Technology is revolutionizing marketing, and business intelligence, the process of transforming business data into knowledge, is leading the way.
BI is helping companies make strategic decisions about which markets to enter, which customers to pursue in those markets and which products to develop and promote. It involves capturing all the data in an enterprise, and then analyzing it with advanced algorithms to reveal valuable information about customer buying habits.
While using BI techniques to fine-tune relationship marketing strategies for several IBM customers, we developed a new way of looking at customer relationships. In studying how companies market to customers, we found most use past sales data to make decisions only about the next “promotion.” To gain a more complete and accurate perspective of the customer, however, that sales information should be used to look forward to determine what the optimal relationship should be across time.
This new method shifts the old marketing paradigm 90 degrees. Instead of a company trying to make the most profit from each sales campaign, this idea looks to maximize profit from each customer over time regardless of the sales campaign. Companies can then focus on building the best relationships possible with customers as a means to maximize profit per customer.
To visualize this concept, think of a spreadsheet. Vertical columns represent product categories. A catalog company might assign one column to outerwear sales, another to women's clothing. A bank might assign one column to personal checking, another to mortgage sales.
Horizontal rows represent individual customers. Thus the data contained in intersecting cells spell out a customer's relationship with product categories. Scanning a horizontal row gives a marketer the full spectrum of a customer's relationship to an organization as a whole, and since the horizontal axis represents time, this “horizontal” view of customers can also give us the past, present and probable future trend of a customer's relationship with a vendor.
Here the concept of viewing customers' relationships from a horizontal perspective comes into sharp focus: it captures the relationship between a company and a customer both spatially (assessing purchase behavior across the spectrum of vendor offerings) and temporally (through time).
Let's step back to put the whole field into perspective. Overall, there are three major steps to BI. In the first step, we build a data warehouse to hold all available customer information.
Step two involves mathematical propensity scoring and segmentation so marketers can identify and group customers that exhibit similar behaviors and define unexpected relationships to help their companies sell more effectively. A good illustration is the drug store chain that mined its data and discovered that 27 percent of the time female shoppers buy cosmetics, they also buy greeting cards. This led the chain to rethink the layouts of its stores.
In step three we do highly refined targeted marketing.
An example of the third step is illustrated by Fingerhut Corp.'s catalog retail operation. Fingerhut, the second largest catalog sales retailer in 1997, mailed nearly 600 million catalogs that year, in 120 different editions. Fingerhut sold products ranging from housewares to electronic appliances to 5 million customers last year. To my mind, Data Mining Executive Randy Erdahl and his team are the best anywhere at constructing sophisticated segmentation models.
Fingerhut is a 50-year-old company, which, Erdahl told me, “almost from the beginning” started building what we now call a data warehouse. Erdahl joined as an analyst 20 years ago, at a time when the company “started to do complex scoring models” to group customers' buying behaviors.
These models provided Fingerhut with invaluable findings, such as the fact that people who move to a new home have a propensity to triple their purchases of certain household goods in the twelve weeks following the change of address.
Although this finding seems simple, the method to discern it was complex. Determining who moved when was the easy part since catalogs contain change of address cards, and the U.S. Post Office also supplies that information for a fee. Erdahl explained how his team labors to uncover this and other marketable gold mines of findings:
“We have recorded every purchase that a customer has made. Out of hundreds, sometimes thousands, of variables we will find a handful that show the ability to discern the difference between buyers and nonbuyers. Customers showing the highest propensity to buy in all probability would be the ones to get the [appropriate] catalog.” The speed of this process enables Fingerhut to mail a movers catalog in a timely manner.
No matter how good their returns, the sheer number of Fingerhut's mailings imposed enormous costs.
This is where IBM was able to help Fingerhut optimize mailings by reducing over-mailings. IBM started by thoroughly understanding the way they made their decisions. We knew that to solve for the “optimal” relationship between customer and Fingerhut, we would have to understand how the stream of promotions affects the customers' purchases and how catalogs mailed too closely to one another actually cannibalized overall customer sales. Each customer or customer group is allotted a limited amount of promotional dollars.
Each week data from 7 million customers is downloaded to an IBM SP2 computer dedicated to organizing the mailings. At the same time Randy's team is optimizing mailing decisions for 20 to 30 catalogs.
The results of optimizing by viewing customers “horizontally” speak for themselves, as seen in this example: For a very active customer who made a purchase within the last six months; in the past, Fingerhut might have spent $13.50 on direct mail for a return of $123.56 in revenue. Today $12.49 in direct mail yields $121.67 in revenue. That's a 4 percent increase in revenue per direct mail effort.
The president of Fingerhut sees this as transforming his business. Instead of sending customers catalogs containing ever more merchandise, the company can now feature merchandise that specific customers are likely to buy. Likewise, the vendor's emphasis is shifted, from “catalog” to “customer,” giving Fingerhut more opportunity to develop a relationship with individual customers that will encourage them to keep coming back.
All of Fingerhut's mailings are now optimized this way. In the final analysis, the new BI techniques developed by IBM and Fingerhut have changed the way most marketers will look at direct mailings in the future.
Michael Haydock is vice president of Global Business Intelligence Solutions at IBM, Somers, NY.