eBay’s Latest Search Technologies Deliver More Targeted Results
By: eBay Inc. Staff
Based on extensive user tests, new search features sift for exactly what shoppers want.
Over the next 12 months, eBay Inc. is launching a series of new search technologies and initiatives, designed to vastly improve user experiences at the company’s sites. Many of the changes are a result of ongoing user experience tests, and there are accompanying shifts in site design that will add to the benefits.
Hugh Williams, Vice President of Search Engineering, is spearheading the new initiatives. Williams joined eBay 18 months ago from Microsoft. There, Williams managed the Bing Search engineering teams that were responsible for image, video, and news search, as well as many aspects of web search. Prior to Microsoft, Williams was a Professor at RMIT University in Australia, where his research focused on information retrieval and search algorithms. He’s published over 100 works, including three books and 70 academic papers.
We caught up with Williams for an update on search changes coming at eBay, and here are his thoughts.
What are some of the primary search initiatives that you’re working on?
Over the next year, several big things will happen in search. One that I’m super excited about is that we’re going to introduce machine-learned ranking. This means that the order of search results will be driven by more data factors than we currently use, and we’ll be able to combine these factors using a computer-generated model. In practice, this means the results will be more relevant to our users, and our search engine will adapt more easily, quickly and accurately to what our customers are doing on the site. We’re already testing this on a small number of our users, and we’re seeing great initial results.
Are you overhauling the entire search engine?
Yes, we’re also working on a major reengineering of the search engine itself. We call this project Cassini, which is an upgrade to our current Voyager search engine and based on cutting-edge technology. It will be much faster, and more flexible – and it’ll really enable more advances in the machine-learned ranking technology and many other initiatives in search.
You mentioned earlier that you’re testing machine-learned ranking on our customers already. What else are you testing?
We’re always testing at least 50 new experiences on customers at any given time. Most buyers are in at least one or two experiments, and typically don’t notice. For example, right now, we have over 15 different search ranking algorithms on the U.S. site, a new crop of experiments to improve the speed of the site, and several tests of new merchandise placements.
How do you decide when to turn an experiment into the main experience that users see on the site?
We measure how our users in the experiment interact with the site, and compare this to the control group of folks who aren’t in any experiments. When we see users benefiting from the experience, we’ll usually make that the default experience for most customers when we do a software release. If we don’t see a benefit, we’ll go back to the drawing board and analyze what was and wasn’t working. So, even failed experiments help us understand our customers better and improve the experience.
Do you have any search upgrades planned to arrive for the holidays?
Yes. We’re moving toward a ”product-based” search experience, where users won’t have to sift through so many of the same types of search results. When you search for a product, you’ll see a great deal for that product right up front.
In fact, we just shipped this new experience for GPS units, MP3 players, and DVDs. Using the new experience, it’s super easy to get a great deal. For example, if you want to buy a GPS unit, you’re now easily able to choose between buying the best deal on a new one from a Trusted Seller, or a refurbished one, and you’re be able to read reviews, see price trends, and more. Try a query such as iPod nano, Garmin Nuvi, or Finding Nemo.
What other technology changes can eBay buyers look forward to in the next year?
We’re launching a new home page—one that we’re really proud of. It has a great look and feel. It’s more personalized too. For example, you can sift through searches you recently did, and explore hot, trending products.
We’re also making more and more progress on our image similarity search, and you’ll see that continue to evolve and improve as part of our new fashion experience. The fashion experience itself is evolving nicely too – I’m excited that we now have personalized the experience, so that you can put in your sizes and only see items that fit you. It’s also got some cool ways to browse by style. You can check it out at fashion.ebay.com.
From a personal perspective, I’m especially excited about the search upgrades we’re rolling. We’re going to have a very modern search engine that delivers great results, and I’m excited about our customers being able to experience that.