Amazon failed Ted Mann, CEO of Slyce, a visual search firm, when it came to the cheesy sweater — the one with, in his words, “little teddy bears all over it.” He tried in vain to search via text for that product, a wedding anniversary gift for pals who apparently have a good sense of humor. How do you translate such an item into a workable search term that will bring up the proper item?
Even Mark Twain might have problems with that.
However, Mann wasn’t about to give up — cheesy sweaters make such memorable presents, after all. He uploaded an image of the apparel item on Pinterest, where the first result “was not an exact match, but spot on.” Victory achieved — score one for visual search.
There is a battle of senses taking place in commerce these days, one in which voice and visual searches are on the rise. Voice-assisted retail has commanded most of the spotlight in recent months, a reasonable development given the sales of voice-enabled devices, especially during the 2018 holiday shopping season.
Visual Search Gains
As Mann told Karen Webster during the latest installment of the PYMNTS Matchmakers interview series, the time has come to recognize the strides visual search has made — and how the retail and marketing technology could help merchants gain and keep more customers.
Instagram and other social media platforms, along with retailers (Slyce serves about 60 of them), have integrated visual search, and the numbers promise to keep rising. In this era of providing consumer experiences instead of mere transactions, visual search offers something extra and more efficient than many other retail tools, according to Mann.
“It can solve real-world problems, like building shopping lists,” he said, noting how tedious it can be to do so on mobile devices without technology that eases the hassles — frustrations with which many consumers can relate. Visual search works with gift and baby registries as well. It can also combine with augmented reality, and, when deployed in certain ways, offer a full product range and experience to consumers.
Take one example from apparel retail (this one without any sweaters). A consumer might be looking for, say, a black skirt, but is likely (if not more) looking for an outfit that has a black skirt.
The consumer is seeking a look, not just a single item. If deployed to reflect that consumer desire, a visual search tool can not only enable the shopper — or the store associate helping the customer — to find that skirt, but instantly call up product suggestions that would round out the outfit. In fact, that is what’s happening now in certain retail stores using visual search, Mann told Webster, which puts a new importance on the relationship between the consumer and store associate.
One of the surprises Mann has experienced from his company’s visual search technology deployments — the company has been around since 2012 — is that the bulk of use is coming from inside stores, and from store associates. When it comes to apparel and outfit suggestions, that serves to “give consumers a lot more confidence” about the overall retail experience, as well as drive conversions. The same principle can also apply to other products, including furniture and home décor.
That’s not to say visual search doesn’t face significant challenges.
Like so much else in digital retail, success comes down to data. In Mann’s telling, pretty much anyone can get visual search matches in the 40 percent, 50 percent or even 70 percent range. However, retailers need rates higher than that — 95 percent or more. That requires “custom training data” that involves “scrapping images off the web, and brute force collection of product photos,” with data being run through an artificial intelligence engine. The principle of what Mann called “garbage in, garbage out” rules the world of visual search.
Even with a high rate of visual search matches, there is always the chance that a consumer — perhaps one scanning an image from a printed catalog — is searching for a product that is out of stock or no longer carried by a specific merchant. A good way to lose a potential sale is to make sure that consumer lands on an empty product page.
Rather than that, Mann said, “you can use visual search to show digitally similar alternatives,” such as a blue dress with comparable attributes to the blue dress the consumer was originally seeking. “Sometimes, you have to go a couple steps beyond” to make sure visual search functions as a solid tool to win over consumers and spark transactions. The key there is to manage expectations — and promote such a visual search feature if it can really be offered, he added.
What about mixing or marrying the visual and voice searches in the coming years? What are the prospects for that?
“In some cases, they can play off each other,” Mann explained, telling Webster that he imagined the two — along with basic, traditional text — serving as search bar options for an increasing number of consumer shopping expeditions, with each offering its own strengths and weaknesses. “It all depends on what you are looking for,” he said.
For now, visual search continues to make strides, while also giving more power and use to store associates.