Facebook has unveiled a new e-commerce platform and several new shopping features in an effort to make its social network more relevant for business.
Facebook CEO Mark Zuckerberg announced the changes Tuesday in a live video broadcast on the social network in which he positioned the new shopping service as a way to boost small businesses, especially those struggling during the coronavirus pandemic.
“Our goal is to make shopping seamless and empower anyone from a small-business owner to a global brand to use our apps to connect with customers,” Zuckerberg said.
The new e-commerce platform, called Facebook Shops, is a further step in Facebook’s efforts to integrate its original social networking site with photo-sharing site Instagram and messaging services WhatsApp and Messenger, which it also owns.
It’s also the company’s latest move to generate revenue from Instagram, which it purchased for $1 billion in 2012, and WhatsApp, which it bought for $19 billion in 2014. Although the shops are free for businesses to set up, Zuckerberg said he thought the improved commerce offerings would lead to increased advertising across Facebook’s platforms, which is the company’s primary revenue source.
He said the new shopping features were “the biggest step we’ve taken yet to enable commerce across our apps.” While in the past, many companies had used Facebook and Instagram to build brand awareness, Facebook has been eager to ensure users stay within its ecosystem to actually make transactions.
Facebook Shops will allow any business to set up a unified Facebook and Instagram virtual storefront, put its catalog online, and take orders through either a link to its own website or, for companies in the U.S. that use Facebook’s “checkout” feature, directly on Facebook’s platform.
Customers will be able to communicate with the retailer through WhatsApp, Instagram Direct, or Messenger. In the future, Facebook said, customers would be able to make purchases directly through these messaging services too.
The company also debuted new e-commerce features for Instagram. One called Instagram Shop will enable users to browse for items they may want to buy through the app’s Explore feature, browsing by brand or item type, and then make purchases directly within the app.
Another feature will allow businesses to tag products in their Facebook Shops catalog and then do a broadcast on Facebook Live or Instagram Live that will enable customers to purchase those items while watching. In essence, it takes a concept pioneered in the television era by the Home Shopping Network and updates it for the mobile age.
Finally, the company said it would test a way for customers to connect company loyalty programs to their Facebook profiles and then enable businesses to manage those loyalty programs through their Facebook Shops accounts.
Facebook said companies that provide tech to help businesses sell online such as Shopify, BigCommerce, Woo, ChannelAdvisor, Cred Commerce, Cafe24, Tienda Nube, and Feedonomics were partnering with it on these new shopping features and would be providing tools to help small businesses set up on Facebook Shops.
In a separate announcement and blog post Tuesday, Facebook touted several artificial intelligence advances that it says will underpin the new shopping features. Manohar Paluri, a director of Facebook AI, said the vision is to give small businesses tools to create the same sort of online shopping experiences—complete with 360-degree views of products and detailed catalog information—that previously only major retail brands could afford to deliver.
Among these innovations is a computer vision system capable of recognizing almost any product—from a Gap sweatshirt to an IKEA sofa to a KIA Sportage—in almost any image.
The system, which Facebook calls GrokNet, not only identifies the product, it can determine many of its attributes, such as its color and style. The idea is that someone will be able to upload a photograph to Facebook’s Marketplace, or eventually Instagram, and automatically create links to places where you can buy the items in that photograph. Businesses will be able to create catalogs of items they can sell more rapidly simply by photographing their products. And, eventually, Facebook and Instagram users may be able to shop simply by taking photographs of items they are hoping to buy.
The system was trained on billions of images, both high-resolution and low-resolution, across seven different data sets, and it outputs 83 different classifications for each object.
The system was fine-tuned to perform particularly well for fashion items, furniture, and automobiles, Sean Bell, the computer vision researcher who developed the software, said. Bell had previously founded a San Francisco company called GrokStyle that was creating this software and which Facebook acquired in February 2019 for an undisclosed amount.
Facebook said in a blog post that GrokNet is currently powering new Marketplace features that will let consumers shop by image and that it is testing using the system to automatically tag items on Facebook Pages.
In what it calls a world first, Facebook also said it had created a system that can accurately identify partly blocked objects in images—such as a T-shirt underneath a jacket. The system isn’t always perfect—it can’t know for certain in some cases, for instance, the sleeve length of a shirt underneath a sweater. But it performs better than any prior system at this difficult task, Tamara Berg, another Facebook researcher who worked on the computer vision project, said.
The company also created a different A.I. algorithm, called Rotating View, that will enable anyone to create three-dimensional and panoramic views from just a few photographs. The software leverages advances made in the past two years by Facebook’s own researchers and those at companies like chipmaker Nvidia and DeepMind, the A.I. company owned by Google parent Alphabet, that can render three-dimensional models from two-dimensional images.
In blog posts related to the A.I. developments, Facebook sketched a future in which an intelligent virtual assistant helps users shop, suggesting items that suit their personal tastes and fashion sense, or that fit in well in with their existing home decor.
Berg said the company is in the process of prototyping a kind of “virtual fashion stylist” along these lines—one that would eventually be able to recommend what items in a person’s wardrobe would work well together or help a person build outfits from different items. She said it has long been her own ambition to create a virtual fashion assistant modeled on the character of Cher, played by Alicia Silverstone in the 1995 film Clueless.
