PERSONALIZED PRICING... EASIER SAID THAN DONE
by Ed Colet

The business-to-consumer space of e-commerce is becoming a mainstream part of the economy. But relative to the business-to-business space, the prospects for revenue growth pale in comparison. Although there are new and interesting business models associated with e-commerce, success in the business-to-consumer space still depends on the ability to maintain a profit margin. But the particular nature of the Internet can make this difficult, so personalized and dynamic pricing strategies are seen as necessary. In this column, I take a look at some of the complex issues that need to be addressed in order to implement personalized and dynamic pricing.

A new report from Forrester Research, Inc. "Pricing Gets Personal" concluded that consumers empowered by information technologies such as shopping agents have a significant advantage over retailers. The end result for retailers is shrinking retail margins and a diminishing likelihood of profitability. According to Forrester's report, the way to restore adequate profit margins is to adopt personalized pricing strategies and the delivery of customized offers to consumers.

Forrester's approach to implementing dynamic pricing involves first identifying and segmenting customer-buying behaviors into three general categories. "Price grabbers" want bargains and price is the most important factor. "Express shoppers" look for convenience rather than price. And "affinity buyers" make decisions based on lifestyle rather than price or convenience. Adaptive pricing involves mapping a different offer and price to different types of consumers. This involves analyzing on-site behavior as well as monitoring prices offered by the retailer's competitors. Lowest prices will be offered to bargain hunters, and the profit margins will be maintained by offering higher prices to the other types of shoppers while stressing convenience and/or affinities.

Although the plan has some intuitive appeal, it is certainly easier said than done. Although personalization is becoming an integral part of information delivery to consumers, the delivery of personalized pricing is a much more complex operation for a variety of reasons. These include limitations on what you can ask consumers, and differences in how to determine what gets delivered or offered to whom.

If the task is to deliver personalized information (e.g. news content) to a consumer, this can be partly achieved by asking the consumer to explicitly indicate their interests in topics and content areas. But when it comes to dynamic pricing, the retailer can not simply ask consumers what they wish to pay for the item (i.e. ask consumers to make a bid) because this would almost always be less than the price that the retailer would originally want to sell the item for. What consumer would indicate a higher price than necessary? (Although consumers can name their price on sites such as Priceline.com, the retailer is under no obligation to then deliver the product at that price). In terms of personalized delivery, the retailer would be obliged to deliver the product at some point.

Unlike personalized prices, the delivery of personalized news content is largely a matter of selecting a subset of content from a larger pool. In terms of dynamic pricing, determining the price to offer a consumer is not a question of selecting a subset of content. In order to be useful, dynamic pricing models have to rely on sophisticated customer profiling tools -- taking into account their purchase history, their demographic information, etc. This touches on potentially sensitive privacy issues -- i.e. should my current income be known to a merchant so that they price a product based on what they think I can pay (i.e. a higher price). And even extensive information about the consumer is not enough. Dynamic pricing modeling also has to take into account factors associated with the retailer -- current levels of inventory, current prices from suppliers, general state of supply and demand for the product, etc. And one has to consider the fact that customer profiles are also likely to be dynamic. I may be a "bargain hunter" for certain types of products and at certain times, but a "convenience shopper" at others. All of these factors contribute to complex and sophisticated models that are necessary in order to determine and offer a personalized price to a particular customer at a particular time.

Last but not least is the notion of fairness. Is it fair or discriminatory that the same retailer is offering a different price to different people for the same item? As long as consumers are generally unaware of the practice, the retailer remains in their good graces. But if consumers find out that they're consistently being charged more, then the retailer could find themselves faced with a difficult public relations problem to deal with.

To conclude, achieving accurate and useful dynamic pricing models requires extensive data analysis that are certainly possible with current data mining approaches and technologies. But it is a task that is certainly easier said than done.

Ed Colet is the Acting Director of Research at Virtual Gold Inc., responsible for developing analytical methods for data mining and for investigating human factors and usability issues of business intelligence systems. At present, he is in the final stage of completing a doctoral dissertation in the Cognition and Perception program at New York University's Department of Psychology. Ed has also worked for IBM Research at the T.J. Watson Research Center. At IBM, Ed was a member of the group that developed Advanced Scout, the data mining application for NBA teams. His research interests focus on statistical methods and human factors.

For more information, see www.virtualgold.com.