InformationWeek
Defining The Business Value Of Technology
|
November 6, 2000 |
 |
Amazon's IT Agenda
continued...page 2 of 4
Amazon.com has come a long way since Bezos founded the company in the summer of 1995. Back then, when Amazon.com needed to relocate its Web operations, the company's chief programmer put the server into the back of his Honda and drove it to the new location, bringing down the site for all of 20 minutes.
When Jeremiah Wilton joined Amazon.com in the spring of 1997 as its first database administrator, he was responsible for two small databases running on a pair of Digital Equipment Corp. Alpha servers. Amazon.com's data-management needs have grown to 100 separate database "instances," running on Compaq/ Digital, HP, and Sun servers, and supporting terabytes of data. There are now 20 database administrators on staff, and Amazon.com wants to add more. "Our greatest challenge, from my point of view, has been scaling our systems," Wilton says. "When time to market and customer service are the main focus at a company, scaling gets left behind."
The agreement with HP helps address the scalability problem--but as computing resources are added, that only makes the systems-and network-management job more complicated. "We have goals for availability that are astounding, considering the size and complexity of our environment," says Ravi Angadi, Amazon.com's senior systems-management architect.
HP is providing Internet servers, storage, and other systems for use in Amazon.com's Web operations and to support its distribution and supply-chain management processes. HP is also supplying desktop PCs for Amazon.com employees and support services to help keep the E-retailer's systems running during the critical upcoming holiday shopping season. Amazon.com hopes to avoid the desperate scenario its IT department faced in the fourth quarter of last year, when it brought in HP to move a 500-Gbyte transaction-processing database to a more powerful computer platform. "We were told by folks that we were nuts to do it, but we didn't have a choice," says Bell. "It was like a military operation."
Scalability, database performance, applications integration, storage, high availability--it's a technology juggling act for any Web-site manager, but Amazon.com's are particularly feeling the heat. Part of the recently disclosed deal with Excelon involves using that company's object-oriented database to improve the responsiveness of Amazon.com's site.
Adding to the challenge is Amazon.com's ongoing effort to refine the process of online personalization, such as product recommendations. "If you ask what they want most from Oracle, it's less latency," says Oracle senior VP Jeremy Burton. "The most important thing you know about a customer is what they last bought. There's too much latency between the action and the recommendation."
Turning customer knowledge into product recommendations and, ultimately, into purchases is one of the things Amazon.com does best--and is a key to its keep-the-customer strategy. The personalization capability--the company refuses to discuss how it works--recognizes when a visitor returns to the company's site and, based on that person's previous purchases, recommends products. With Amazon.com's customer base exceeding 25 million, it's one of the world's largest business-intelligence applications.
"These are huge technical challenges," Bezos said at PC Expo in June. "We have a big team working on this, and it is very, very hard, especially in real time."
Amazon.com is leveraging its personalization know-how in novel ways. For instance, shoppers can create online communities of friends and family to share gift lists and recommend CDs, books, and other merchandise. When Bezos demonstrated the feature at PC Expo, he had nearly 100 acquaintances, including his mother and Amazon.com VP and CIO Richard Dalzell, as part of his personal network.
It's a nifty application, but Amazon.com needs to go further if it's serious about becoming the "Earth's most customer-centric company." It's been expanding customer-service operations by opening new call centers. State-of-the-art customer-relationship management and business-intelligence software should also help.
Amazon.com has assembled a team of data-mining experts, led by manager Jaya Kolhatkar, that's using SAS Institute's Enterprise Miner data-mining algorithms to detect patterns in customers' online shopping habits, with a goal of increasing repeat visits and improving customer retention. "We were already doing this kind of thing, but not as intensely as we expect to," Kolhatkar says. The E-retailer calls its best customers--those who shop often or spend a lot--Amazing Amazonians, and it wants to make offers that keep them coming back.
The company is also using the SAS tools to experiment with the effectiveness of delivering different kinds of content to different customers. "In essence, this is conceptually the same as what supermarkets do with promo and display mix in stores," says Robin Way, SAS Institute's CRM strategy director. Experimenting with content is one thing; tinkering with pricing another. When customers learned that Amazon.com was using some of them as guinea pigs in a test that involved selling the same item at different prices, it didn't go over well. In September, Amazon.com said it would discontinue that experiment. It emphasized the test was random, not based on customer data.
continue on to page 3
return to page 1
Back to This Week's Issue
Send Us Your Feedback
Top of the Page
