It's a conspiracy of sorts.
The pervasiveness of the Web and the rapid acceleration of online information-gathering technologies have conspired to make the e-customer the most demanding and fickle demographic yet to populate the ever-splintering pie charts of beleaguered marketing execs. Your ability to acquire, up-sell or retain these canny customers depends more on how much you know about the mouse catcher--and his mice--than whether you can build a better mousetrap. If you know, for example, that the targeted rodents prefer peanut butter to cheese, you can push the appropriate bait to the mouse catcher. When he realizes you've personalized your product for him, he'll come back. And when he does, he'll expect even greater personalization.
"In the e-business environment, we don't choose a vendor because that vendor has the products we want," observes David Cody, senior marketing manager for SPSS, an e-business solutions provider based in Chicago, Ill. "We go there because the vendor has better information. The products become secondary. The value from that site is highly personalized information that's highly useful to me."
This highly personalized information, or intelligence, has become the valued commodity for both customer and marketer. For the former, it increases the usefulness of the site; for the latter, it increases the value of individual customers. As Aaron Zornes, executive vice president and director of META Group, notes, this type of customer intelligence is the "currency of the digital marketplace." But to fully leverage this currency, e-businesses must be equipped to gather and process enormous amounts of information about their customers, then process and apply this information to the experience of the individual user. "A real-time, panoramic single-customer view is the Holy Grail of CRM," says Zornes. "Everyone is searching for it."
Many e-businesses have arrived at the same conclusion: that to recruit and retain customers, they have to personalize customer interactions. Effective personalization, however, requires a complex choreography of data marts, operational data stores, analytics, psychographic profiling, demographic profiling, histograms, archives, audits, traces and much more. It can also require a shift in the company's business model to match the changes in its IT structure.
But once this organizational tour de force has been accomplished, the company can begin to apply analysis to customer and key-performance data to yield true business intelligence--the real-time, single-customer view that empowers the customer relationship, as well as broader strategic decision making.
The Power of One: The Real-Time,
The critical intersection of business intelligence with e-business occurs in the front office. "In a lot of respects, e-business is just business," explains Cody. "Business intelligence plays just the same role as it does in any kind of traditional business. The new and interesting areas right now that let us exploit the intelligence opportunities in e-business are within the realm of CRM."
E-commerce is uniquely customer-centric, as compared to other customer touchpoints. For instance, a point-of-sale system only tells you what a customer has purchased from you in the past. E-commerce systems not only keep you informed on a customer's purchase history, they potentially allow you to follow that customer down virtual aisles. You know which products that customer picks up, which he puts down--and in what order. This information provides a powerful basis for predicting behavior, since this detailed data can be applied to very specific, accurate kinds of analytical models, which in turn yield highly accurate scores, or behavior ratings, for individual customers.
Most current e-businesses rely on rules-based personalization, which depends on assumptions made about customers from aggregate data gathered from a variety of sources. A quality score is based on real-time activity, enabling you to always present the most appropriate goal-meeting message possible to each individual. If you have 100 different people on your site at any given time, they may be viewing 100 entirely different messages.
"Everyone's looking for some value add," says Todd Nash, vice president of Faber Consulting, a professional services company specializing in the integration of business intelligence applications with e-business solutions. "Value add comes in the form of knowledge or information. That's part of the key to personalization--that you can get the customer to the crux of their decision faster, and you can supply them with information to help them better their decision process."
In the case of a real-time product offer, your analysis might tell you that a customer, while clicking through your site, has ordered a gray sport coat and a dress shirt. Based on this behavior, you might initiate a proactive push, offering that customer a 5 percent discount on the entire purchase if that customer also buys a tie. You might even suggest some patterns or colors. The analysis involved in this scenario can take into account any number of parameters, including that customer's past behavior and preferences. By factoring in key performance indicators (KPIs) from existing financial data, analytic models can even assess the profitability of offering that particular customer the discount while he's still browsing.
Real Time Versus Rules
Intelligence on customer interactions and all aspects of operations has been and is currently derived for most businesses through a combination of collaborative processing and rules-based customer interactions.
Collaborative processing involves the manual integration of customer and market information with operational systems that contain back-office data. Analysis occurs offline using operational documents, such as e-mail and marketing presentations, pulled from large clusters of aggregated data. The results of this type of analysis generally inform the business rules that govern the logic of the site's decision engine.
Not only does this approach delay the deployment of business intelligence, it merely mimics personalization, and its logic seldom progresses beyond facile conclusions on the order of, "Anyone who buys shoes must also need socks." Real-time analysis can take into account the fact that some customers may not wear socks at all.
"We look at rules-based personalization as rigid," explains Jennie Grimes, director of e-intelligence and portal solutions at Hewlett-Packard (HP). "There's a certain value to it, but it is flexible enough to incorporate clickstream, which is a large body of information that e-businesses have access to."
As an adjunct to its E-Services initiative, HP's E-Intelligence Division has launched E-VUE, a data-integration and analysis solution that links both internal and external data sources to provide a "360-degree view of the entire business ecosystem." The goal is to empower strategic decisionmaking on all fronts. E-VUE was designed to accommodate the growing intelligence needs of e-businesses seeking to leverage their clickstream data.
"We've announced very publicly that we now have the back-end of Amazon.com," says Grimes. "We've learned that they generate 500 gigabytes a week of clickstream data, and they have no idea what to do with it. Ninety percent of it is going to be worthless, but 10 percent will be absolutely pristine data that they need to incorporate against their histogram. Through that data they're tracking the behaviors, versus the rules, that they've been able to build up over time."
The transition from rules-based, offline processing to real-time analysis involves combining historical data on both individual users and user segments, as well as potential external data, with context-sensitive clickstream data in real-time to provide the single customer view. Whereas offline processing delivers the same offers to everyone who has purchased a given product, real-time analysis enables the database to make recommendations that are unique to that customer on that day during that session.
Dealing with Data Load
Unlike rules-driven personalization, real-time analysis demands the massive quantities of data that the Web generates. "With business rules, the more volume you get, the faster the business rules break down and the fewer business rules you can have," observes Cody. "It's the opposite for analytical systems. When you're using predictive modeling, the more data you have, the better the model."
SPSS's analytical CRM solution, CustomerCentric, allows analysis to begin with third-party data, gradually introducing actual proprietary data as it becomes available through operational systems and customer touchpoints.
Cody elaborates: "Say your company just went IPO. You have 9,000 customers. We'll start with a model that's already built based on other people's data. But as time goes by, that model will be refreshed with your information and history, making it that much more accurate and specific to you as you build the data.
Within a year, as we refresh weekly or daily, you'll wind up with analysis that's driven entirely by your data.
Cody foresees the extension of the Web's capabilities to all the other customer touchpoints. "It not that the Web will become like every other channel," he says. "It's that every other channel will become more like the Web. Your point-of-sale system will become a Web-based, data-driven system: But instead of you logging on, the sales clerk at the cash register will do it. But that clerk will still enter information about the customer as a unique individual, deriving a score, making a recommendation, providing information to the customer that engenders loyalty."
So, thanks to business intelligence, all CRM may soon become eCRM.