For e-commerce giants like Amazon and eBay, personalization is the name of the game. We live in an age where Internet pages are increasingly customized to individual users, all in the name of maximizing potential advertising or product revenue. EBay has been one of the companies at the forefront of this practice; back in late 2012, eBay launched a major redesign centered around Pinterest-like feeds. These feeds, which push content based on eBay search histories and browsing habits, now dominate eBay’s homepage.
Behind eBay’s customized homepage, app content, and landing pages lies the larger tale of a company transitioning from a traditional auction site to a middle person for brick-and-mortar companies in the digital world. This requires a staff of researchers--primarily data scientists and machine learning researchers, but also from the social science sphere--who can wed quantitative and qualitative research traditionally found in academia to the world of e-commerce. Elizabeth Churchill, eBay’s director of human-computer interaction and a veteran of Yahoo and Xerox PARC, has a unique mandate: Getting data scientists inside the heads of different kinds of eBay customers.
Churchill, whose academic background is in experimental psychology and knowledge-based systems, supervises a staff of three researchers and six interns. The interns come from PhD programs in STEM disciplines, ethnography, and communications. A large part of her team’s role is understanding different varieties of customers who use the service--and wedding eBay’s internal data to sociological work to figure out how to tweak the service’s appearance and behavior for different users.
“One of things we have is different forms in data,” Churchill told Co.Labs. “Not just behavior data, but transaction data, a lot of data from interviews, surveys, and ethnographic work. We really do a lot of 'experience mining' to look at what the data doesn’t tell us, so we can find the questions we want answered. We drive ethnographic process by looking at data that exists in scale to sample the right people to talk to to find people to speak about what they do off eBay in their general life experiences, as well as what’s on eBay.”
In real-life terms, that means Churchill and her team research specific subgroups of site users--ranging from new eBay users to purchasers of vintage clothing to purchasers of low-cost bulk items to different kinds of resellers--to find trends in the items they purchase or the way they navigate the site. That information is combined with ethnographic research to help eBay’s team tweak the site experience users have.
By email, Churchill added that “We use data science techniques to classify activity types, use ethnographic research to dig deeper into the motivations behind these behaviors and to classify user types beyond the classic marketing categories, develop behavioral ‘traits’ that correspond to different shopping orientations and activities, and use our eBay data in the small and large to more deeply investigate onsite activities and develop predictive models.”
This means more than just the items that show up on the homepage or what auctions are most prominently featured in the mobile app. Demographic and site use data about eBay users is used, Churchill says, not for homepage design but for notifications. The emails users receive from eBay are shaped considerably by demographic information. “Demographic data is used most effectively for notifications and marketing campaigns, rather than algorithmic recommendations,” she added. A big part of this is using data about a user to figure out the sweet spot that will get them to visit eBay more often without annoying them.
For eBay, which faces competition from Etsy, Amazon, Target, and a host of smaller competitors, combining data science with the social sciences makes sense. Analytics help them understand user behavior, and anthropological and sociological fieldwork allows them to leverage behavior their competitors might not perceive.
It’s an unorthodox job: In our conversation, Churchill explained that eBay is able to parse data points relating to purchasing of vintage band T-shirts by Japanese consumers (which is apparently a very valuable market segment) and identify blips like Sears’ flying jackets that are popular with small market segments. If a particular band’s T-shirts sell more among consumers in a specific geographic area, that data is then leveraged for the future.
Churchill added that, for her team, empathy being able to place themselves in the shoes of users who use eBay in different ways is the most important aspect. “I build multidisciplinary groups because understanding users’ emotional journeys means a mix of computer scientists, front end developers, game designers who look at gamification elements, and social scientists for ethnographic fieldwork.” In the world of commerce, data science needs all the data points it can garner to be useful. For researchers, this means embracing the social sciences as well.
[Image: Leon7 via Wikimedia Commons]