
What every roadmap was blocked on
The parts of every form and listing the country teams were about to redesign.
- Checkbox
- Radio
- Toggle
- Input field
- Accordion
- Breadcrumb
- Segment
- Badge
- Tab
- Chip
- Button
- Alert
How CARS24 tested a new look and feel on real screens, picked its first components from teams' roadmaps, and got 44 composable parts adopted in three countries.

01 the starting point:
The 2024 rebrand changed how CARS24 looked. The components underneath it hadn't changed at all.

The rebrand moved the palette to OKLCH and carried a new blue through a four-layer token system. That is its own case study. But tokens only recolour what's already there. The buttons, chips, cards and headers were the same old components in new colours, and the app and mobile web still had two different homepages.
So the rebrand became the chance to rebuild the interface too. The team called the new system LEGO: small, solid parts that any team can put together.
Aniket led it as Design Systems Lead at CARS24, with a small group of system designers and product designers from each country.
The chapter before this one: one colour system for 10+ products in three countries.
02 who it was for:
India, the UAE and Australia each had their own business, product and design teams, and their own plans.

That was the hardest part of the job. A system made for one team would be ignored by the other two. The UAE also needed every screen in Arabic, reading right to left. And redesigns of the homepage, listing and product pages were already planned, so LEGO had to be ready before those teams started, or they would build their own again.
How might we give three teams with three roadmaps one set of components they'd actually choose?
03 look and feel, first:
Shadows, strokes, fills, radius, icons and contrast, each tried on real pages and reviewed with designers from every country.

Before building anything, the team decided how LEGO should look and feel for the next few years. Nothing was too small to test: should shadows pop or barely show, what colour should they be, how dark can a border get before everything looks like a form, which corner radii are allowed.
Every option went onto real pages, not empty artboards: busy listing pages, light marketing pages, banners and forms. Designers from the country teams reviewed them side by side, so the decisions belonged to them too.


04 typography:
Fonts were tested on the busiest pages first: listings and product pages full of prices, EMIs, kilometres and badges.
$lego-font-size-17017px · 130% · semiboldHyundai Creta SX, 2021
$lego-font-size-15015px · 150% · regular32,410 km · Petrol · 1st owner
$lego-font-size-13013px · 140% · regularInspected at 140+ points · 7-day return
How a token reads
$lego-font-size-150
The number is the size: 150 means 15px. Every size carries its own line height. Small text gets more space between lines, big text gets less.
Shown at twice the size. Prices are illustrative: every market's longest price had to fit the same slot, and Arabic needed the taller line height.
Candidate fonts were set into real pages from the apps: listings and product pages first, then light landing pages and marketing banners. The team checked how each font handled numbers and prices, how it sat next to Arabic, and how many lines a listing card took at each size.
The scale that came out of it is short, and every token says what it is. Designers and engineers read the same name and know the size without opening a spec.
05 on real screens:
Every decision was checked against the live homepage, so we could see what users would actually get. Drag to compare.


06 what to build first:
Three teams, redesigns already planned, and a small system team. What we built first would decide whether anyone used LEGO.
The obvious first step is the button, then type, colour and the other basics. But no product team needs another team to give them a button, and every team already had one. A system that starts there spends months on parts nobody is waiting for, while the teams redesign their pages with their own components.
We treated LEGO like a product with customers. Instead of asking what a complete design system should contain, we asked who we could help next, and with what. The product teams were in it from the start: their roadmaps set the order, their screens showed us the components, and their best versions became our starting point.
What each country planned to ship next quarter
Product and design leads showed us their screens and drafts
Every component we saw, by team: 26 on the first list
Most teams, on the roadmap, hardest to get right
Start from the best version a team already built
Test it in their screens, then install it together

| Component | Who needed it next | Why it went in batch 1 |
|---|---|---|
| Choice chips and filters | Listing, search and filters, in all three countries | Used on almost every page; four versions already existed to merge |
| Input field | Sell car, loans, login | Every form on the roadmap was blocked on it, and it's the hardest to get right |
| Alert | Loans, car check | New verticals needed clear states for errors, warnings and success |
| Tabs and segment control | Homepage, explore, listing | The homepage redesign was built around them |
| Badge | Every listing card | Three badge styles on one screen were the loudest problem in the audit |
| Button | Everywhere | In batch 1 because every form needs one, but never the starting point: every team already had their own |
Reasons summarised from the walkthroughs and the audit.

