Design System Patterns

Design System Patterns

Overview

A design system is not just a component library.

It is the shared contract between:

• design decisions
• reusable UI primitives
• accessibility expectations
• documentation
• contribution rules
• release and adoption workflows

Good design systems make teams faster and more consistent.

Bad design systems become a second product that nobody fully trusts.

What a Design System Usually Contains

If one of these is missing, teams usually compensate with ad hoc workarounds.

Goals of a Design System

A healthy design system should help teams:

• build consistent UI faster
• reduce duplicate component logic
• encode accessibility as a default
• make design decisions reusable rather than one-off
• support change without rewriting every app screen

It should not try to eliminate all flexibility.

Tokens First

Design tokens are the smallest reusable design decisions.

Common token categories:

• color
• spacing
• typography
• radius
• shadow
• motion
• z-index

Raw Tokens vs Semantic Tokens

Raw tokens:

--gray-100: #f5f5f5;
--blue-600: #2563eb;
--space-4: 16px;

Semantic tokens:

--color-surface: var(--gray-100);
--color-action-primary: var(--blue-600);
--space-control-padding: var(--space-4);

Raw tokens define a scale. Semantic tokens define meaning.

Semantic tokens are what make themes and redesigns easier, because components depend on meaning rather than a hardcoded visual choice.

Token Rules of Thumb

• Prefer semantic names where components consume tokens
• Keep scales predictable
• Avoid one-off tokens for single components unless they are proven patterns
• Use CSS variables or an equivalent runtime-friendly token format when theming matters
• Keep token naming boring and stable

If token names encode current visual taste instead of intent, the system becomes fragile.

Component Layers

Most systems work better with clear layers instead of one giant component pile.

1. Primitives

Low-level building blocks with minimal opinion:

• Box
• Text
• Stack
• Inline
• Button

These should be composable and predictable.

2. Patterns / Composites

Composed building blocks for recurring UI:

• FormField
• Card
• Dialog
• Tabs
• Toast

These encode more behavior and structure.

3. Product-Specific Components

App- or feature-level components that belong outside the core design system:

• InvoiceStatusCard
• SubscriptionUpgradeBanner
• TeamMemberRow

Not every reusable component belongs in the shared system.

API Design for Components

Component APIs should be small, composable, and difficult to misuse.

Good API Characteristics

• predictable prop names
• clear variant model
• accessible behavior by default
• escape hatches without encouraging chaos

Prefer Composition Over Prop Explosion

Hard-to-maintain API:

<Button
  primary
  danger
  rounded
  fullWidth
  iconLeft="plus"
  iconOnly
  dense
  loading
/>

Better:

<Button variant="danger" size="sm" loading>
  <IconPlus />
  Create
</Button>

If a component needs dozens of boolean props, the abstraction is probably wrong.

Variants, Sizes, and States

A design system usually needs a clear model for:

• variants
• sizes
• states
• slots

Variant Model

Good examples:

• variant="primary" | "secondary" | "ghost"
• size="sm" | "md" | "lg"
• tone="success" | "warning" | "danger"

Avoid variants that encode page-specific meaning like dashboardPrimary or settingsAlt.

State Model

Components should define expected behavior for:

• hover
• focus
• pressed
• disabled
• loading
• invalid

These states should not be reinvented in every screen.

Slots and Composition

Complex components often need named regions instead of giant prop bags.

Example:

<Dialog>
  <Dialog.Header />
  <Dialog.Body />
  <Dialog.Footer />
</Dialog>

This is often more maintainable than a component with many “content props”.

Slots are especially helpful when layout structure should remain consistent but content should vary.

Accessibility as a System Default

Accessibility should be built into the system, not left to each feature team.

System-level defaults should cover:

• keyboard interaction
• focus styling
• semantic markup
• color contrast
• reduced-motion handling
• error and busy states

If the design system makes the accessible path harder than the inaccessible path, teams will drift.

Styling Strategy

A design system can work with different styling approaches:

• CSS variables
• utility classes
• CSS Modules
• CSS-in-JS
• static extracted styles

The important part is not the styling brand. It is whether the approach supports:

• tokens
• theming
• predictable override behavior
• component portability
• runtime and build performance needs

The system should not depend on “magic” that only one or two maintainers understand.

Theming

Theming works best when it is token-driven.

Common theme axes:

• light and dark mode
• brand or tenant customization
• density
• motion preference

Theme Pattern

:root {
  --color-surface: white;
  --color-text: #111;
}

[data-theme="dark"] {
  --color-surface: #111;
  --color-text: white;
}

Components should consume semantic tokens like --color-surface rather than branching manually in each component.

Documentation Patterns

A design system is only useful if teams can discover how to use it.

Good docs usually include:

• when to use the component
• when not to use it
• examples
• variant guidance
• accessibility notes
• content guidance when relevant

Useful Documentation Types

• component API docs
• design rationale
• do/don’t examples
• migration notes
• layout and composition recipes

Docs should answer real usage questions, not just dump prop tables.

Contribution and Governance

A design system needs rules for what gets added.

Good governance questions:

• Is this pattern reused across products?
• Can an existing component solve it with composition?
• Is the proposed API stable enough to support long-term?
• Does it improve or harm consistency?
• Who owns maintenance after merge?

Without governance, systems grow into inconsistent wrappers around product-specific needs.

Versioning and Change Management

Shared UI systems need stable change practices.

Useful patterns:

• semantic versioning
• deprecation before removal
• migration guides
• codemods for common API changes
• changelogs focused on adoption impact

Breaking changes in a design system ripple into many applications quickly.

Adoption Patterns

A design system succeeds only if product teams actually use it.

Common adoption strategies:

• start with a few strong primitives
• replace high-duplication components first
• provide examples for common flows
• make migration easier than staying custom

If the system feels slower or harder than copying local code, teams will bypass it.

Testing a Design System

Design systems need more than unit tests.

Useful test layers:

• interaction tests for behavior
• accessibility checks for semantics and keyboard flows
• visual regression for stable appearance
• token/theme regression coverage

The goal is confidence that the shared foundation stays reliable across many consumers.

Common Anti-Patterns

Decision Guide

Use this rough mapping:

• Need reusable visual values: tokens
• Need layout and text primitives: foundational components
• Need recurring interaction pattern: composed component
• Need feature-specific business UI: keep it in the product, not the core system
• Need multiple brands or modes: semantic tokens plus theming
• Need consistency across teams: docs plus governance, not just code

Rules of Thumb

• Start with tokens and primitives, not a giant component backlog
• Make the default path accessible and visually correct
• Prefer composition over API sprawl
• Keep product-specific abstractions out of the core system
• Treat documentation and governance as part of the system itself
• Optimize for adoption, not only architecture purity

Related Skills

• accessibility-interaction-patterns for system-level accessible defaults
• developer-experience for tooling and workflow ergonomics
• performance-patterns for styling/runtime tradeoffs and shipped UI cost
• testing-strategies-patterns for regression and confidence patterns