Case study
Clearway
A driving school that tells you where you stand.
A concept for a New York driving school for adults learning to drive for the first time. Students book lessons themselves, see what’s done, and hear from their instructor, not from a progress bar, when they’re ready for the road test.
- UX Design
- UX Writing


About this project
Key results
- decisions loggedeach with what I turned down
- 15
- things Clearway refuses to doeach replaced by something calmer
- 8
- named road-test skillsinstead of an hours bar or a percentage
- 7
A concept, not yet tested with learners. These numbers count decisions, not outcomes.
Overview
- The problem
- A real driving school promises self-service and runs on phone calls. Learners can’t book after hours, can’t see their progress, and can’t tell when they’re ready for the road test.
- The goal
- A product where an adult learning to drive can answer three questions without calling anyone: where am I, when is my next lesson, and am I ready.
- What I did
- Wrote a brief from the real school’s site and New York’s rules, then designed the dashboard, booking, readiness and the landing page. Then the instructor’s side, the iOS and Android app, and the brand.
- The result
- One product across four platforms, with every major decision logged next to what I turned down. It hasn’t been tested with learners yet.
The problem
Context
The school promises self-service and delivers a phone call.
I started from a real driving school’s website. It tells students they choose their own driving slot, and that the school adapts to them. Right below, it explains how: the office arranges lessons by phone or email and sends an SMS reminder. The only thing the website itself can do is take a name and a phone number.
That gap leads to three problems.
Booking waits for office hours
Someone who remembers at 10 p.m. that they need a lesson on Thursday can’t do anything about it.
Progress is invisible
Over weeks or months there’s no record of what’s done, what’s paid for or what’s left. Every question is a phone call.
Readiness is a guess
Adults have no required hours, so people book the road test on a hunch: too early, or long after they were ready.
Repainting the site wouldn’t fix any of it. The problem was how the school worked, not how it looked.
What this is
A concept, on purpose.
There was no client. I haven’t tested Clearway with learners yet, so nothing on this page is presented as proven. What I can show is every decision I made, what I turned down, and where I got it wrong.
The goal
Brief
Answer three questions without calling anyone.
The brief I set first
A desktop-first website: four screens and three states. No app, no instructor side.
What it became
A website, an iOS and Android app for students and instructors, and the school’s brand.
The scope grew twice, and both times for a reason. Readiness only means something if you can see who signed it off, so the instructor’s side had to exist. And lessons happen out of the house, so the product had to fit in a pocket. The process below shows both.
What success had to look like
- Honest
Readiness that never overstates
It shows what’s missing and who made each judgment.
- Prevent, don’t report
Rules visible before you choose
Not an error message after you click.
- Attributable
Every decision explained
With the options I turned down written next to it.
How I measured it: I haven’t tested with learners yet, so each criterion was checked in design review, and the decision log is the record.
Out of scope, on purpose: checkout, sign-up, theory study and the school’s admin side.
The process
How each step led to the next.
01Starting point
I invented a school with a different promise.
- The question
- Was the problem how the site looked, or how the school worked?
- What I did
- Read the real school’s site line by line, comparing what it promised with how it actually ran.
- Decision
- Invent a New York school, Clearway, instead of repainting the old site.
The question I kept asking was simple: what would this look like if it were built for the person learning, and not for the person selling lessons?
I called the school Clearway and set it in New York. Plenty of people there learn to drive later in life, because they never needed a car. New York also requires a 5-hour course from every first-time driver, which gave the product real steps to track: learner permit, the course, practice, road test.
02Who it’s for
One scenario, instead of a persona.
- The question
- Who am I designing for, without research participants?
- What I did
- Defined the audience by their situation: adults learning for the first time, with no one to practice with.
- Decision
- One scenario instead of a persona: Ana, kept true on every screen.
The audience is adults who never learned to drive and are learning now, for the first time. They grew up with transit, or without a car at home, or moved to the US as adults. They pay for lessons themselves and book them around a job.
- No one to practice with
Every practice hour is paid
No parent in the passenger seat and no family car. The school’s car is the only one they’ll drive before the test.
- Lessons around a job
Office hours mean no booking
A lesson that has to be arranged by phone during the day is a lesson that doesn’t get booked.
- No way to tell
Readiness is a private guess
Nobody sets required hours for adults, and not knowing at 30 feels more awkward than at 17.
To keep myself honest I wrote one learner and kept her on every screen. Ana is 33. She moved to Queens from Belgrade, where she never needed a car. She works full-time, so she books at night. She has her permit, she’s done the 5-hour course, and she’s six lessons into a package of ten.
A persona would have looked like research. Ana is something else: one learner with real dates and a real next lesson. If a number on the booking page disagreed with the dashboard, the design was wrong.
03Three questions
A learner opens the product with three questions. I built everything on them.
- The question
- How should the product be organized?
- What I did
- Reduced the brief to the questions a learner actually asks, and checked every screen against them.
- Decision
- The dashboard answers all three at once. Booking and the road test get their own pages.
Where am I? When is my next lesson? Am I ready for the road test? Each one became a block on the dashboard, so all three are answered the moment it opens. The second and third also got their own pages: booking and the road test page.
Every screen after that was checked against the same three questions. If something didn’t help answer one of them, it had to justify being there.
04Where am I
Where am I?
- The question
- What does a learner need to see first?
- What I did
- Mapped New York’s path to a license and compared the path with an activity feed and a big progress graphic.
- Decision
- The dashboard opens with the four-step path. Finished steps show when they were finished.
Getting a license in New York is four steps, and most learners can’t name them. The dashboard opens with the path: learner permit, 5-hour course, practice, road test. Finished steps say when they were finished. The current step says how far in you are.
I rejected an activity feed, because it buries the next lesson under history, and a big progress graphic, because it spends the best part of the page on decoration.

