Data Mining Examples
Letís briefly review some examples of data mining applications. Some applications include: predicting which customers are likely to buy which products and when; improving credit/loan/mortgage risk analysis; identifying new untapped market segments that might be profitable; predicting which securities to buy/sell when; improving customer service, support, satisfaction and loyalty; understanding which factors affect profit and productivity; and detecting fraud earlier to avoid losses.
One example is identifying characteristics of users of ATM cards at points of sale. Some people never use their ATM cards at points of sale, others use their cards only a couple of times per month, and some use their cards quite frequently. Frequent users generate the most revenue for the financial institution that issues the card. Genetic data mining was used to evolve prediction models for several levels of card usage, based on parameters such as customer age, average checking account balance, and average number of checks written per month. Using these models of frequent users, the financial institution is able to target people matching the frequent-user profile for promotional campaigns (cf., http://www.ultragem.com/sample.htm).
Firstar Bank used data mining to determine which customers are likely to be interested in a new service. Data mining allowed Firstar to do target mailings saving the company time and money compared to broad mailings to all customers. As a result of the targeted mailing the response rate to the mailings increased by a factor of four (Freeman, 1997).
Siemens uses a DSS built using case-based reasoning to aid technical customer support services staff. The program uses the results of previous customer inquires to help quickly answer the questions from current inquires.
As the result of a data mining project done at ShopKo, managers discovered that the sale of film does not cause the sale of a camera, however the sale of a camera generally causes the sale of film. Data mining may find relationships that managers already knew existed. One hopes new knowledge and relationships are also discovered.
American Century used data mining to find information to help them cross sell financial products to existing customers. Developers shared a number of lessons they learned from this project. One lesson is that senior executive support, as well as IT support is necessary for success. Another lesson is that business issues must drive project development. If the project will not benefit the company, resources should not be allocated to it. They found that data mining often yields specific results rather than general rules. The quality of the data had a direct effect on the usefulness of the results. Finally, they found that data mining requires statistical skills, business skills, and analytical skills in order for the company to get the most benefit from the tools.