treating drift as a style problem ("make it prettier") instead of a decision problem ("who decides, once?").
  • Starting a system by drawing new components, before looking at what already exists.
  • Promising that the system will make every screen good.
  • Counting only the design time saved, and forgetting development, testing and accessibility.

7. Vocabulary

Term Meaning
Design system Shared decisions (tokens, components, patterns, guidelines) and the people and process that keep them alive
Drift The slow spread of near-copies of the same decision
Token A named design decision, like color.action.primary
Component A reusable piece of interface, designed and built once
Source of truth The one place a decision lives; everything else points to it

8. Try it (no tools needed)

  1. Open two screens of a product you work on, or one you use every day.
  2. Find every button that means "the main action here". Write down how they differ: colour, radius, height, text.
  3. For each difference, ask: was it a decision, or did it just happen?

Reflection questions: If the brand colour changed tomorrow, how many places would someone have to edit? Who would know when they'd found them all?


9. If you already practise

  • Describe your system's value in costs you can see: review time, bugs fixed twice, accessibility issues found late. "Consistency" alone doesn't fund a team.
  • Keep a short list of what the system will not do. It protects the team from becoming the place every problem gets sent.
  • Watch for new drift inside the system itself: two components that do the same job are drift with a version number.

Checklist

Learn more

Books, courses and videos to go deeper

Book links go to Amazon. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.

What a design system is · When you don't need one yet · Anatomy of a design system · The interface inventory · Colour and contrast

Saved only on this device.

↑ ↓ to move · Enter to open · Esc to close