eCommerce is an industry at an exciting time. It is a highly competitive, yet still burgeoning space. The market is growing steadily year after year. Societal shifts in behavior have contributed to the gradual acceptance, use and reliance on eCommerce. Global Internet penetration relentlessly increases over time. Global shipping costs continue to steadily decline. Consumers are learning to trust online merchants. These secular trends, though promising invite intense hyper-intense competition. Amazon has established itself as “The Everything Store”. Small, niche players pop up seemingly every day to specialize in and capitalize on the latest consumer trends. Retail giants invest heavily in their online presence. Acquiring customers through digital channels continues to be more and more expensive. How then can an Internet age dinosaur compete with these new age eCommerce brands? eBay’s answer is to invest in Artificial Intelligence and Machine Learning capabilities to improve the marketplace. eBay’s CEO, Devin Wenig made clear his belief that investing in machine learning capabilities is where the industry is headed, “I believe that commerce, in particular, will be the focus of some of the most immediate and exciting applications of AI…The holy grail for commerce has always been deducing the buying intent of consumers — when they walk into stores, when they browse online, when they order on apps”. Wenig believes in the power of Machine Learning to transform commerce itself. More specifically, his strategy is to use machine learning to understand buyer intent and ultimately improve the buyer experience.
eBay’s machine learning strategy is focused on the buyer experience in the near term. When a marketplace has millions of listings, it can be difficult for consumers to find the right item, especially when they are searching across several different criteria (e.g. price, quality, seller rating, age etc.). Wenig and the eBay team believe that machine learning can eventually become powerful enough to “render the search box redundant”. Improving the buyer experience with customized recommendations is an important point of differentiation for the marketplace. While personalization has existed in some form for quite a while, machine learning can use personal preferences and generalize to other purchases that you may make. For example, if you buy a toothbrush, it is not hard to guess that you might also want toothpaste. However, machine learning can help eBay to take a person’s toothbrush preferences and generalize to what type of scarf they would be inclined to purchase. Custom recommendations work well when buyers are using the platform for discovery or are just browsing through listings.
In terms of timeframe, eBay has made clear that they are investing for the long haul. Many of the functions that they are working on today to improve through machine learning could have been outsourced to a third party. While faster and cheaper to utilize experts who specialize in machine learning, the company would not have retained the knowledge and skills after the projects were over. So, eBay has made explicit its strategy of building the skill set in house and hiring its own team dedicated to machine learning. In accordance with this strategy, eBay acquired ExpertMaker in May 2016 to add skill and brainpower to its machine learning team. Albeit slower and more expensive, the choice to in-source machine learning capabilities demonstrates eBay’s long term commitment to this technology.
Despite eBay’s strong start to machine learning applications for the buyer experience, there is still plenty of unexplored territory in which to experiment. The improvements to the marketplace mentioned previously are very focused on the buyer experience once on the site. When eBay is ready, I recommend that they look at using machine learning to understand how to support their sellers and grow the community. They could look at what distinguishes power sellers from those who are less successful. Or they could use machine learning to understand how to source new sellers from the plethora of small business owners out there already. Alternatively, eBay could use machine learning to refine and develop the efficacy of digital buyer acquisition channels. Using machine learning to deeply understand the LTV of one customer profile vs another would help the company spend money bidding on the right consumers for the marketplace, thus reducing wasted ad dollars and increasing efficiency.
Machine learning and eBay’s business model seem like a perfect fit. But are they? The most recent evidence that the company gives pegs the incremental GMV impact at about $1B each quarter. $1B is a huge number of sales, but only represents less than 5% of eBay’s total GMV. Based on the company’s own excitement around the technology, should we expect this number to be larger? What is a reasonable target? Only time will tell. For now, enjoy the search box – it may be gone before you know it.