Every car has a story. What makes each one special?
Cazoo
2022
At Cazoo, a UK company that sold used cars online, I was the designer in the Select team, the team that helps people choose between cars at the end of their search. Our focus was the vehicle detail page, or VDP, as we called it internally: the page you see when you open a car. It was already good. But it had turned into a long list of facts, changed one small test at a time, and the team didn't have a shared sense of where it was going. This is the story of how we created a vision for it, around the thing Cazoo was really selling: trust.

Selling cars without the salesman
Used car salesmen don't have the best reputation. Pushy, a bit sleazy, never quite telling you what's wrong with the car.
Cazoo set out to be the opposite. Cazoo bought every car itself, fixed them, and told you exactly what it had fixed. What it didn't fix, small scratches and chips it called imperfections, it photographed and showed you.
That honesty was the whole deal! People were buying a car they had never seen, committing to it online, and waiting for it to arrive at their door. The only thing that made that work was trust, and every page had to earn it.
Past the first version
When I joined Cazoo, the website was really good. Searching for cars, looking at a car's vehicle detail page (VDP): none of it was random design work. And the proposition was solid. If you didn't love the car, you had 7 days to send it back for a full refund, and every car came with a 90-day warranty.
That meant every team I joined was an optimisation team. The product was well built, so the work was making it better, bit by bit. On the VDP that meant A/B tests on small things: detail A, then detail B. A banner here, a counter there.
Each test made sense on its own. Together, they didn't add up to anything. The page kept changing, but no one could say what it was supposed to become.
A spec sheet only tells a story if you know your car stuff
Open a car on the live page in 2022 and you'd get the facts: mileage, year, gearbox, fuel, then lists of labels and values. All correct. All useful. But every car on the site read the same way.
For someone who knows cars, a spec sheet is a story. They read "2.4L, 228 bhp, 46,625 miles" and already know if it's a good deal. But research told us those weren't the people actually visiting Cazoo. Many of our customers were people who didn't want to deal with salespeople at all. They didn't want to go to a car dealer, and they half expected to get ripped off if they did. Not having to do that was a relevant part of why they came to Cazoo.
A used car has a story to tell: who owned it, what was fixed, what it came with, what it'll cost you to run. When you buy in person, someone walks you through that story. Online, the page has to do it. And the story isn't about the horsepower or the number of doors.
The page also had a job to do for the business. Cazoo was losing money at the time, and every car sold had to bring in more for the company to reach break-even. We wanted people to click "Buy now", but finance and extras like paint protection mattered just as much. On the live page, those extras sat in a block next to the price and again at the bottom, a bit out of context.

What research already told us
Research was a really big thing at Cazoo, and that shaped this work more than any single study.
Take customer interviews in other countries. To learn how German customers saw the VDP, for example, we would write an interview script, and an agency would run the interviews in German. Live translators made sure English speakers could follow. Anyone at Cazoo could sit in and listen, mostly designers and sometimes product managers. As a designer, it's rare to work somewhere that does research almost by the textbook, and I knew I was lucky.
Then there were the support calls. Cazoo had customer support by phone, and every call was recorded. There was no AI to transcribe them, so the whole team, engineers, product manager and designers, would pick calls at random and listen together, for about an hour. What were people asking? What could we fix? It became a ritual, started by Matthieu, our product manager, who wanted to build that habit with the team. Listening together also gave us a shared sense of who we were building for.
I listened a lot. What I heard in those calls and interviews stayed with me, and much of it found its way into the page.
The VDP bashing sessions
Remember those A/B tests on small details? They're great at making one detail better. But run enough of them and you can end up polishing corners of the page while nobody looks at the whole. Every time we changed one part of the VDP, we lost a bit more of the bigger story: how do you sell a car to someone who has just landed on it, and what questions do they have?
There was a human side to it too. A team running test after test on small details can lose sight of why it's doing them. People need a sense of mission: something bigger than the next test, that helps them focus, gives them direction and keeps them motivated.
So when I joined the Select team in June 2022, I gathered every UX person I could in about 10 sessions. We called them the VDP Bashes. The goal: a clear picture of where the VDP should be in a year's time. Two fellow designers, Imogen and Andreia, shaped it a lot.

This was mid-2022, about five months before ChatGPT came out. A conversation on a web page couldn't be generated. It had to be designed, section by section.
What the vision looked like
The vision gave every section of the page the same flexible pattern: a title, a short paragraph in plain language, and the good stuff highlighted.
Inside that pattern, the page started to talk:
Say what's good about this car. Not "Mileage: 3,376" but "3,376 mi only". Low mileage on a used car is a big deal, so present it as so.
Answer questions where they come up. The three questions support heard most on the phone were: can I test drive it? How does finance work? Can I part exchange my car? We put them in the middle of the page, not at the bottom. Every question the page answers is a call support doesn't have to take.
Sell extras in context. After the features and the photos, when the car looks great, comes the natural question: how do you keep it like that? That's where paint protection belongs. Not in a block of its own.
No dead ends. Working on media websites taught me that a page should never be a dead end: there's always something worth reading next. The VDP ended with nothing, not even other cars. But Cazoo had written a lot about cars, and before joining the Select team, I had worked in Cazoo's content team on how those articles were organised. So the vision ended each car with the articles about its make and model. Many of them put cars up against each other, so they showed we weren't playing favourites. One more way to keep the conversation going, and to build trust.

