Leveraging the Power of Visual Search
Project
Search cars by image
Role
UX Design
UI Design
Team
UI/UX Designer
Android Developers
PM
where
AutoScout24
year
2021
Find Your Dream Car with a Snap
Imagine being able to find your dream car with just a snap of a photo. Searching for the perfect vehicle can be time-consuming and overwhelming. To revolutionize the way our users discover their ideal cars, we introduced the feature: “Search a Car by Image.” Available exclusively in the Android app, this cutting-edge technology harnesses the power of visual search, similar to Google Lens, to provide a seamless and intuitive car browsing experience.
User need
From our user and market research we know that deciding on which car to buy is an offline and online experience. People are moved by their needs, the influence of other people’s experiences, by what they see on the street and in the ads. They get inspired and then visit the online car marketplaces spending countless hours scrolling through classified car listings, inputting specific details, and sifting through countless search results.
We also have identified that a significant percentage of our users are falling into the category of inexperienced users that need guidance in their journey.
The goal of the feature is to support that type of users by making less hectic the first step of the search phase in the car browsing experience.
Users, while browsing the AutoScout24 Android app, can stumble upon the “Search a Car by Image” feature in the homefeed or in the search page.
Users can select to either take a photo or upload a photo.
Whether they’re captivated by a car they spot on the street, a magazine ad, or even a friend’s photo.
We also provide a disclaimer to avoid uploading content with personal details or faces as well as a reassuring message that the photo will be deleted and not stored.
There is the possibility to crop or rotate the uploaded image.
Within seconds, our advanced image recognition technology analyzes the photo and presents the matching make and model and a curated selection of vehicles that closely match the desired make and model.
If the process encounters an error, users have the option to retry and attempt again.
Future enhancement
As the next step, we aimed to utilize user feedback to enhance the data model training. To achieve this, I explored the concept of incorporating a satisfaction metric to evaluate the accuracy of the search results. Users would have the option to provide a thumbs up or thumbs down rating. In the event of a thumbs down rating, an open field would be presented to gather qualitative feedback from the user. Additionally, we offered users the opportunity to retry the search by uploading another photo, ensuring they had multiple attempts to find the desired results.
This step was not implemented as the adoption of the feature was very low.
What my team says about me
Designed & built by kerry.gr