Web Mining Crucial to E-Commerce Success
But the Web's growth has created intrinsic difficulties beyond the ability of any one company to solve. The Web has more than 800 million pages spread over 3 million servers with a collective storage capacity of 15 terabytes. Some large portals have millions of users, hundreds of thousands of pages and tens of millions of page views in a day.
To retain customers, these Web sites need to maintain the responsiveness and interactivity of smaller sites. At these levels, the data in the portals stretches the working limits of existing hardware and software. The data mining algorithms supporting customer-centric applications also need to be more scalable, and the predictive models - which are the backbone of real-time personalization and interactivity - also require instant validation and more frequent updating.
The Web is bringing us closer to reading the minds of consumers and influencing their decisions. We no longer just observe and record the purchase event but track and log all consumer activity pertaining to prepurchase considerations. Data mining is what makes it possible to predict behavior and respond in a relevant way.
Web mining is the application of data mining techniques to discover actionable and meaningful patterns, profiles and trends from Web resources. It couldn't come at a better time because forecasters estimate a 40 percent boost in visits to e-commerce sites this holiday season: an eye-popping 1.3 billion prospective hits to e-commerce sites.
Most businesses, however, are not yet focused on e-commerce enough to maximize its potential for sales or for greater knowledge about their customers. While individual businesses cannot advance hardware and software technology or invent better data mining algorithms, they can learn more about the likes and dislikes of visitors.
Mining your data is the key to maximizing the e-commerce potential of your site. Gathering information about the traffic hitting your site may be useful, but it has its limits. Knowing which search engines refer the best customers is more valuable than knowing which search engines were used.
Web mining includes three areas:
• Usage mining is the discovery of site access patterns as logged in the server access logs. This analysis includes custom reporting, usage profiling, banner ad targeting, real-time recommendations and cross-sale analysis to such customer relationship management applications as customer attraction, segmentation, retention and Web-time value.
• Web structure mining refers to the application of data mining techniques to improve the structure and design of pages and sites. Structure mining could help assess the problem areas on your sites such as major traversal paths associated with quick exits and paths that lead to sales and cross sales.
• Web content mining refers to the automated search, extraction and classification of primarily textual content information resources available online. Application areas include customer support and service, automated e-mail routing and reply and knowledge management.
The first step in unlocking your Web site's potential is to hire an e-mining service provider to identify strategic and proactive uses of this data. The electronic paths of all pre-buy actions and site visits allow you to optimize the arrangement of the pre-purchase site areas.
Once Web mining has established the most common click paths through your site, redesign the site to make it as simple and direct as possible for visitors to find what they need. Experiment with the best placement for special offers and personalized messages and locate and fix bottlenecks.
Data mining is now on the cusp of commercial success. By using its techniques, you will be able to identify your site's most popular areas, predict consumer actions, analyze the results of targeted marketing campaigns and recognize when areas need improvement.
The businesses that respond to their best customers in a meaningful way will be the winners.
Ismail Parsa is senior director of quantitative analysis of the analytic consulting group at Epsilon, Burlington, MA. His e-mail address is email@example.com.