Don would like to thank Andrew Harrison for his contribution to this month’s column.

As with any BI effort, once the data has been captured, stored and presented for a historical report of what happened, the aggressive analyst asks, “What more can I get from this?”

Enter predictive analysis. With the vast amount of data that is captured online, it is easy to think of a long list of questions about Web site performance and effectiveness. But not every answer is equally valuable. For example, developing a model that predicts how many pages a prospect will view on a Web site is much more valuable if that Web site relies heavily upon advertisement revenue derived from page views, and less valuable when the key goal is to increase the number of new customer acquisitions. While it may be valuable to understand the optimal product sequence to serve to a given customer, or the best creative to display to a given customer segment during evening hours, it is fundamental to predict customer value potential first.

Predictive Web analytics pick up where paid search analysis ends in terms of inducing desired behaviors in users of a Web site. Even the best paid and organic search strategy can only drive people to a landing page – once a user gets to a given site, it is critical to retain that individual throughout the sales pipeline. By leveraging key data points, such as where visitors came from, their demographic and profile data, and their history of behavior on a Web site, it’s possible to predict their likelihood to complete a given sales process (such as click on a link, enter lead information or make a purchase and even forecast the potential value of future purchases.

Using statistical models to predict the likelihood of prospects to convert to customers, and the points at which a given prospect is most likely to abandon a sales process, are critical tactics for improving customer acquisition rates. By identifying points of abandonment we can target visitors with alternate content or offers, leading prospects down pathways that are custom tailored to convert them to customers. It is also important to test and measure the impact of these tactical efforts. Plugging the leaks in the sales pipeline by understanding which prospects to intervene with when will steadily improve conversion rates and acquisition cost savings.

After acquiring a new customer, the most critical thing to do is develop his/her value. In order to determine which new customers are most likely to become high value and ensure that new customers receive additional attention, predictive models of customer potential need to be developed. By looking at historical customer data and understanding which behaviors or profile attributes are predictive of value, new customers can be segmented into high, medium and low potential groups. Focusing additional sales and marketing efforts on high potential customers and reducing expenditures on customers with low potential can ensure the capture of more business per dollar invested, often increasing efficiency by more than 50 percent. Ultimately, catering to the needs of the minority of customers that drive the majority of business is an efficient way to allocate resources that also provides focus to Web design and marketing efforts.

Predictive analytics should be applied to more specific behavior questions only after understanding which metrics predict whether a prospect will become a customer (or when a customer will become more valuable). Product purchase likelihood models, optimized offer strategies or simple propensities to determine whether a given visitor is likely to click on a partner advertisement are all possible with the wealth of data available, but these activities should take a lower priority. While the answers to these more specific questions are important, understanding how to improve customer acquisition rates and developing value in high-potential customers should serve as the foundation upon which other insights branch out and inform more nuanced tactics.

The Web provides an ideal application for predictive analytics – there is a vast amount of data to harvest and much to learn to inform strategy. The biggest pitfall of working in the Web context is that there is so much data, and the number of metrics and details can be overwhelming. It is critical when getting started with a predictive Web analytics strategy to remain focused on traditional sales and marketing objectives – customer acquisition and the development of existing customers to produce more sales. Only after these basic concerns are met does it make sense to use predictive analytic techniques to build product purchase propensity models, optimize offer strategy or predict whether a customer will click on a specific link.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access