Comparing and booking rides across apps

Why the best ride is hard to find

• Lots of supply, spread across apps. London has 93,300 licensed private hire vehicles, according to the Department for Transport’s 2026 statistics. FREENOW became part of Lyft in 2025, and in 2023 Uber announced that black cabs would join its app.
• Prices move trip by trip. University of Oxford researchers analysed 1.5 million trips by 258 UK Uber drivers. After dynamic pricing arrived in 2023, passengers paid more per trip, and Uber’s average cut rose from about 25% to 29%.

So the cheapest app for one journey may not be the cheapest for the next.

A phone, in one hand, at the kerb

Steven Hoober made 1,333 observations of people using phones in streets, airports, bus stops, cafes, trains and buses. Of those touching the screen, 49% held the phone in one hand, and people changed grip often. That set the rules before any screen was drawn: big targets within reach of a thumb, text readable at a glance in the dark, and nothing that breaks when the signal drops.

Sixteen public sources, including the law itself

• Comparing has been tried. Karhoo, a London comparison app, raised $250 million and closed in 2016. Comparison apps today show the options, then open the provider’s app for you to book. Even Bolt in ChatGPT sends you back to Bolt’s app to finish.
• The largest provider rules it out. Uber’s developer terms for Europe say an app may not “include any underlying Uber platform or product with competitors in any aggregated view” without its written permission.
• The law decides who books. In London, nobody may accept a private hire booking without a TfL operator’s licence. TfL’s guidance says the operator that accepts a booking contracts with the passenger “as principal”.

As I read the Act and TfL’s guidance, the moment that makes the idea worth having, “it is on its way”, is the moment the app becomes the operator, answerable for every journey it books.

Comparing and booking sit with different companies

Four ways to make “it is on its way” true

• Becoming the operator for every ride delivers the whole idea, and makes the app answerable for journeys it does not run.
• Handing every booking over is what exists. It never books anything.
• Reading the rider’s own apps with an AI agent runs into terms like Uber’s, and breaks when an app changes.
• Booking through partners and handing over the rest was the only option that books for the rider without relying on access the providers have not given.

One pick, one reason and one button under the thumb

• One pick, with the rest one tap away. Kerbside makes the call and shows its working, so the rider only has to agree or overrule.
• The reason fits on one line. “Cheapest. The black cab is 9 minutes quicker for £3.80 more.”
• Solid outline for a fixed price, dashed for an estimate, so the meaning never rests on colour. Estimates are compared at the top of their range.
• The main button at the bottom, with what it will do written above it. Book, ask or hand over, with the reason.
• Booking without asking is a rule the rider sets: partner rides only, under a price and time limit, with 5 seconds to undo.

No signal is normal in a station, so prices say how old they are and booking waits. Handing over goes through the phone’s own “Open in” prompt with the trip filled in. Once the phone is back in a pocket, the pickup time sits on the lock screen.

A small design system, built from tokens

• Three tiers of tokens, in the Design Tokens Community Group format. Components only use the theme tier, so dark mode never touches a component.
• A palette from the kerb. Kerb yellow sits almost opposite cool asphalt grey on the colour wheel, so one accent marks the pick. It is too pale for white, so it only sits on dark asphalt.
• Atkinson Hyperlegible Next, made by the Braille Institute for low vision, with fixed-width numbers so prices do not jump.

How it would make money without bending the pick

Partners would pay a fee for each ride booked through Kerbside. That creates a pull towards favouring partners, so the pick is worked out on price and time first. Partner status only decides whether Kerbside books or hands over. There are no “2 cars left” or “most people choose this” prompts. Kerbside earns more only by signing up more partners.

How I would know it works

What I got wrong first

• It booked everything. The law and the providers’ terms cut that back to partner rides only.
• It used sliders. On a phone held in one hand, a slider is hard to set exactly. Steppers replaced them.
• It was a web page. It showed how the pick worked, but nothing about using it on a phone. I rebuilt it as an app with its own design system.
• The button hid the reason. On a real phone screen, the pinned button covered the line explaining the pick. A shorter map and a smaller header fixed it.
• It opened on the wrong default. It now opens on “cheapest”, the question the rider has just been asking.

What the sources could not tell me

The research shows what the market and the law allow. It does not show how riders choose. The first test I would run is at a station: five riders book the same trip by hand and with Kerbside. I would time both and ask whether they would have overruled the pick. The legal reading is my own, and I would check it with a licensing specialist before building.

If you see a flaw in how I read the sources, tell me.

Sources were checked against each organisation’s own page in October 2026. Provider names, prices and drivers in the prototype are made up.