New Car Experience

Rebuilding the new car journey around the car customers want, not the individual vehicle in stock.

Project overview

cinch made its name selling used cars online, delivered to your door. It then became the first company in the UK to sell a new car completely online, end to end. To get live quickly, the new car journey launched as an MVP built on the used car journey. I was part of the launch team and later led the work to fix it.

The gaps soon showed. Our insights team flagged rising cancellations from customers who only found out after ordering that the car was far away. There was an opportunity too: get the journey right, and Marshall salespeople could use it to sell new cars in store.

The problem

New cars aren't used cars. cinch prepares used cars centrally, but new cars are sold through the dealerships of Marshall Motor Group, cinch's sister company, with stock always on the move. And because UK law gives online buyers 14 days to cancel, customers had to collect from the dealership instead of getting home delivery, so distance mattered.

Reusing the used car template made every physical car its own card and page, which caused four problems:

Duplicates. One model returned dozens of near-identical results.
Scattered choices. Colours and trims (the versions of a model, from basic to top-spec) were spread across listings, not shown as options.
Hidden distance. Customers only learned where a car was after ordering, so many cancelled.
Lost demand. Cars in transit or only available to order couldn't be listed, and brands without live stock or finance quotes could only take enquiries.

Before: one model, dozens of near-identical listings

My role

After launch, I led design on the next phase, working with product, engineering and service design. I owned the options deck, the page architecture, the design principles and the prototype used to align stakeholders.

Reframing the problem

All four problems had the same cause: the journey started with an individual vehicle. People don't start by picking one specific car off a list. Our insights showed they choose the model, trim and colour first, then work out how to get it, which is also how a salesperson sells one in store. Duplicates, scattered choices, hidden distance and invisible order-only stock all came from starting in the wrong place. I reframed the brief with product around the order customers actually follow: choose the car, see how each choice affects price, then see the best way to get it.

Approach

Making the case for trim

The working assumption was one page per make, model, trim and colour. I built a deck comparing the options and made the case for one page per trim, with colour chosen on the page. Salespeople, alongside online buyers, were a key customer group. During in-store interviews at Marshall dealerships, our service designer got their feedback on my early prototypes, showing the downsides of a page per colour against the alternatives. With that evidence, I walked product and engineering leads through the trade-offs, and the deck changed the direction:

Better for SEO. Each page ranks for its own trim name.
Easier to compare. Every colour in one place.
Accurate specs. Fixed to one trim, so they never conflict.
Less to maintain. Fewer pages as new cars are added.

A slide from the options deck: three ways to structure the page, from one listing per colour to one per model. One page per trim gave the best balance.

Designing the architecture

I mapped two page types: a trim product page for buying online, and the same page in an enquiry state for brands that only take enquiries.

I moved search to one card per trim, with colour swatches that update the price, and put order-only trims in a collapsed "Explore models" section so hidden stock was visible again. On the product page, customers pick their options and see the best way to get the car, from a nearby store today to cars still on their way. Customers can add a postcode on search, where it's optional, or on the trim page. They only enter it once, and every distance and nearest store updates from it.

Because the journey starts from the customer's choice, salespeople can use it in store too. I mapped the web journey while our service designer mapped the offline one, so we could see where each change helped customers and colleagues.

Making this work meant changing how stock is stored, from individual cars to a catalogue of trims and colours. That suits emerging Chinese brands in particular, with their simple line-ups. Traditional manufacturers, with large option catalogues, would need a full configurator, which needs its own discovery.

The end-to-end journey, mapped with our service designer, from search to collecting the car in store. Blue shows the new steps, green is Marshall’s part, and pink marks the open questions we worked through.

Setting design principles

I set four principles to keep decisions consistent across squads and edge cases:

Never define "too far." The obvious fix for cancellations was to hide distant cars. I pushed back: the issue was surprise, not distance, and hiding cars would lose customers happy to travel. Show the distance and let the customer decide.
Respect what the customer chose. Never silently swap a colour or trim they picked, even when a closer car exists. Colours are rationed between dealers, so their choice may be far away.
Only stop a sale when there's no car anywhere. If the car is far away, in another colour or arriving later, tell the customer and let them carry on.
Keep buying consistent, whatever the stock. In stock, far away or enquiry only, the buy box follows the same order: price, colour, availability, finance, then buy or enquire.

The same buy sheet in two stock situations. Whatever’s available, the customer always sees price, colour, collection and a clear way to buy or enquire.

Working with AI

Tools: Claude for early wireframes and prototyping; Figma for the final UI.

I used Claude throughout:

Weighing the options. Laying out the pros and cons of each page structure for the options deck.
Stress-testing. Running the principles against awkward scenarios to find gaps early.
Prototyping. An interactive greyscale HTML prototype of the full journey, so product and engineering could click through the architecture before visual design started.

The decisions stayed mine; AI got me to them sooner.

The greyscale prototype I built with Claude, used to click through the full journey with product and engineering before any visual design.

Outcome

I played back the work and product leadership agreed the trim-level direction. My job was to set that direction and lay the groundwork for squads to deliver it: an agreed page architecture, four principles to guide decisions, a prototype to build from and a clear first release, one card per trim on search, using stock we already have. I also fed the new car card back into our design system, so other squads can reuse it. Getting it right would also give Marshall's salespeople a tool to sell with in store.

Key learnings

Start from how customers choose, not how stock is stored.
Clear principles scale further than individual screens.
Knowing where a solution stops working matters as much as the design.

What I would do next

Measure each release against clear goals: fewer cancellations, especially those linked to distance, to test whether honesty converts better than hiding; more people moving from search to a product page; more enquiries turning into sales; and how quickly salespeople can find the right car with a customer. Then help customers compare trims, showing what each one includes over the others, since trim is now the main choice they make.

Before and after: from a page of near-identical cars to one clear card per trim.

Proposed search results on desktop: one card per trim, with a postcode lookup that shows each car’s nearest store location, so customers know how far away it is before they choose.