Brand Identity System: Design Tokens Across Every Platform
How token tiers, role-based type, perceptual colour and shared component names keep one brand identity system consistent across web, iOS, Android and product UI.
Give your brand book to an iOS engineer and ask what colour the primary button should be in dark mode. The book will not say.
Brands start to drift apart at exactly that point. The marketing site stops matching the app, and the app stops matching the logged-in product. Before long, three teams are shipping three slightly different blues. A brand identity system built for digital fixes this. It turns visual decisions into named, versioned data that every platform reads from the same place. Here is how the pieces fit together.
Get a token architecture that keeps your brand identical across web, app and product interfaces.
Why the brand book stops at the edge of the screen
Brand guidelines describe end results: what the logo looks like, which colours are approved and which typeface sets the headlines. Interfaces need rules for situations the brand book never covered. Think of a disabled toggle, a focus ring on a dark card, or an error message inside a modal when the user has text set to the largest accessibility size.
When a team hits one of these gaps, it makes a local decision. Each decision makes sense on its own. Added together, they produce drift.
A longer PDF will not fix this. The source of truth has to move out of documents and into a format that design tools and codebases read directly. That format is design tokens. Brand strategy still decides what the interface should express, and our piece on brand strategy and design decisions covers that part. Tokens are how those decisions get applied the same way everywhere.
Token architecture: three tiers, references in one direction
Most mature systems split tokens into three tiers. The names vary, but the jobs do not.
- Primitive tokens are the raw palette and measurements: color.blue.600, space.4, font.family.brand. They say what exists. They do not say where to use it.
- Semantic tokens give those values a purpose: color.text.default, color.action.primary.background, color.feedback.danger. This is where most brand decisions are recorded.
- Component tokens attach semantic decisions to a specific part, such as button.primary.background. Use them only where a component really needs to differ from the shared default.
The rule that makes this work is that references only point downward. Components point to semantic tokens, and semantic tokens point to primitives. A component never points straight at a primitive. If you follow the rule, dark mode, a high-contrast variant or a full rebrand becomes a change to the semantic mapping, and no component file needs editing.
The semantic tier is where the brand actually lives. Primitives are the paint box. Semantic tokens decide which paint goes where, and those decisions are what a rebrand changes.
There is now a shared standard for this. The Design Tokens Community Group announced the first stable version of its specification, numbered 2025.10, on 28 October 2025. It supports aliases, component-level references, themes and multiple brands, and one token file can generate platform code for iOS, Android, web and Flutter. The announcement lists Style Dictionary, Tokens Studio and Terrazzo as reference implementations, and Figma and Penpot among the tools adopting the format.
Google’s Material Design 3 works on the same principle. It describes design tokens as the building blocks of all UI elements, with the same tokens used in designs, tools and code. If a supplier proposes a token structure that cannot be compared to either model, ask them why.
One warning: do not tokenise everything. A system with thousands of component tokens is impossible to audit, and nobody will remember which ones matter. Add a component token only when the component genuinely needs to differ from the semantic default.
Typography that survives three rendering engines
Each platform handles text size differently. On the web, sizes are set in rem units, often made fluid with CSS clamp(). On iOS, Dynamic Type text styles resize when the user changes their system text size. On Android, sp units follow the user’s font scale setting. A type scale drawn in fixed pixels in Figma will not work properly on any of the three.
The solution is to name text by its role, not its size. Material uses display, headline, title, body and label, and that set covers most products. Each role is a composite token that bundles family, weight, size, line height and letter spacing. The DTCG specification includes a typography composite type for this. Each platform then maps the role to its own scaling system. On iOS, for example, text.body follows Dynamic Type, but its name and intent stay the same as on the web.
Modular scales built on a fixed ratio look tidy at headline sizes and tend to break down at the small end. Expect to adjust body and label sizes by hand. Line height should also get tighter as size grows. A ratio that suits paragraphs makes a large headline look loose.
Two practical points often surface late:
- Licensing. A web font licence often does not cover embedding the font in a native app. Check this before the typeface is approved, not after the app build has started.
- Performance. Brand fonts are render-blocking assets. Subsetting, variable fonts and a sensible font-display strategy decide whether the typeface hurts load time. We cover the trade-offs in our article on fast Core Web Vitals.
Should the app use the brand typeface or the platform’s system font? It is reasonable to split them: the brand face for display and headline roles, and the system font for dense interface text. Keep the token names the same either way, so the decision can be reversed later in one change.
Colour as roles, not swatches
Start with how the ramps are generated. HSL is misleading: two colours with the same HSL lightness can look very different in brightness. Building ramps in a perceptual colour space such as OKLCH gives steps that look evenly spaced across every hue. The stable token specification supports OKLCH, Display P3 and the other CSS Color Module 4 colour spaces. That means wide-gamut brand colours can be defined once, with sRGB fallbacks generated for displays that need them.
