Allegro develops visual search tool
- Allegro, the go-to commerce platform for Polish consumers, develops a new feature on its mobile app.
- Visual search uses machine learning to match the product featured on a picture made by the user and finds similar items for sale on allegro.pl.
- The tool currently supports over 100M offers in the following categories: fashion (including shoes and sport fashion), health, beauty, grocery, media, books, comics, games and music, and also a major part of home and garden.
Each month Allegro attracts visits from an average of 21 million users. Visual search engine will make shopping on the platform even easier now. So how does it work? The new feature is built into the Allegro app (available for both Android and iOS). Users can take a picture of a chosen piece of clothing, bottled water or book, upload it and the Allegro tool will use machine learning to search through articles available on the platform and show most relevant offers. What's more, the user doesn't even need to go to the store and take the picture of the product on his / her own. Visual search recognizes pictures of products presented e.g. in magazines or screenshots made on one’s phone.
Visual search required a completely new searching architecture and machine learning technology. The simplicity of the customer experience belies the complexity behind it. We expect that the visual search tool will be able to provide shoppers with multiple options of similar products - comments Wojciech Bogdan, Chief Data Officer at Allegro.
For years Allegro team has been developing different functionalities that help to manage the sales as quick delivery services or convenient payment methods. Now it's entering a completely new era with a tool that will make finding and buying dreamed-of items even easier.
We always rely on modern technologies. Besides, our main goal from the very beginning was to make online shopping easier for Poles. Thus, it was kind of a natural step for us to support hardly descriptive categories with a cutting-edge solution. We can easily identify a song that’s playing on a radio or a plant growing in our neighborhood. Why not do the same with a dress or a grocery product that we saw on the street or in a store? Visual search was inspired by a situation that we all have been in – who hadn’t come across a cool t-shirt or shoes somewhere in public and then spent a lot of time trying to find it – adds Wojciech Bogdan.