Telgani came to us with a product that worked and a funnel that leaked. Cars were getting found, carts were getting filled, and then, at the payment step, a fifth of customers quietly left. The brief was narrow on purpose: do not redesign the brand, fix the flow.
The challenge
The old checkout was built in a hurry during a growth spurt. Seven screens, three of them asking for information the backend already had. On mid-range Android phones, the payment page took over four seconds to become usable, and the session data showed exactly where patience ran out.
Speed was not a nice-to-have here. In rental, the customer is often standing in a parking lot when they book.
The approach
We ran design and engineering as one track. While research mapped the seven screens against what the API actually required, engineering built a prototype checkout against production data in week two. Every design decision after that was tested on a throttled connection before it earned a place in the flow.
The seven screens became three. Saved payment methods moved to the first screen instead of the last. Every field that could be inferred was inferred.
The version we shipped is the boring one. That is why it converts.
What shipped
A three-step checkout rebuilt on the existing design language, a payment page that becomes interactive in under a second on 4G, and a component library the Telgani team now extends without us. We stayed on for a quarter of measurement, tuning the flow against real traffic before calling it done.