Next, assign roles. A working set usually covers surfaces, text on those surfaces, actions, borders, focus states and feedback (success, warning, danger). Test every foreground and background pair against WCAG contrast requirements at the semantic level. Every component that uses those pairs then inherits the compliance, so you are not checking contrast one component at a time.
This often turns up an awkward fact: the signature brand colour fails contrast as text on white. That is normal. The brand colour becomes a background or accent role, and a darker step from the same ramp is used for text. The brand is still recognisable, and the interface stays readable.
Treat dark mode as a theme. It is not an inversion of the light palette. It is a second mapping of the same semantic names onto different primitives. Surfaces get lighter as they rise in elevation, and saturated brand colours usually need less chroma to avoid glaring.
What do you call a button?
Naming sounds like admin work, but it decides whether design and engineering are talking about the same thing. Use a fixed grammar and apply it everywhere. One common pattern is namespace, object, property, variant, state: acme.button.background.primary.hover.
Names should describe purpose, never appearance. “Blue button” stops making sense the day the brand moves to green. “Big heading” means nothing on a watch screen. Primary, secondary, danger and the role names above stay accurate through a rebrand.
The same words should appear on every surface. The Figma component, the React component, the SwiftUI view and the Jetpack Compose composable should share one name. Their variant properties should match too: if the code has variant, size and state props, the Figma component should have the same properties with the same values. Each platform can use its own casing (kebab-case in CSS, camelCase in Swift), but the words must not change. The namespace also lets one codebase serve several brands, which our article on a multi-brand design system covers in detail.
If a designer and an engineer cannot say a component’s name aloud and mean the same thing, the system is already drifting.
Building it, in order
Order matters more than tooling. Teams that open Figma first tend to end up with a well-organised version of the inconsistency they started with. This sequence follows the path Moburst’s own process takes, from Discovery through Brand and Strategy to UX and then UI. For the full component build, see our guide to building a design system.
Audit What Exists:
Catalogue every colour value, text style and component variant across the website, the apps and the product. The number of near-duplicates shows how far things have drifted.
Translate Brand Intent:
Decide which brand attributes the interface itself must express, and which belong only to campaigns.
Define Primitives:
Generate colour ramps in a perceptual colour space, set the spacing scale, and lock the typefaces after checking licences for both web and native use.
Map Semantic Roles and Themes:
Assign colour and type roles, test every contrast pair, and build light, dark and any brand themes as alternative mappings.
Agree the Naming Grammar:
Write the naming convention down, apply it in Figma and in code at the same time, and reject names that describe appearance.
Wire the Pipeline:
Store tokens in version control in the DTCG format and use a transformer to publish packages for CSS, Swift, Kotlin and any other target.
Set Governance:
Define who can propose a token, how deprecations are announced, and how a version bump reaches each platform team.
Questions that separate a competent supplier from a plausible one
A polished Figma file can make any supplier look capable. These questions show whether the system underneath it will hold together:
- Can you show me the semantic tier, and explain why each role exists?
- When a token value changes, how does the change reach the iOS app, and who has to approve it?
- If we rebrand, which files change? (The right answer is “the semantic mapping and the primitives, nothing else”.)
- Which token format and transformer do you use, and why?
- Where are contrast pairs tested, per component or per semantic role?
Clear answers to those questions matter more than a moodboard. You can also look at how this has held up on real launches in our case studies.
Before you commission anything, ask for one artefact: the semantic colour and type token map for a single screen, shown on web, iOS and Android. If the supplier cannot produce it, the brand will not stay consistent after launch.
FAQs
What is the difference between a brand identity system and a design system?
A brand identity system defines what the brand expresses: its colours, typography, voice and visual principles. A design system turns those decisions into tokens, components and code that product teams use. In digital work, the token layer connects the two.
How many tiers of design tokens does a brand need?
Most brands need three: primitives for raw values, semantic tokens for purpose, and component tokens only where a component genuinely diverges. The semantic tier carries most brand decisions and does most of the work during theming and rebrands.
Should the app use the same typeface as the website?
Not necessarily. Many brands use the brand typeface for display and headline roles and the platform system font for dense interface text. Keep the role-based token names identical so the choice can be reversed later. Check that the font licence covers native app embedding.
How do design tokens handle dark mode and multiple brands?
Both are handled as themes. The semantic token names stay the same, and each theme maps them to different primitive values. Components reference only semantic tokens, so they pick up dark mode or a different brand without any change to component code.
What file format should design tokens use?
Use the Design Tokens Community Group format, which reached its first stable version in October 2025. It is vendor-neutral and supported by tools such as Style Dictionary, Tokens Studio and Terrazzo, which generate platform code for web, iOS, Android and Flutter from one source file.
Make your brand hold together on every screen
A three-tier token architecture, role-based type and semantic colour are what keep a brand consistent as it moves from website to app to product. Moburst Digital Experience designs and engineers that system so a rebrand or dark mode becomes one controlled change, not dozens of redesigns.