How does an enormous platform like eBay — which, as of the end of 2018, had an estimated 179 million active users and over tens of million sellers across 190 markets — use AI to deliver great buyer and seller experiences? Chiefly by supplying structure to the more than 1.2 billion listings its users interact with every day.
In a blog post published this morning, eBay highlighted the many ways it’s using machine learning to “enhance experiences” and inspire “economic empowerment” throughout its ever-changing marketplace. The company’s more than 20 years’ worth of data provides the starting point — the kindling that fuels the computer vision, machine translation, natural language processing (NLP), personalization, and recommendation algorithms at the heart of eBay’s purchasing and discovery funnels.
“Artificial intelligence touches every experience within eBay,” writes Sanjeev Katariya, VP and chief architect of AI and platforms at eBay. “It is woven into all aspects of the eBay marketplace, anticipating the needs and wants of buyers and sellers, inspiring shoppers on the hunt for something special, empowering entrepreneurs looking to grow their business, and making the platform more accessible to everyone. Over the last few years, we’ve worked to make our shopping experience more relevant to every shopper’s needs, whether customers are looking to discover, be inspired, find a specific item, or browse.”
A tangible outgrowth of its work arrived early last year. Interests is a module on the eBay homepage that susses out shoppers’ passions, hobbies, and styles and uses the insights it gleans to tailor product recommendations. (According to eBay, the more than 2.6 million people across five countries who use Interests average 18 recommendations on eBay’s smartphone app and 10 on the web.) Enhancements to search coincided with the rollout of Interests. eBay says it leverages AI to better understand areas such as units of measurement and natural language, so that a search for “watches,” for instance, will surface a range of bands, prices, materials, and movement types.
Computer vision has come to play a larger role in the shopping flow, according to eBay. Late 2016 saw the launch of Image Search, which lets users find items by snapping photos of real-world objects, and Find it on eBay, which surfaces listings of products similar to those on virtually any webpage. “All of these improvements use AI to power a sharper experience for our buyers,” eBay says. “We have a better understanding of who our shoppers are, what they might be looking for, and how they want to shop.”
It’s not just shoppers who’ve benefited from eBay’s infusion of machine learning. The company says that sophisticated NLP algorithms have enabled it to accelerate listing times by refining searches to uncover best-matched items. And it claims its work in translation has dramatically reduced processing time for international customers. (One study found that English-to-Spanish AI-powered translation on eBay increased exports by 17.5 percent.) More broadly, eBay says it has worked to improve guest onboarding by halving the number of steps required to purchase any given item. And it says its new product-based shopping models not only help most buyers find products more quickly, but also opt in “millions” of listings it believes will drive search engine optimization and social traffic going forward.
“We used open source technology to build an in-house AI platform that reaches across eBay to enable collaboration and training of our AI models at scale. It allows our data scientists and engineers to experiment, build products and experiences for customers, and leverage AI at scale,” Katariya writes. “While we’ve elevated the experience for our customers with AI for more than a decade, the recent advances in deep learning have … enabled us to scale to larger and more complex data sets [consisting of] billions of data points and [to approach] human-level competency [with deep learning models].”
AI is only one part of the $4.88 trillion ecommerce puzzle, of course. To that end, eBay has in recent years introduced a suite of APIs for Image Search, machine translation, and its marketplace, which it provides to developers at partner sites at no charge. Separately, it has applied specialized domain science to issues like shipping times, risk, and trust and investigated ways to improve AI models in deployment and reduce bias.
It’s evidently doing something right. In the fourth quarter of 2018, eBay recorded net income of $763 million, beating year-ago losses of $2.6 billion. And last year Tom Pinckney, eBay’s VP of applied research, told VentureBeat in an interview that AI was driving “north of $1 billion per quarter” in incremental marketplace sales.
“Our approach allows us to determine what methods work and at what time so that we can create these meaningful experiences for our buyers and sellers,” Katariya writes. “eBay is a robust marketplace, and AI is at the center of that journey, positioning the platform as a leading intelligent marketplace, and we’ll never stop aiming for the best customer experience. Looking ahead, we will continue down the path of innovation through eBay’s AI managed marketplace, knowing that the scope of what’s possible with AI expands every day.”
VentureBeatVentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact. Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform
- networking features, and more