05Booking
When is my next lesson?
- The question
- How does someone book at 10 p.m. without making a mistake?
- What I did
- Turned the school’s scheduling rules into the flow: two lessons a day at most, shifts including Saturdays, a free change of instructor.
- Decision
- Instructor, then day, then time, one week at a time. Every time you can’t book says why.
The next lesson sits beside the path, with the pickup point and the last moment it can be changed for free. Booking another one runs in a single order: instructor, then day, then time. The current instructor is already chosen, and “Any instructor” is one click away.
The rule I held to was rules before you click, not errors after. A time the student can’t book is shown as unavailable, with the reason written on it. Booking shows one week at a time, because more than that turned a decision into a search.

06Readiness
Am I ready?
- The question
- How do you show readiness when New York sets no required hours?
- What I did
- Ruled out an hours bar and a percentage, then built readiness from the instructor’s sign-off on each skill.
- Decision
- Seven named skills, each with who signed it off. Highway merging sits apart, beyond the test.
This was the hardest one. New York doesn’t require a set number of practice hours, so an hours bar would measure the one thing that doesn’t matter. A percentage would claim a precision nobody has.
Readiness in Clearway is a list of seven named road-test skills. Each one is Not started, Practicing or Signed off, and each sign-off shows who gave it and when. The summary is a count: 5 of 7 road-test skills signed off. It’s never a ring or a gauge.
One skill didn’t fit. Highway merging matters for real driving, but the New York test is driven on local streets. Counting it would block readiness on something the test doesn’t check, so it sits in its own group, called beyond the test.
The page is also honest about whose rule is whose. The permit and the course certificate are the DMV’s requirements. The skill sign-offs are Clearway’s own rule for lending its car on test day, and the page says so.

