Car-photo verification flow

Context

In Mercado Libre's vehicle financing, fraudsters were closing contracts on cars that didn't match reality — about four cases every month. The MVP asks the person at the dealership to photograph the actual car, turning a back-office fraud problem into a two-minute product moment that had to work for people who are not photographers.

The problem

Fraud in vehicle financing was costing cases every month. The existing flow had poor guidance, leading to retakes, support tickets, and frustrated users who could not self-correct their photos.

Process & decisions

01

Making fraud logistically impossible

The flow demands presence: you must be at the store, near the car, and the photo has a two-minute window once you start. Location and time constraints quietly remove the room fraud needs to operate — before any human validation happens.

02

Guidance that gets the photo right the first time

Instead of a bare camera button, the flow teaches by example: a reference photo of exactly the expected shot — car from behind, plate visible, store around it — with a short checklist for framing and lighting. The camera opens with that mental model already set. 85% of users nailed the photo on the first try.

03

Self-service when the photo fails

Rejection isn't a dead end. The contracting timeline flags exactly which step needs attention — "fix the vehicle photo" — and validation expectations are honest: up to 10 minutes, business hours stated. 75% of users who had to retake the photo solved it alone, without ever contacting support.

The solution

I redesigned the photo verification flow with clear instructions, real-time feedback, visual examples, and progressive guidance so users could take the right photo on the first try and fix it alone when needed.

Demo

Watch with sound

The photo-verification flow in motion: guidance, location check, capture and validation.

Results

Screens