ebay's secret weapon: stimulating buying with big data

1.8 million of buyers and sellers are active on ebay, and the site generates a lot of data every day. At any given point in time, there will be about 3.5 million items listed for sale through ebay's auction search engine that has more than 2.5 million queries per day. ebay's Hadoop cluster and Teradata devices typically hold 10PB of raw data, said Hugh Williams, vice president of the ebay search platform. Online auction site ebay uses many features of large data, such as measuring site performance and detecting fraud. But one of the more interesting uses for collecting large amounts of data is to encourage users to buy more goods on the site.

While ebay cannot force users to buy every product they encounter, ebay makes great use of big data to promote it. One way is to optimize the search engine and search results, through the data collected to analyze the user's behavior patterns, to adjust the results.

"If you go back a few years and use a search engine on ebay, you might find it too ' literal '," Williams said. "There are things you can say to a search engine that will literally find the information the user needs, but it doesn't really understand the user's intentions. ”

"We've been trying to make our search engine more intuitive. "For example, by using large data, ebay found that if users want to buy a pilzlampe, this is a collectible German mushroom lamp, when they enter" Pilz Lampe "in the ebay search engine is more likely to buy, because this input will have more results.

In search engines, simply add a space bar to the middle of a word, and ebay can improve sales opportunities through the site. With this information, ebay has changed and rewritten user search queries through its search engine, adding synonyms and alternative terms to bring more relevant results.

Not only that, ebay uses big data to make predictions about whether the listed products will be sold, what prices will be sold, and how much will affect the search engine of the auction site.

All of this can increase the likelihood of users buying.

Wlilliams that the model of the search query implementation factors are risky. "It takes a few months to implement a factor, and there is a very high risk because we don't know if it really works for the customer when it comes to helping our customers find the project," he said. That's why ebay usually runs some tests on the site and gets the user's sample group to measure the response.

Another challenge is to take the environment of the search query into account. An example is that if a user looks for "geelongcats", the ebay search engine may simply use "Cat" as a keyword and search in the pet category-which is not much use when the user is searching for sporting goods.

"There is a potential for very subtle problems within our control, so we need data for scientists to study these issues," says Williams. ”