At the time, Figma couldn't handle some of the interactions we wanted in the new vision, and building a real prototype in it was very hard. So I went back to my front-end roots and built the vision from scratch in HTML and CSS, using Jekyll: real cars, real photos, real articles, all linked to each other through the same organisation we used for the articles. It took a lot of extra hours. But you could scroll through a car and feel the difference.
We never planned to rebuild the VDP in one go. That wasn't the priority, and it didn't need to be. Our goals and priorities kept deciding what we worked on. But whenever a piece of work touched a section of the VDP, we already knew what that section should become, so we did the extra work to move it there.
What makes this car special?
To put the vision in place, our strategy was to move each section of the VDP towards it whenever a piece of work on our roadmap gave us the opportunity. The features section got that opportunity first.
Picture two Ford Kugas, same year, similar price. One of them came with £2,895 of extras when it was new. How would you know?
On the VDP, you wouldn't. The features section only listed what Cazoo's technicians had checked on the car itself: a short, reliable list, the same few lines for almost every car.
Then the business wanted to try something new. An outside data provider knew a lot more about each car, including the extras it was fitted with when new, some of them worth a lot of money. Showing them might help people choose, and sell more cars. And if we stopped relying on technicians to check every feature, cars would spend less time in refurbishment, and Cazoo could get more of them ready each day.
For us, it was the opportunity we'd been waiting for. We were going to change this section anyway, so we could move it towards the vision at the same time.
The extras
The extras were the first thing that came to mind when we asked what makes a car special: the things someone once paid more for. We couldn't highlight everything. If everything is a highlight, nothing is. So we started with those.
The section followed the vision's pattern:
A sentence instead of a list: "This car was fitted with 4 extra features worth £1,278 when it was new."
Each extra highlighted, and every other feature one click away.
Interior photos right there, because another experiment had shown they increased engagement and conversion.
Paint protection pitched in context, to get more people to add it.
But extras only tell part of the story. A Tesla comes with almost everything as standard, so it has very few extras to show off. What makes a car great is often what it comes with as standard: heated seats, a reversing camera, sat nav. But a car can have close to 100 of those. Which ones do people actually care about?
Which features do drivers love?
We used four sources to find out:
Card sorting. We asked people to sort 64 cards, one per feature, by how much each mattered to them. Two rounds of 32, then a final round of 24 with the best of both.
Search data. We looked at which features people filtered by on the search page. It was a small signal, since only a few people used those filters and only 24 features were available there, but it showed us what people went looking for.
Motoring press. Car magazines and websites regularly publish lists of the features buyers want most. We read a lot of them. We even went to Google's second page of results!
ChatGPT. It had just come out, so we asked it for the most popular car features in the UK. It was a bit scary how much sense it made: about 70% matched what we already had.

Where the vision paid off
Because we already had the vision, we didn't need to spend time exploring layouts and UI for the features section. We knew what it should become, and the pattern was waiting for it. That freed us to focus on the part that was actually new.
The challenge was fitting key features and extras together, clearly, without overwhelming anyone. With help from engineers, we looked at the data: a typical car had about 7 popular features and about 4 extras. We also designed for the worst case: no key features and 17 extras.
After a lot of iterations, we settled on a maximum of 16 highlights, shared between key features and extras. On mobile, a "Show more" button kept them from making the page endlessly long.

Beyond the VDP
Back to those two Kugas. On the VDP, knowing why a car is special is a nice touch. On the search results, where cars sit side by side, it matters even more. That's where people actually compare them.
The search results weren't our team's page. They belonged to the team where Andreia was the designer, the same Andreia who had helped shape the vision in the VDP Bashes. She already shared the picture, so bringing her on board was easy. The Bashes turned out to be good at that too: building bridges between teams that would otherwise each optimise their own page.
The new data about each car made it possible. With fitted extras and key features now known for every car, the search results could show badges like "£2,895 fitted extras" or "Key feature loaded" right on the card. Now you could tell, at a glance, why this Kuga is better than the average Kuga.

Cazoo. Yeah, you can't.
Cazoo's ads said "Cazoo, yeah you can". But while we were building and planning all this, the company was starting to crumble financially.
The company had grown fast and spent big: four new countries in Europe, sponsorship deals with football clubs, and a huge operation to buy, fix and deliver every car itself. But it made very little on each car it sold, far less than other dealers, and nowhere near enough to cover all of that. Every car had to earn more, and it wasn't happening fast enough.
In September 2022, Cazoo pulled out of Europe. In January 2023, it announced it would close the Lisbon office, where I worked. Our plan never fully played out. Reality got there first.
The extras went live. Key features went into implementation, and the search results stayed a design. We didn't get around to measuring. The plan was to watch conversion.
Matthieu, the product manager on my team, loved the vision. He was very data-driven, and every change still had to prove itself with numbers. But the vision helped him decide which changes were worth testing in the first place.
What I took away
You don't have to wait to be asked for a vision. Nobody asked us for one. But once it existed, it changed the conversation: instead of debating the next small test, we could talk about where the page was going, and which tests would get us there.
A vision doesn't have to ship in one go to be useful. Ours gave the team a mission, and when the next real piece of work came along, it was already half designed.
And people need to feel a vision, not just look at it. Static frames in Figma show what a page looks like. A prototype you can click through shows what it feels like to use. Ours had real cars and real photos, and everything was linked: from an article to a car, from a car to its make, from a make to all its models, and back to more cars. Seeing articles and car pages mixing like that is what sold the idea that content and the VDP could work together. That's what got people on board, and why I still prototype whenever I can.
