The case studies described below illustrate a variety of uses of data mining ranging from assessment of the credit worthiness of untried borrowers, lowering the costs of market promotion, price adjustments in real time, fraud detection and supply chain management. A common denominator in these case studies is the fact that companies become more agile and innovative as data mining helps to improve their existing business processes.

Capital One

Capital One, a financial services company, pioneered fact based decision making, based on analysis of large data sets, in its lending operations. In the past, credit cards were issued to only those customers who had a sound history of repayment and loans were extended to borrowers who could secure their loans with collateral. Capital One changed the rules of the game by using data to determine credit worthiness. Its strategy was based on the premise that customers, without a credit history, could have the potential to be good borrowers. A typical example is a recent college graduate, with limited means and a large student debt, who could go on to make a successful business career. Capital One has created a huge data warehouse, an equivalent of 20 pages for each of its 30 million customers, with information relevant for sub-prime lending. It uses this information to conduct experiments, as many as 45000 product and marketing tests in 2001, to find financially sound offers for its various segments.

Source:
www.infoconomy.com/pages/in-depth-reports/group64290.adp
www.virginiabusiness.com/magazine/yr2002/mar02/capone.shtml

Victoria's Secret

The Limited experienced rapid growth in its sales when it launched its web site in 1998. Concurrently, catalogue sales dropped precipitously. The company had to rethink its channel strategy especially because the costs of serving customers by printed catalogs were higher and more customers were acquired by the web. The company still believed that it needed to distribute catalogs to complement its web channel.

Consequently, the distribution of catalogs was targeted to customers based on the RFM or RECENCY, FREQUENCY and MONETARY value figures. Customers with higher scores on RFM are more likely to receive catalogs. In addition, the company segmented the customers for purposes of customer acquisition, retention and extension. This helped the company to lower its distribution of catalogs from a peak of 400 million to 350 million.

Source:
www.virginiabusiness.com/magazine/yr2002/mar02/capone.shtml

Harrah's Entertainment

Casino operators routinely offer low hotel rates to motivate their guests to spend more on games. Yet not all customers spend enough on games to justify the discounts on the hotel rates. Harrah's Entertainment, with 26 casinos in 13 states, used data mining to analyze spending on games in real time and customers are offered incentives while they are still in the hotel. It processes 16 terabytes of data everyday emanating from the slotting machines and analyzes data on time spent on each of the games, the preferences of tourists and local visitors, gender differences in game expenditure and a host of other variables. The share of expenditure on games has increased from 36% to nearly 50% after it started to use data mining technologies.

Source:
"Too much Information" by Daniel Lyons, Forbes, Dec 13th 2004
"The Road to One-to-One Pricing", Executive Technology, May 2004

Visa

Credit card fraud is endemic at a rate of 0.93% for off-line transactions and a higher 1.97% for on-line transactions. The challenge of detecting fraud is avoiding false alarms. In the past, ninety seven false alarms were generated for every genuine alarm. Credit card companies have to meet the conflicting demands of keeping fraud low without irritating customers with false alarms. This is hard to achieve with statistical models which compare the normal behavior with abnormal behavior because criminals are savvy enough to circumvent them.

VISA introduced its new software, VISOR, VISA Intelligent Scoring of Risk, across all banks in Europe to lower the rate of fraud. Its new system analyzes abnormal behavior of not only the card holder but also for each merchant. In addition, it now uses artificial intelligence software which changes the rules of identifying fraud based on the most recent data. The rate of fraud has dropped from 1,576 to 458 cases. The false alarms have declined to ten.

Source:
The Economist, A Golden Vein, June 10th 2004

Wal-Mart

Retail stores struggle to balance their inventory with consumer demand and their suppliers are often unable to dispatch the right amounts of goods to warehouses. Wal-Mart set up a massive data warehouse in the 1990s to store information about sales and inventories. Its Retail Link program shared information with its suppliers like Proctor and Gamble which has expertise in predicting consumer demand. The availability of real time information enabled suppliers to stock goods without waiting for orders to be placed. What's more, Wal-Mart is now able to respond quickly to exceptional circumstances like the recent storms in Florida. Based on past experiences, Wal-Mart could predict that the demand for Strawberry Pop-Tarts would rise. A streamlined supply chain has enabled Wal-Mart and Proctor and Gamble to refocus their energies on category management and to use information for merchandising.

Source:
What Wal-Mart Knows about Customers' Habits, by Constance L Hays, New York Times, November 2004

Sponsored Links: 5 Minute Guide to Business Intelligence | Business Intelligence Software Guide | Sarbanes Oxley | Balanced Scorecard Software