Few companies generate as much data as the online auction and shopping website eBay. Each time a user searches for a product, bids on an item, or navigates to or away from the site, the company collects information about the user interaction. Every day, the amount of data eBay processes adds up to an astonishing 50 petabytes—or 50 quadrillion bytes.
The company has been one of the biggest and longest lasting success stories of the early dot com era. And a big reason for eBay’s success has been its effectiveness at collecting and using data, according to Michael Goul, chairman of the information systems department, who developed an academic case based on eBay’s data management approach. He found that strategic use of data has enabled eBay to dramatically reduce transaction costs for users, enhance the experiences for both buyers and sellers, and save millions of dollars for the company.
“This is a key for eBay,” says Goul. “The company is all about data. If they can find ways to use data to build trust and to improve the user experience, that is exactly what they’re going to do.”
In carrying out his study of data management at eBay, Goul worked closely with eBay Vice President of Analytics Bob Page, and he also relied on the writings and presentations of then Senior Director of Architecture and Operations Oliver Ratzesberger. A written report of Goul’s findings are available as a case study for faculty and students in information systems. It is available for free for academics through the portal: www.TeradataUniversityNetwork.com
Ways to use ‘big data’
Businesses of all kinds are struggling to manage what has been an explosion in data, triggered by the extraordinary advances that have occurred in recent years in computers, communications, and storage. The volume of business data worldwide doubles every 1.2 years, according to estimates.
Much of the data comes from new hardware and software systems that firms have installed to manage information that comes from within their companies or from interactions with clients. The Internet and mobile phones are other major sources of new data.
“The term that is being used now is ‘big data,’ ” Goul says.”We’re finding that big data can change a company’s strategy. It can create information products for the organization. Different companies are trying to figure out right now how to leverage their big data. It is a real challenge—and an opportunity.”
A company built on data
Founded in 1995 by then 28-year-old software developer Pierre Omidyar, eBay grew rapidly in the early years of the Internet and went public in 1998. By 2010, the total value of goods sold on eBay was nearly $61.7 billion—more than $2,000 per second. Over 200 million items are listed on eBay on any given day.
Officials at eBay recognized early in the firm’s growth that the huge volume of data being generated by the company was getting too big and unwieldy for eBay’s existing systems to manage, according to Goul.
“Frankly, what was happening was that their big data was getting too big,” Goul says. “They had to figure out new ways to go about storing it and to be able to manipulate it.”
The company hired Teradata, an Ohio-based data solutions firm, to manage eBay’s data warehousing and analysis. With operations in 42 countries, Teradata is the world’s largest data solutions company.
Teradata’s approach involved first collecting eBay’s vast data stores and making sure the information was usable.
“You have to do some things along the way to storing data in a data warehouse,” Goul says. “You have to clean it and scrub it so you can make it useful for the people that want to leverage it.”
Data often arrives from disparate sources, and, in the case of eBay, Teradata made certain that formats were compatible and that data arriving from different places could be integrated seamlessly, according to Goul.
Renting space: A “virtual” private data mart
Once the data warehouse was established, the next challenge for Teradata was to find a way for eBay to use the data.
According to Goul, Teradata devised a system whereby the many different business units at eBay looking to analyze data could have temporary access to reserved space within the data warehouse. In the parlance of data analytics, eBay’s analysts could “rent” space to create a “virtual private data mart.”
“The enterprise has a big data warehouse, and the company can say to its business units, ‘We’re going to allow you to work with a portion of it,’ ” Goul says.”The finance department may be allocated one portion of it, while marketing is allocated another. Eventually, they release the space to another group. We call it ‘multi-tenancy.’ ”
Another term that has arisen in the world of big data is “sandboxes,” which refers to the secure environments or virtual data marts within the data warehouse where analytics are conducted by business units. Because the structure goes away once the analysis is completed, the process has come to be known as, “analytics as a service.”
This approach is more efficient than providing each business unit with its own collection of data, according to Goul. “This way the individual units have a space where they can access the data and have the tools they need. They can do their analysis, and then the virtual data mart goes away.”
In keeping with eBay’s identity as an enterprise built on data, the company employs about 5,000 data analysts spread across 50 business units.
Turning clicks into profits
A visitor to eBay might search for an item on the site and buy it immediately. Another customer might browse several items without buying, then purchase something completely different. By analyzing these and hundreds of other consumer behaviors, eBay can learn whether revisions to the site and changes in business practices will improve the company’s bottom line.
According to Goul, eBay has already scored some significant business wins through its data analytics. The company has long paid search engines for prominent placement in the results for certain keywords. By analyzing its big data, eBay was able to determine which keywords were worth buying and which ones were not, according to Goul.
“When they looked at the details, they were able to tell pretty much what the most effective use of their ad words was,” Goul says. “They learned what ad words were bringing people in and which were the best ones by time of day and day of the week.”
Data analytics also helped eBay to find ways to improve product result displays and matching, according to Goul. Matching searches and products is an important and challenging task for eBay, which has more than 50,000 item categories.
“They learned the approaches that work best to match buyers and sellers,” Goul says. “That’s really the bottom line.”
Other gains from eBay’s data-based approach have been reduced fraud, faster transactions, and overall better experiences for participants, according to Goul.
“Their goal has been to build trust,” he says.