UX essay · System design
A design system is only living if it is used as code
I have built a design system at every company I have worked for since 2017. One difference decided which of them were still in use after I left: whether the library was compiled into the product, or filed as a document to consult. This page walks through the process with Findhotel as the example, then shows what the other systems shipped.

Problem
Nobody decides to build four different buttons
In a company of any size, designers sit with product teams, for good reasons: each learns that team's problem and ships on its roadmap. The cost shows up later. Put two teams' screens side by side and the differences are obvious.
It happens at every scale: a font style one step off, a grey sampled from a screenshot, or a whole flow built because one team could not know another had already solved it.
This is not a quality problem. Every screen was made carefully by someone competent. The inconsistency comes from how the teams are organised, so the fix has to change the organisation too. A document asking designers to be consistent is only a request.
So I do not build a style guide. I build a place where designers and the front-end developers who care about user experience agree on a shared language, and a route that turns what they agree into part of the product.
Goals
What a system has to do to be worth the sprint
Deep diveWho uses the library
The beneficiary is the customer. The users are four groups in a company.
Team & audience
One sprint in four, and no product owner
What has worked best for me is a borrowed team, not a standing one: a designer and a front-end developer from each product team, meeting on a fixed schedule.
For three sprints everyone works on their own product and runs into what the library is missing. The fourth sprint starts with a queue of real needs, not a wish list made up in a meeting room.
At Findhotel the team was six people: four front-end developers, one from each product team, and two designers. We ran a kanban board and had no product owner.
Without one, the ticket writing fell to me. Most of it was small and specific. Occasionally it was not: bringing every colour and type token up to W3C contrast guidance for example.
Scope and constraints
What did not need an experiment
Almost none of this work was tested, and that was the correct call. Two key decisions made up nearly all of it, and neither is the kind of question an experiment can answer.
The first is removing a variation nobody chose. If a product has three greys where two teams each invented one and nobody can say what the third is for, collapsing them is not a hypothesis. There is no version of the result that argues for keeping an accident.
The second is applying guidance that already exists and is better evidenced than anything we could run ourselves. Contrast ratios are the clearest case: W3C has done the work, at a sample size no single company reaches, and rerunning it locally to see whether our users also prefer being able to read things would be theatre.
When a change genuinely was a question, it left this team. It went down to the product team that owned that flow, where there was a product owner to prioritise it and a data scientist to design the test.
Process
Six steps, and the first three are politics
The order matters. Three of these six happen before a single component is made, and skipping them is how a design system becomes a folder somebody made once.
Mapping what is already there
An inventory before a proposal. Colours, text fields, font families, toggles, sliders, popups, buttons, split buttons, galleries, and every variation of each that the product turned out to contain.
The deliverable is the list, drawn out. Someone can dispute a count of thirteen button styles. Nobody disputes thirteen buttons laid out across three slides.
At Findhotel the primary button appears three times, in three different blues: #3979F5 on the buttons slide, #0074F8 in the header, #4476FB in the dialogues. Nobody picked three on purpose.
Buttons
13- #3979F5Buttons slide, Search
- #0074F8Header section, Search
- #4476FBDialog, Back to form
One primary button, three blues.
Deep diveThe rest of the catalogue
Form
13Display
9Containers
8Images

Menus
Navigation
Overlays
Buying the time
With the inventory in hand, a lead developer and I presented to the people who decide where sprints go. We invited the people who had been complaining that the product did not look like itself. They had already seen the problem; they just had no name for it.
Nobody funds a design system as an idea. They fund it once they have seen the thirteen buttons.
The deck asks what the problems are and answers with the plain answers: features ship slowly, nobody is sure which components already exist or how to use them, small bugs in components never get fixed, accessibility is missing across the application. Read on its own it is the case that could be made at any company.
Composing the team from every team
A designer and a front-end developer from each product team.
With someone from every team, the library has a voice in every room where it will be used. By launch day, adoption has already happened. That is the difference between a library the company uses and a library only its authors use.
Four product teams, and the six people who put their strengths together: four front-end developers and two UX designers.
Choosing tools
On the design side this came down to two decisions: what to draw in, and where to publish. At Maersk in 2019, InVision was already licensed and widely used. At Findhotel it was Zeplin. Both companies used Sketch. In each case the answer was the tool already in use.
On the development side there was a new requirement rather than an inherited one: the main consideration was to make the process of publishing and debugging a component as automatic as possible.
Making components, and agreeing what one is
Divided among the designers, one component at a time, and the unit of work is the component in all of its states.
Two designers splitting the work will make two halves that do not fit, unless someone writes down what "finished" means. So I wrote nine acceptance criteria every component had to pass.
- Flexible to resizing with autolayout
- Using the official list of named font styles
- Using the official list of named colours
- Fitting exactly within the grid: one unit is 8px, and it can be halved to 4px when needed
- Carrying a mobile and a desktop version, where that applies
- Exposing the fewest editing permissions that still make it useful
- Properly named, including every sublayer
- Tested on a canvas for resizing and for permissions
- Published to the official platform
Three of them are worth seeing rather than reading, because each is a rule about a shape.
Shape rules
Publishing, which is where the argument is settled
My earlier attempts got this step wrong, and at the time it did not look like failure. I published the library to designers. Designers used it, and the products drifted anyway: developers who rebuild a design from a picture produce the same drift, one handover later.
Design only ships once it is code. So this time I published for the developers. I mapped each component between design and code, each with its own URL in an index file the front end could read.
At Coolblue in 2017 and at Maersk in 2019 what I built was aimed at designers, with limited developer involvement, and both were left at the mercy of whoever was sitting in a given team: it was that designer's job to reach for the components, to keep them current, and to apply them properly. That is too much to hang on one person's diligence, and it is not a criticism of any of the designers involved. A system whose enforcement mechanism is somebody remembering is a system that degrades on the first busy sprint.
Below is one component, the star rating, in all the three places it exists. It is the same object with variations.
What shipped
Pages that came out of the design system
These pages are assembled entirely out of parts made in their corresponding design systems.
Findhotel, 2020
The whole funnel, on desktop and mobile. The mobile captures are full page exports.
Deep diveFloorplanner, 2018
Floorplanner, 2018
A drawing tool rather than a funnel. The two marketing pages at the end are the interesting ones, because a marketing page assembled from the product's own components is an example case where the guide reached beyond the core product.
Deep diveMaersk, 2019
Maersk, 2019
Twill, Maersk's platform for the small and medium exporters and importers who book LCL (less than container load) cargo: shipments too small to fill a container, sharing the space and the cost with other people's freight. The screens below follow one booking from the destination search to the confirmation screen.
Deep diveZimmer Biomet, 2021
Zimmer Biomet, 2021
A surgical planner, built over eighteen months. Most of the dialogues in this product are one component with different content inside it. Change the component once and every dialogue changes with it.
Deep diveTrackBee, 2025
Five flows shown at three widths each: signing in, creating an account, connecting ad platforms, choosing a plan, and coming back after a lapsed subscription.
Conclusion
A living design system is a system that is coded
By the time I left it was a live library, imported by every product in the portfolio rather than referred to by them.
Bringing every component up to W3C accessibility guidance improved every product at once.
Lessons
What changed since
Since then Figma has become the default design tool, and Storybook has kept its place. The bigger change is underneath the tools: a designer working with agentic AI like Claude Code can write and commit a component directly, instead of waiting for a developer to build it.

