Ready is a judgment, not a number of hours.
07What it won’t do
Most of the design is things Clearway refuses to do.
- The question
- Where would a school that sells lessons be tempted to push?
- What I did
- Kept a log of fifteen decisions and wrote down, next to each, what I turned down.
- Decision
- Eight things Clearway refuses to do, each replaced by something calmer.
I kept a log of fifteen decisions, and next to each one I wrote down what I’d turned down. Reading it back, the rejections were the product. A school that makes money from lessons has every reason to pressure a learner. Clearway is what happens when it doesn’t.
| Clearway won’t | What it does instead | Why |
|---|---|---|
| Counts named skills, with who signed each one off | A percentage implies a precision nobody has | |
| Updates the slot in place, offers the three closest free times and keeps your instructor and day | A timer adds pressure and wastes the choices you already made | |
| Uses the lesson from your package; no card is charged | The package already holds the money, so a fee is only a punishment | |
| Inside 24 hours, Keep lesson is the main action, with a way to message the school if something serious happened | The cheapest choice should be the easiest one | |
| Shows a low package inside the package module, with the instructor’s estimate and the smallest top-up first | A pop-up at login is a paywall in disguise | |
| No reason field, and your sign-offs stay with you | Any field implies an explanation is owed | |
| Moves the lesson; the old time is held until the new one is confirmed | You should never end up with no lesson | |
| Shows the product itself on the landing page | For an invented school, both would be made up |



08Instructor side
Readiness is only believable if you can see where it comes from.
- The question
- Where does a sign-off actually come from?
- What I did
- Went past my own brief, which left the instructor out, and designed what an instructor fills in after a lesson.
- Decision
- One lesson log on a phone: tap each skill, add a set phrase, save.
If the student’s readiness comes from the instructor, the instructor’s side had to exist. Leaving it out would have made the whole readiness page a nice-looking claim.
I designed it as one lesson log on a phone, because that’s what an instructor has in the car. After a lesson they tap each skill to Practicing or Signed off, add a set phrase if it helps, and save. It takes about thirty seconds, which matters when you teach seven lessons a day.
The set phrases are written in the product’s voice, because the student sees them exactly as the instructor chose them: “Curb contact on final turn,” not a grade.
Signing off a skill asks for one confirmation, because it changes what the student sees. If a sign-off is ever reversed, the student sees that too, with a name and a date. Readiness never changes silently.



09Voice
Clearway speaks like a good clinic: clear, factual, unhurried.
- The question
- What should it sound like to an adult who’s a little embarrassed to be learning?
- What I did
- Wrote the copy for every screen and state before designing anything.
- Decision
- A calm clinic: state, don’t cheer, and name who made every judgment.
I wrote the copy for every screen and state before I designed anything. It was the cheapest way to find out whether the product was honest, because pressure shows up in words first.
| Instead of | Clearway says |
|---|---|
| 5 of 7 road-test skills signed off | |
| Free to change until Mon, Sep 28, 6:00 PM | |
| 2 lessons left. Marcus’s estimate: 2–4 more. | |
| Marcus is booked | |
| Book |
The rules behind them are short. State, don’t cheer. Name who made every judgment. Put a number where a decision is. No urgency, no exclamation marks, and nothing on screen a colleague could read over your shoulder and learn something private.
The landing page follows the same rules. It has no reviews and no statistics. It shows the booking page, the path and the readiness list, and lets them make the case.
From the landing page
“We won’t tell you you’re ready before you are, and we won’t sell you lessons you don’t need.”


10Four platforms
It started as a website. It needed to be in a pocket.
- The question
- Lessons happen out of the house. Is a website enough?
- What I did
- Carried the website to mobile web, iOS and Android, keeping the structure and the words.
- Decision
- One app for students and instructors, with different tabs and native controls on each platform.
Lessons happen out of the house, and the reminder that matters arrives the evening before. So after the website I built the app, for iOS and Android. Students and instructors sign in to the same app and get different tabs.
I kept the structure and the words, and let each platform keep its own habits.
Signing in
Face ID on iOS and fingerprint or face on Android, through each system’s own prompt.
One reminder
“Lesson tomorrow,” the evening before, for students and for instructors.
Native chrome
A Liquid Glass tab bar on iOS, a Material navigation bar on Android.
Sheets
Side panels on desktop become bottom sheets on a phone.




