Shopping for clothes is a damn near impossible task. There’s a reason why many women make it a whole-day affair. Getting caught up in the idea of a specific thing that you want, down to the cut and style, gives you a narrow lens to view from but also limited options.
Shopping online doesn’t prove to be much better: text search can only get you so far, and filtering can take time if you’re searching across many brands.
Donde Fashion, a mobile app for iOS released today, cuts through all that nonsense by turning the search for women’s fashion into an image-based experience. Users simply tap on the clothing item or accessory they are interested in, and continue narrowing down through a series of taps to determine ideal cut, color and pattern.
“We’re the first ones classifying millions of items around the world based on visual features,” said Liat Zakay, founder and CEO of Donde Fashion.
Visual search is a hot thing among ecommerce and design-focused applications. The deep learning algorithms at Pinterest and the hyper-tailored search methods available on Yahoo-acquired Polyvore both immediately spring to mind. But neither of those allow you to skip the typing process entirely — something that Donde Fashion readily provides.
Once you’ve found an item you’d like to purchase, Donde Fashion dumps the item directly in the affiliate’s shopping cart and sends you to check out. At launch, the company is working with more than 6,000 brands and retailers, including big names like Bloomingdale’s as well as fast fashion stores like Zara.
Users can also ask a friend for an opinion or request the app to send a push notification for when it goes on sale. They can also filter to only see certain brands or leave out others.
Donde Fashion accomplish this experience through a machine learning algorithm that Zakay said the team developed in-house. Combining both text and machine vision, Donde’s algorithm is designed to tag fashion products based on how people understand an item’s different features. So, the app can discern in its database between a round-neck and a v-neck, and serve results accordingly.
Zakay explained further how that information gets applied to the retailers on the app:
To make our indexing accurate, we crawled the Web and extracted data from all major brands and retailers, building a semantic network of products. The semantic network is calibrated using machine learning techniques reflecting relations between the products and their components and encapsulates the similarities and differences between products and their composure allowing the search to be focused and contextual.
Donde Fashion is a useful tool if you know exactly what you already want. If you’ve been envisioning a blue-striped tote bag or a pair of slouchy harem pants, then Donde will get you to your destination and ostensibly push you to the shopping cart in just a few taps.
That is a powerful force, and something to consider when, according to display ad company Criteo, mobile accounted for 29 percent of US and 34 percent of global ecommerce. As more people turn to their phones to shop online, the application (or applications) that figures out the magic formula to prevent shopping cart abandonment will make the big bucks.
For all of its posturing, however, Donde Fashion is far from perfect. For example, when narrowing down to a gold shoe, the results often turned up a black shoe with gold trim. When looking at green pants, a pair of turquoise pants are found among the mix and can certainly read to the eye as blue. These are no doubt kinks in the app’s algorithm that will improve over time, but those expecting a 100 percent magical experience might be disappointed.
All in all, Donde Fashion reduces hours of casual browsing into a brisk few minutes of tapping, which is a potent and intriguing ability.
➤ Donde Fashion [iOS]