The parts of every form and listing the country teams were about to redesign.

Split across three system designers, each owning a set from skeleton to QA.

Navigation and overlays, plus components teams requested during the first two batches.
07 built to be composed:
Every team needs its own version of an input. Drawing each one as a variant doesn't scale. Slots do.
2 labels × 2 info icons × 2 leading icons × 2 trailing icons × 2 buttons × 2 helper texts × 7 states. Every one drawn, named and kept in sync, and still missing the case a team needs next.
One text input. Designers switch slots on and off and pick a state. Nothing gets detached, so a fix to the input reaches every screen.
_Base Input Label ↓
info · optional · required ↓
.iconplaceholder →
← .Text container
↑ _Base Input Helper Text
Five private atoms sit inside the input. Designers never place them directly; they switch the slots.




Search: before and after


The new search is made of LEGO parts: the header, the text input, choice chips for recent and trending searches, icon buttons and cards. Compositions like this became the pattern library.
08 the library:
Every sheet carries its states, sizes, usage notes and the date it last changed.














09 definition of done:
Every state is its own property, named the same in Figma and in code, and every component was QA'd before release.


The guideline set one rule: never merge states into one property like hoverErrorDisabled. Interaction, disabled, read-only and validation are separate props. That keeps the number of variants small, makes impossible combinations impossible, and gives engineers the same API they see in Figma.
Every component then went through the same six steps, and nothing shipped until its QA result was written down.
10 adoption:
A library nobody uses is a museum. Getting LEGO into three products took as much work as building it.
Designers had to change habits, not just files. LEGO's props worked differently from the old kit: states became separate properties, and variations came from slots instead of detaching a component and drawing over it. If teams didn't understand that, they would detach LEGO too, and we'd be back where we started.
So the rollout was planned as carefully as the components. Each batch shipped with a session that showed what was new. Office hours ran every week. And the system team sat with each country team while they rebuilt their first screens, fixing what broke straight away.
Type, surfaces, shadows, radius and icons tested on real pages, with designers from all three countries reviewing.
ResearchThe 12 components every roadmap was blocked on, each shown to the teams in a walkthrough.
12 componentsSplit across three system designers; each component QA'd and the result recorded before release.
44 componentsHow the new props work, how slots compose, and how to make your own version without detaching.
All three teamsBring a screen, leave with an answer. Questions that kept coming up became usage notes on the sheets.
OngoingThe homepages in India, the UAE and Australia, then buy, sell, listing and product pages.
ShippedDesigners rebuilt one of their own screens from LEGO slots and atoms, in the session.
The fastest way to find the gaps. Each repeated question became a note on the component's sheet.
Right-to-left checks on every component, and help adapting their screens to the new system.
11 shipped:
Sell, car check, loans and explore each got their own homepage. Header, search, tabs, cards and list rows are all LEGO.




Pages rebuilt on LEGO, top to bottom





Each phone scrolls through the full page. Select one to see it whole.
12 what changed:
The homepages for India, the UAE and Australia moved to LEGO.
Buy, sell, listing cards and product pages, redesigned on the new system.
Built from five shared atoms, with slots instead of endless variants.
Built in the order the teams' own roadmaps needed it.
Don't start with the button. Start with the component the next roadmap is blocked on.
Test the look and feel on real pages before drawing a single component.
Composable slots beat a variant for every case, and let teams make their own versions.
13 for your team:
Each step of this case study maps to how we work with clients now.
$1,500
In this case studyRoadmap and product walkthroughs, an audit of your live screens, and a list of what to build first, with the reasons.
See what's included$5,900
In this case studyLook-and-feel tests on your real pages, then the first batch of composable components.
See what's included$2,900/mo
In this case studyNext batches, QA, office hours and workshops, team by team, until it's in the product.
See what's includedThe work above was done by our founder as a CARS24 employee. Layer One now runs the same method as a studio, for product teams of any size.
Aniket Pandey led this as Design Systems Lead at CARS24 (2025), with the CARS24 system designers and the product design teams in India, the UAE and Australia. This is employee work, not Layer One client work. Screens and component sheets belong to CARS24 and are shown for portfolio purposes. The sketch boards are annotated for this page; the planning board is a recreation and its tallies are illustrative.