Complaisant brand managers beware. There’s a new player in town and shoppers will like it a lot more than you do. Functionally, the new visual search tool lets the consumer shoot a picture and then scan a database of photos to find online matches.
None of the components of the SuperFish Visual Search Tool are entirely new, but combining them certainly is. In principle, if you can render a wireframe image to create a 3D object, you can de-render a 3D object to get a wireframe. Getting to 3D from a two-dimensional image has been around for a while, though other applications rely on human cognitive input.
Clever Artificial Intelligence
Without know the inner workings of the algorithm we can only speculate, but … Applications that rely on human cognitive input use our visual processing ability to identify edges, then use basic geo-semantic relations like the orthogonality, concentricity, and parallelism of human design to extract the image. To compensate for the lack of human input, presumably the visual search tool relies on the human desire for good photographic composition to identify the search target.
How It Works
The process begins when the consumer uploads a photograph from either disc or the Internet to the application. Next, the visual search tool de-renders the image to create a wireframe model of the image. It then queries the vast modern database of images to generate a ranked list of the best results. Finally, the application presents that list in the form of a visual online tombstone-style catalog. Consumers then drill down to explore possible matches for the product of interest.
Why That Matters to Marketers and Product Designers
Years spent building a brand can now be erased anywhere and at the blink of an eye. At one time cheap knock-offs of brand name products were at least hypothetically limited to street corner markets. With the advent of visual search, similar items will be at the shopper’s fingertips wherever she happens to be and whatever she wants to buy.
What You Should Learn in a Trial
Five minutes conducting a trial using the ‘Like That Décor’ app on the SuperFish.com site. Google for pictures of a common object like a ‘custom greenhouse.’ Save a copy of one of the photos to your hard drive. At the Like That Décor prompt, upload the photo. The app runs astonishingly fast. Within a few seconds Like That Décor returns pages of greenhouses for you to shop through. As you will see the app doesn’t return an exact match, even for the most commonly returned photos on Google. Instead, shoppers get a beautiful catalog listing of their competitors’ products, including prices, shipping costs, and other essential information.
Implications and What To Do About Them
Since the visual search tool returns a broad selection of images of items similar to those the consumer looked for, ‘Image rank’ will quickly take on importance similar to that of ‘Page rank,’ or keyword search. Just when you thought you had the Internet knocked, expect the emergence of an entirely new branch of search engine optimization. Looking ahead, that might include some or all of the following:
· Archive more images of each branded product
· Shoot images from more angles
· Try to predict the angles/views shoppers will actually see
· Archive some images taken ‘in the wild.’
· Incorporate easily machine-recognizable design elements in products
Thankfully revolutions don’t come along every day, but this is definitely one of them. Proactivity prospers and whether this helps your business or eventually wipes out your market share depends on what you do right now to prepare for the future.