Overstock Is Turning to Machine Learning to Personalize the Path to Purchase for Individual Shoppers

brought to you by WBR Insights

Internet retailer Overstock.com wants to provide more intelligent and more immediate personalization for its online customers. Though personalization has been an important facet of the company's website for many years, what started as a journey to offer individual shoppers more accurate product recommendations has now turned into a mission to tailor all content - from page layouts to emails - to each visitor's individual interests.

"We want to improve the customer experience, which started our journey toward personalization," said Amit Goyal, who served as Senior Vice President of Product and Engineering at Overstock until September 2018.

Overstock knows that individual visitors to its website all have individual needs. And this holds true for all websites, of course. A first-time visitor, for instance, will have completely different requirements than a third-time visitor - one who's already made a purchase, is familiar with the products, and is coming back for more.

The goal for Overstock, then, has been to move away from static web pages that display identically to each and every visitor, and towards a more dynamic approach that presents personalized content in real-time, based on an individual's history and behavior.

"When you enter Overstock.com, no matter what page, if we know what you're looking for, that page should personalize to you," said Overstock's Chief Marketing Officer John Paul Knab. "It should not be the same type of page every other consumer is seeing."

The Asset Data Platform

There are two major systems that Overstock has in place in order to achieve the levels of personalization it believes will be winning for customers. The first is an in-house built "asset data platform", which stores all of the retailer's 25 million separate pieces of content - including product images, marketing materials, offer overlays, and videos - ready to be delivered to visitors.

Using the asset data platform, Overstock is able to, for instance, present different product recommendations to different shoppers, based on their search history, in-session behavior, and previous purchases. Goyal gave the example of a consumer who's previously searched for and purchased a couch on her last visit, and is now on the site searching for ottomans. The platform is able to adjust and recommend ottomans rather than couches as she journeys around the site and changes her search behavior.

Based on this information, Overstock's asset data platform surmises that the shopper is likely redesigning her living room, and so populates the pages she visits with relevant products in the shopper's preferred color and style and also suggests relevant articles and other content.

"We believe that if we can personalize a consumer's page to her style, with the colors she likes, with items within her budget, Overstock will be top of mind for her the next time she's looking to make a purchase," said Knab.

The Customer Data Platform

The second major system is what Knab calls the "customer data platform", which Overstock created in partnership with technology provider mParticle. This system collects real-time data about customers from every available touchpoint - including interactions with Overstock's web and mobile sites, emails, social media feeds, and online ads. This information is fed back into the data asset platform and the various other systems Overstock uses to deliver personalization across its website, email marketing campaigns, and online advertising.

Driving both the asset data platform and the customer data platform is machine learning technology, which monitors the 40 million monthly visits to Overstock.com to learn which types of content resonate best with which types of customer. The system is then able to more effectively rank the impact that each of Overstock's content assets has on particular groups of shoppers. It also allows Overstock to segment customers based on preferences and purchase history, as well as where they are in the path to purchase. From there, the most relevant content can be served.

"If the shopper is early on the purchasing journey, she may be looking for an inspirational piece of a buying guide, or looking for reviews," Knab said. "The system attempts to understand what part of the journey the consumer is on to serve the appropriate content. [...] The system is learning. It doesn't just ask whether the asset is relevant, but whether it is the right asset to serve to that customer."

Final Thoughts

Retailers that are able to offer shoppers a tailored, personalized experience - on their websites, in their marketing messages, and across devices - can gain a sharp competitive edge in the market, and win more customers for the long term.

However, this is no mean feat. For a company the size of Overstock - where there are literally millions of possible variations of each page that can be displayed based on what's known about an individual customer and the products he/she may be considering - generating positive results is undoubtedly a challenge. Nonetheless, early results for the retailer are promising, and Knab says that Overstock will continue to deploy the new personalization technology to further channels in the retailer's sales and marketing stack going forward.

"The content the system is choosing is leading to much higher engagement," said Knab. "Engagement click-through rates climb on pages where we're implementing these systems."

Return to Blog