11Brand
I designed the identity too: a C drawn as lane markings.
- The question
- Does an invented school need an identity at all?
- What I did
- Designed a mark that has to work on a car door and in a 16 px browser tab.
- Decision
- A C drawn as lane markings, solid below 24 px, with two app icons.
An invented school still needs a name on the car. The C is a lane line: evenly spaced dashes that read as road markings, with one blue dot. Below 24 px it switches to a solid C, so it still holds up in a browser tab and as an app icon.
On the cars the same dashes run along the lower doors as a lane stripe. The roof sign sits on ink so it reads at night, and the app gets two icons: ink for students, white for instructors.
Light and dark
Light and dark.
On the website the switch sits next to Help, because for some people it’s an accessibility setting and not a preference.






12Accessibility
I checked the whole file against WCAG, then fixed what failed.
- The question
- Do the colors and tap targets hold up, or only look like they do?
- What I did
- Checked the whole file against WCAG and fixed what failed.
- Decision
- All 25 dark-mode pairs I checked pass, targets are 44pt and 48dp, and no status relies on color alone.
The check found problems in places I thought were fine. Tertiary text and the toggle track were too light. Text on photos needed a stronger overlay. Some links and chips were too small to tap.
Contrast
The failing colors were fixed, and all 25 dark-mode pairs I checked pass.
Target size
Links and chips are at least 44pt on iOS and 48dp on Android.
Never color alone
Every status has text and a dot or icon. A struck-through time also says why.
Focus and hover
Desktop controls have both. Phone-only controls don’t pretend to have hover.
This was a design check, not a test with assistive technology. That’s still to do.
The outcome
What held up, what’s still unknown, and what I’d test first.
Did it work?
Everything I could check in design held up. None of it has met a learner yet.
| Goal | What happened | How I know |
|---|---|---|
| Three questions at a glance | The path, the next lesson and readiness sit together at the top of the dashboard. | Design review |
| Honest readiness | Seven named skills, each with who signed it off and when. No percentage, no hours bar. | Design review |
| Prevent, don’t report | A time you can’t book is marked before you click, with the reason on it. | Design review |
| Hard moments designed | Slot taken, late change and a package running low are drawn as screens, not described. | The screens above |
| Attributable | Fifteen decisions logged, each with what I turned down. | The decision log |
| Works for real learners | Not yet known. | Still to test |
Reflection
Four platforms, and it still had to feel like one product.
Designing one website was manageable. Carrying it to mobile web, iOS and Android was the hard part. Every screen raised the same question: what has to stay identical, and what should follow the platform?
- 1
Same structure, same words.
The order of information and every label stay the same on all four, so a student who books on a laptop recognizes the app right away.
- 2
Navigation follows the platform.
A Liquid Glass tab bar on iOS, a Material navigation bar on Android. Copying one onto the other would have looked consistent in Figma and wrong on the phone.
- 3
Panels become sheets.
A side panel works on a wide screen. On a phone the same content comes up from the bottom, where a thumb can reach it.
- 4
Touch isn’t a small desktop.
Targets are at least 44pt on iOS and 48dp on Android, and hover exists only where there’s a pointer.
The rule I settled on: the content never changes, the controls can.
Consistent doesn’t mean identical.


Next steps
Every decision here is still a hypothesis.
I believe in these choices, but I haven’t watched anyone use them. These are the first things I’d find out.
- Does “5 of 7” read as progress or as a verdict?I’d test the readiness page with five first-time drivers and ask what they think happens next.
- Does the late-change rule feel fair?Losing a lesson is the hardest moment in the product. I’d watch whether people understand it before they decide, not after.
- Can an instructor really log a lesson in thirty seconds?That number is my estimate. It needs a real instructor in a real car.
- Are the New York facts right?The road-test route, the accompanying-driver rule and the course requirement need checking against the DMV’s own pages.
- Does it work with a screen reader?Contrast and target sizes are checked. Reading order and labels aren’t.
I set out to redesign a website. I ended up building a brand.
The brief I gave myself was a better driving school website. By the end there was a name, a logo, a voice, the school cars and four platforms that all belong together.
That’s what I’d take forward. A brand like this is a foundation: whatever the school needs next, a new page, a new feature or a sign on the street, already has a starting point.
— Iakov Grechukhin, UX designer & writer










