A picture paints a thousand words. And so does a screenshot, especially when shoppers are gazing at the latest pair of basketball shoes gracing the feet of their favourite influencer.
This is the simplified premise of visual search, regarded by some as the evolution of traditional keyword searching, as it renders the world instantly searchable and infinitely shoppable, from a piece of jewellery on a celebrity’s finger to a sofa on the set of a TV chat show.
Words can be awkwardly phrased, misspelt and sometimes a user won’t even know how to describe what they’re looking for. But all those limitations are lifted when searching visually – just take a picture, feed it into a visual search engine and you’ll be directed to the online retailer that will get it to you in the shortest time and at the best price.
How visual search works
The technology works using a combination of computer vision, which enables mobiles, tablets and laptops to ‘see’ (as with Face ID or QR-code scanning); and image recognition, which allows the digital eye to ‘understand’ and categorise what it’s seeing. As with all machine learning, the understanding comes from training.
To find that pair of basketball shoes from your camera roll, engineers will have trained an image search engine on hundreds of thousands of images of basketball shoes in varying styles, sizes and colours. The AI engine will then study and learn from every pixel in every image so it can refine and expand its understanding over time.
“Leveraging visual search effectively involves a combination of technology integration, user-experience design and strategic marketing,” says Blake Morgan, author of The Customer of the Future.
“It is a threat to retailers that aren’t considering how visual the modern customer is. If they don’t build customer experiences with visual search, they stand to be left behind because research shows that 62% of millennials like the ability to search visually; they are visual learners and that’s how they prefer to shop as well.”
The opportunities in visual search
One of the greatest opportunities in visual search, according to Katarzyna Dorsey, founder and CEO of Yosh.AI, is that “it allows brands to respond to the way people want to shop in an inventive and impactful way and achieve significantly higher conversion rates.”
Take the type of shopper categorised as the spearfisher because they are laser-focused on buying one specific item.
“Visual search could be so attractive to this group because it reduces friction; no typing, no scrolling, just one button pushes and they have their item,” explains Dorsey.
Similarly, a visual search engine might direct online users to a website for their product even if it’s unavailable. That user might be tempted to purchase an alternative, especially if the website is sophisticated enough to immediately engage them in a recommendation carousel, showing similar or related items.
The technology can also inspire, for instance, an image of basketball shoes to return the image of an entire outfit. This might prompt the user to buy even more items than they originally intended.
Some products are already optimised to be visually searched, like basketball shoes, for example, which are already branded with a logo that will provide little room for error for a well-trained visual search engine. But products with more generic, or fewer distinctive, features are more open for interpretation from the AI, which could lead to the product that features in the user’s visual search being bypassed for a lookalike.
The difficulties of visual search
The visual search for a sofa might throw up several near-identical sofas, at different price points. Seen through an ecommerce lens, these look nearly identical and will draw the buyer’s attention away from the original and towards one of the substitutes.
This can be a problem for retailers that rely on more traditional brand interactions, but there are some measures they can take to increase the chances of their specific products being recognised by AI.
Crystal Carter, head of SEO communications at WIX, says that: “Retailers need to have a solid understanding of how visual searches are informed.” To identify products, AR relies on the ability to recognise “billions of entities, like logos, brand colours, landmarks, text and named objects.” These all need to be visible in a product’s online display if it is to be picked out by AI in a visual search.
“For instance, if you’re selling a teapot, make sure the online image shows the handle and spout, ideally in profile. If you have an in-store display, ensure logos are given high priority, any text is unobscured, and objects are well-lit and well-staged. Busy backgrounds are a no-go.”
Where retailers might struggle is imagining use cases where visual search can be leveraged to bring people in-store. After all, how do you bring the technology to bear when you are already surrounded by shelves crammed with physical products? The answer could be to purposefully empty those shelves.
Bringing visual search in-store
This is what footwear brand Eobuwie did when it wanted to create a unique in-store experience, hiding all its product in a stockroom loaded with more than 110,000 pairs of shoes, leaving a clean, future-fit retail space that contained just tables and tablets where customers can search for the shoes they want. Once chosen, its shoes are sent from the stockroom, via racking, to appear behind the cash desks to create the kind of retail theatre that customers can’t get online.
Visual search can move this concept on to prompt customers to walk in from the street to find a match in-store. Not only would this engage young consumers, who usually shop online, but it would also appeal to consumers who still wanted a more traditional retail experience. If the item isn’t in stock then the customer will already be in the store, making an alternative purchase a possibility.
What is clear is that regardless of whether visual search is viewed as a threat or an opportunity, the reaction to the technology needs to be similarly proactive. It may be the next step in digital Darwinism that retailers must actively prepare and optimise for.