A new Pinterest Inc. feature aims to eliminate the need to ask questions such as: “Where did you get that shirt?”
Taking advantage of the rise in mobile photography, Pinterest says its image-discovery app in coming months will let users point their smartphone cameras at something, and the app instantly will search within its vast online showroom of 75 billion images for a visually similar match of the items in the photo.
The end goal for the six-year-old San Francisco startup is to make it easier for users to buy what they see in the real world with links to retail sites.
If successful, Pinterest’s camera-search technology could usher in a new kind of consumer shopping behavior. It also could help transform Pinterest, which makes money mainly from advertising, into an e-commerce destination that would help support its $11 billion valuation.
Pinterest first must prove it can persuade consumers to embrace a technology that for years has been riddled with slow and inconsistent results. The technology has flummoxed companies such as Amazon.comInc., Alphabet Inc.’s Google Inc. and a range of startups.
“Is it making an impact on retail? Not yet,” said Marshal Cohen, chief retail industry analyst for research firm NPD Group. “But the one thing we’ve learned over the last decade is when something is good with technology, and it hits successfully, it comes very quickly.”
Pinterest, whose site encourages amateur interior decorators and fashion mavens to hunt for inspiration, last week offered a sneak peek of what it would look like when people search for a baby stroller they see on the sidewalk, for example.
After a user takes a photo, Pinterest will search through its database and return images that are visually similar. Users can tap on dots that appear over the objects to get recommendations of similar items.
“What we’ve built…is the capability to use your phone so that when you’re in that moment—that magical moment of inspiration—you can find out where to buy that T-shirt,” said Pinterest President Tim Kendall at a June 28th event at the company’s San Francisco headquarters.
Retailers advertising on Pinterest now sell about 10 million items such as decorative pillows and hiking boots. While Pinterest doesn’t currently earn revenue from these transactions, it hopes to entice retailers to increase their advertising budgets.
Last year, Pinterest generated about $100 million in revenue by showing ads to its roughly 55 million active monthly U.S. users, according to a person familiar with the matter.
Pinterest isn’t the first to attempt to bridge the physical world with e-commerce. Amazon a few years ago built a stand-alone app called Flow designed to identify book covers and millions of products, but it has since incorporated the technology, now called Firefly, into its main Amazon app.
Google has been working for years on image-search technology, though its e-commerce ambitions are less clear. In 2009, it launched Google Goggles, which could identify and fetch information about a particular object or place by taking a photo, but the app failed to resonate with mainstream users.
Many startups also have jumped into the fray, including Mode.ai, a chatbot available on Facebook Messenger that helps users find outfits to buy after they snap photos of clothes.
Another company, Toronto-based Slyce Inc., incorporates its visual-search technology into apps from other retailers such as Neiman Marcus Group Ltd., Nordstrom Inc. andHome Depot Inc.
Still, photo-recognition shopping has yet to broadly take off for reasons ranging from technological limitations to lack of consumer awareness.
The number of monthly scans using Slyce’s technology can range from as few as 30 to more than 100,000, depending partly on the size of the retailer, said the company. Slyce’s own two visual-shopping apps, Pounce and Craves, have less than 5.5 million users each.
Neiman Marcus, which has used Slyce’s technology for two years, said it has been a “viable channel” to help consumers look for certain items, said Catherine Davis, vice president of marketing at the retailer. Still, she said, customers continue to use traditional search and navigation methods rather than photo search.
Ensuring accurate search results is a steep hurdle. Ted Mann, president of Slyce, said consumers don’t come back when the accuracy rate drops below 90%. “If you’ve tried any of these apps and had a bad experience you’re likely not to try it again,” he said.
Slyce’s technology, he said, averages a 92% accuracy rate, the criteria of which varies by client. For instance, identifying the correct brand logo is imperative for a shoe retailer, while an apparel company might give more weight to surfacing a gender’s clothing category.
Many analysts, retailers and competitors note that Pinterest’s scale could help the technology break through the mainstream. “I think a rising tide lifts all boats here,” Mr. Mann said.
Pinterest said it is still months away from launch, as it works out kinks and increases the accuracy of results. In one demo following the June 28th event, a photo of a black flat shoe delivered images of black flats with and without bows as well as pumps and flats in different colors, but not the actual shoe in the photo.
Source: WSJ (by: Yoree Koh).
