Rebuilding a Georgian ticket app around the trip, not the catalogue

What I did, and what was not mine

• Mapped the structure of a funded travel startup before this one: seven sections, thirteen branches under them, seven more under settings alone.
• Audited the live app from the outside: ten categories, 98 items of stock, four parallel navigations.
• Redrew the information architecture into four tabs and moved the old side menu under the last of them.
• Designed 35 screens at 390 px: onboarding, sign-in, discovery, event page, travel, checkout, profile.
• Built the kit under them: 54 variables in three collections, 14 text styles, 35 components.
• Rewrote the interface in English, replacing a catalogue that ran in three languages at once.
• Modelled the onboarding key visual in Cinema 4D.

Three things I did not have

No access to the company. Every number about the old app was counted from outside: twelve screenshots, and the stock counters the app prints beside its own categories.

No data from the startup. Four months, a small funded team, nothing in production and no analytics left behind. The structure map survived, so that is the only thing from it in this case.

No sessions. Five people on the prototype would settle whether four tabs beat four navigations. They have not happened, and nothing below is written as though they had.

A travel marketplace that never shipped

In 2023 I spent four months as the designer on a small travel startup: hotels, activities and flights in one place, with crypto settlement planned on top.

I mapped the whole structure — search by place and by date, filters down to the type of bed, a merchant side with bookings and payouts, taxes, referrals and account settings.

It never shipped, and not because the idea was wrong. Nobody owned the plan, calls never turned into dated steps, and a team that had only met online came apart.

Biletebi sells the same shape of thing — a place to be, a date, and the way there — and unlike us it is running. So I took the structure work there.

The app had everything

Which was the problem. You open it with a free evening in Tbilisi, and the first screen gives you eleven carousels, a side menu of ten categories and five tabs along the bottom.

Not one of them shows a price. Behind all four systems sat 98 items, counted from the numbers the menu prints beside its own categories.

The stock I counted from the side menu's own numbers: ten categories, 98 items between them.

A card told you the crowd, not the price. “Attendance: Good”, “Attendance: Cult”, “Attendance: No Info” — that was the fact every card carried.

Price and time were on neither the carousels nor the category lists. Comparing two concerts meant opening both.

The Eng switch moved the interface, not the content. Georgian, Russian and English titles shared one screen, and the switch in the header left every event name as it was.

The trip was already there, cut in half

Two of the ten categories were travel — Transport & Cableway, Leisure and Tourism — and two of the five bottom tabs were Railway and Bus.

So the redesign did not add travel. It stopped selling it separately: four tabs, with what is on, the catalogue by interest, travel, and your tickets.

The cost. The long tail loses its own front door. Museums, education and the partner clinics sit one level down now, so a person who came for a museum pass has one more tap than before.

The flow is drawn three times, and I cannot tell you why

Each pass fills in more: grey blocks and layout, then real events and prices in Russian, then the English screens that went into the prototype.

What changed between passes is visible in the file. Why it changed was never written down, and that is the first thing in the takeaways.

Seven bets, written down before the first screen

Restructuring the app, an FAQ to cut support load, social login and payment, an interactive calendar with reminders, filters and sorting, reviews and sharing, and letting organisers submit events themselves.

None of the seven was tested. They are design bets, and the case calls them that rather than findings.

The fees arrive before the commitment

The seat map breaks the price into three lines on the screen where you pick the seat: ₾20 × 2 tickets, taxes fee, service fee.

The cost. The total looks bigger earlier, and some people will leave at that screen. The alternative is a surprise at payment, which costs trust instead of a session.

The kit says iOS, and four of its eight sizes actually are

The type scale runs 38 / 32 / 24 / 20 / 16 / 13 / 12 / 10. Apple's runs 34 / 28 / 22 / 20 / 17 / 16 / 15 / 13 / 12 / 11, so four line up: 20, 16, 13 and 12.

Calling the scale “aligned with iOS” would be a claim the file does not support.

Under the screens: 54 variables in three collections, 14 text styles, 35 components, two effect styles. Colour is only ever a variable — the file has no colour styles at all.

The colours carry two modes; the screens do not. Light and dark are both defined, and all 35 screens are drawn light.

What I would measure

No numbers are claimed in this case. The first three I would instrument: share of event pages that end in a paid ticket, drop-off on the seat map where the fees appear, and share of event purchases that add a travel ticket.

Before any of that, five people on the prototype buying a concert ticket in another city. Structure is the one thing a map cannot confirm.

What I would do differently

• Ask who owns the plan before starting. The startup had a team, an idea and nobody accountable for sequence or dates.
• Keep the reasons next to the file. Three passes of every flow, and not one line saying why a pass changed.
• Treat structure as untested until someone tries it. Four tabs are an argument, not a finding, and this case never pretends otherwise.