Denis Kharchenko — Lead Product Designer
EVmarket · Case Study
Poland's marketplace for electric cars, designed for people who already spoke EV and used mostly by people still learning the words.
People were leaving the marketplace before they'd opened a single car. On paper that's a conversion problem, and that's how it had been treated. What was happening was simpler and more stubborn: the product had been written for buyers who already knew EVs, and most of the people arriving didn't yet.
So I rebuilt the first session around the people still learning: an inline glossary, a cost-of-ownership calculator right on the profile page, a car layout reordered to answer questions before it listed specs. After the redesign, compared to the pre-redesign baseline, first-session drop-off fell 13.2 points, first-visit dealer contacts rose 12%, and 32% of new users used the inline glossary. The product no longer needed a separate glossary sitting next to it.
Results: first-session drop-off down 13.2 percentage points; first-visit to dealer contact rate up 12%; 32% of new users used the inline glossary. Compared to the pre-redesign baseline, 09/2023 to 06/2024.
EVmarket assumed its users already understood electric cars: range, kWh, charging connectors, the whole total-cost-of-ownership picture. Everything followed from that assumption. Filters were labelled in technical terms, car profiles opened with specs, and the comparison tools only paid off if you already knew what you were comparing.
Most of the people showing up hadn't crossed that line yet. They arrived curious and left confused, usually before they'd opened a single car profile. The dashboards read it as a conversion problem. It was a vocabulary one.
I started in the Polish EV forums, where curious buyers were already putting their questions to each other. The busy threads weren't about which car to buy. They were about whether to make the switch at all.
Then I sat down with five current EV owners, people who'd already crossed that line. I wasn't interested in which cars they liked. I wanted to know what had finally reassured them enough to commit.
What tipped the decision wasn't price or sustainability. In every interview there was a particular moment where doubt turned into permission to buy. One person did the home-charging math and saw it beat petrol. Another realised the battery warranty outlasted how long they'd even keep the car. Savings and emissions mattered, but only once the practical worries had practical answers.
Other themes that kept surfacing:
None of the doubts were irrational. They'd just never been answered.
None of these are buyers' questions, exactly. They're what people ask while they're still talking themselves into the idea, and the product had nothing to say to any of them.
I wrote one persona, on purpose. Adam, 34, lives in the suburbs of a big city, curious about EVs, knows the basics, has never driven one, and doesn't quite trust what he knows yet.
He's not who the original product was built for. He's who it was quietly losing.
Designing for him meant taking on a constraint the product had spent years dodging: the first session has to teach a little before it's allowed to sell.
The buying journey runs through six stages, from first curiosity to advocacy. Doubt doesn't spread evenly across them. It piles up at two: the first time someone opens a car profile (Exploring), and the moment the cost question comes due (Doing).
Those two became the centre of gravity for the whole redesign. Pretty much every decision that follows is aimed at taking doubt out of one of them.
Everything on the first-session screens had to answer one question to earn its spot: does this help Adam work out whether he could own an EV? If it did, it made v1. If it didn't, it waited, even the things that had tested well.
What I cut: three things pulled from v1, all of which had tested well:
Five moderated sessions, 30 minutes each, using a Figma prototype.
Watching five people work through the low-fi prototype told me more than any amount of internal debate had. Most of what follows came from where they hesitated, and from what they said when I asked why.
What broke:
What worked: The cost-of-ownership calculator was the clear standout. People adjusted inputs, ran it again, and came back to it later in the same session. It was the thing most of them named when I asked what made the product feel useful.
Seven users took the high-fidelity version, and it turned up one finding worth acting on.
The EV part is almost beside the point. This is what happens to any product that quietly assumes its audience has already done the homework. Build for the fluent and you stop teaching everyone else the language, and the numbers catch it before you do. First a dip in conversion, then a churn pattern, and eventually a quiet consensus inside the team that you've got the wrong audience. The audience was never wrong. The product was mistranslating itself.
More case studies and notes: denisxko.com · full text of everything: /llms-full.txt
© Denis Kharchenko. All rights reserved.