Forty tons, highway speed, six inches of screen.

Context
The user is driving
Problem
Where the flow broke down
The app got in the driver’s way in four places.
Place corrections
A driver sees wrong data about a place, and the app asks him to describe it in words.
Constraints
Everyone but the driver
Everything below was decided inside these:
Not one conversation with an end user. Everything I knew about drivers came through experts and open sources.
No metrics before or after. Every number on this page counts work done, not its effect.
Three months — the entire engagement, including learning the industry.
The first two aren't an excuse, they're the terms of the job. A designer who says "we can't work without research" doesn't work; a designer who passes their own guesses off as user needs works badly. What follows is the third option.
Method
Someone you can't ask
Three independent layers that check each other
The layers were chosen so the sources wouldn't depend on each other: where conclusions overlapped, confidence went up; where they diverged, that was the signal to check.
Competitor analysis
Direct

Trucker Path

TruckMap
Indirect

Waze

Google Maps
Benchmark and opportunity map
What I had to learn first
The domain was studied across all three months, in doses, alongside the work. I started on tasks two weeks in. What had to be learned from zero:
Drivewyze, PrePass: what they are, how they work, what they save a driver
Bypass systems
and above all, why drivers dislike them and avoid them whenever they can
Weigh stations
Day four: yellow instead of black
The first test of the method arrived before I had a name for it.
Day four, a small task: change the colour of markers for truck stops with fuel discounts. They were black. I proposed yellow, and argued it not from taste but from what yellow already means to an American driver, twice over. It's the colour of a discount, fixed by retail practice, Walmart among others. And it's the colour of fuel itself — the very thing these stops discount. The app is strictly US-facing and its audience is local drivers.

Three months later the same move would give the project its core. Only instead of a price tag in a supermarket, it would be a road sign.
Solution
Removing unnecessary steps
Three separate user flows, each with its own solution: one entry screen for signing in and signing up, feedback sent from inside the app, and place corrections made directly in the place's data. All three were designed alongside the main work on weigh stations.
01 / 03
Sign-in: one screen instead of two
Before
Solution
One screen for both, labelled sign in or sign up, with three ways in instead of one: Apple, Google, email. Whether the account already exists became a check on the backend, not a choice the driver makes at the door.
Rejected
Doing exactly what was asked. Adding an email button to the existing split flow — faster, safer, no questions asked of the designer. Rejected because it would have closed the ticket and left the problem.
Before

After

One screen for both, and three ways in instead of one.
02 / 03
Feedback: don't push people out of the app
Before
Solution
Feedback inside the product: three types — suggest a feature, report a problem, share feedback. Plus an optional screenshot or video and two optional contact fields, phone and email, so a report can be acted on rather than merely received.
Rejected
Before

After

Three report types in tabs — you see what's being asked before you commit.
03 / 03
Place corrections: don't describe it — fix it
Before
Generic checkboxes and a text area. Someone who sees a specific error — wrong opening hours, wrong address — has to translate it into free text. At the other end it becomes a vague report where it’s unclear what to fix.
Solution
Six structured edit types with the current POI data prefilled. You correct the exact field in context instead of describing the problem from scratch. Less typing, more precise edits.
Where six came from
From the structure of the new POI data — what is actually editable — and from the competitor teardown: which corrections they allow and what they call them. The set was derived, not invented.
Hardest place first
I started from the hardest place on the map. A weigh station carries the most data, and some of it exists nowhere else — station type, bypass support. Designing corrections for it produced a template that the simpler places, truck stops and parking, fit into as a reduced version of the same set. This wasn’t about saving time: design from the simple case first and every next place type needs its own pass, until editing spreads into several incompatible mechanics.
Before

After

Weigh stations
Reading the road, not the map
The main task of the project. I started with it and came back to it three times:
Stations → Sign-in → Stations → Feedback → Stations → Place corrections.
Why this is more than a map point
A weigh station means delay, inspection, a fine and a broken schedule. Understanding why drivers avoid them is half the task.
What the app showed: the station and where it is. Bypass support wasn’t on the map yet, that data was still being collected.
What the driver needs: is it open right now, does it support my bypass provider, is there parking? Whatever was there had to be dug out by tapping. At highway speed a tap isn’t interaction. It’s distraction.
The marker as a decision model
I stopped treating the marker as an icon and started treating it as layers of answers. The goal was to move the most critical signals into the zero-tap layer — what can be read at a glance along the route.
The difficulty wasn't inventing the layers. It was fitting four signals into one marker without turning the map into noise. The marker had to stay readable in peripheral vision.
An arrow instead of a scale
The most domain-specific decision of the project, and a direct continuation of the yellow-marker story. On US highways weigh stations are announced by large green signs — "WEIGH STATION →". A driver has been trained for years to read that arrow as an instruction: get ready, slow down, decide whether you're pulling in. The response to it was built long before he opened the app.
I replaced the scale icon with that arrow. The old blue marker denoted a place but not a meaning — it had to be decoded. An arrow is simply recognised on sight, because the person sees it every working day on the road.
That is what reading the environment means: the interface doesn't invent a language, it borrows the one already working outside.


BEFORE
1 dimension
Location only

AFTER
3 dimensions
Status · Provider · Parking
The interface borrows a language drivers already read at highway speed.
Filters with counts
Toggles control which stations appear on the map. Each one carries a count: how many stations along this route support that provider. That stops being a filter and becomes an answer to "how many times will I have to stop on this route" — available before pulling out.

A count next to each toggle answers a route question before departure.
Not shipped. The weigh stations flow as a prototype.
Outcomes
Two of four reached users
On weigh stations. I presented the first iteration to stakeholders — as work in progress, not as a finished solution. The task needed further options and I kept looking for them. The state I stopped at was never presented: the work was cut short when I left the company.
On trip planning. Three concepts for comparing routes, surfacing key indicators earlier — including how many weigh stations sit on each path. None of them won out: the task ran in the background the whole time.
On place corrections. I left and don't know what became of it. I can't find those screens in the published app — I assume backlog, but I won't claim it.
What reached users: two tasks out of four. That part is publicly verifiable — the app is in the App Store and on Google Play.
Where this case study’s evidence ends
Validation with real drivers. Solutions were checked by domain experts — former truckers. They know the industry, but they aren't current users of the product.
A complete picture of the user base. Forums and YouTube represent the drivers who post, not all of them.
Takeaways
The environment is data
The best interface here was the one that required less thought. A six-inch screen in the cab of an eighteen-wheeler isn't a browsing environment, it's a decision-making environment. That changed my approach: instead of adding detail, I subtracted effort. For products where the cost of a mistake is real mileage and real money, reducing mental load matters more than visual richness.
When you can’t ask the user, you can read their environment. Distance from users is a reason to look harder. You can’t bring a trucker into a lab, but you can look at what he sees every day: a green sign on the highway, a yellow price tag in a supermarket. The environment is data that’s already been collected — by someone else.
The method travels. It isn't about trucks. It's about whatever domain I'm dropped into next: find what the audience already knows how to read, and don't make them learn it again.
Thanks for reading
Let’s talk about your product

Next case
From one platform to four connected experiences
Waiter Jan · Restaurant Tech · 2025
→



