One of the most commonly cited reasons for companies being disappointed with the results of building a data warehouse is that they just didn't get the financial return they expected. When asked to prove the value of the warehouse, they usually dance around the issue unless forced to come up with a figure, in which case they use soft numbers based on unverifiable assumptions.

Data mining is different. It often produces impressive quantifiable benefits across a broad range of industries in a wide variety of applications. Data mining yields firm numbers that can make the case not only for data mining, but for your whole data warehouse effort.

For example, a large wireless company dramatically increased their profitability using data mining. Faced with a high churn rate (percentage of customers leaving), 40 percent of the customer base still using analog as opposed to digital services, and a low monthly minutes usage that resulted in an average revenue per user of less than $50, they turned to data mining. If they could keep and upgrade more customers, the potential payback was significant. According to the published material, they might otherwise lose 700,000 customers per month, at an annual replacement cost of $360 million!

The data consisted of hundreds of fields, with approximately one-third coming from call detail records. Using SPSS's Clementine to mine the data on a Teradata platform, they built a series of models that scored customers on their likelihood to leave and succeeded in finding sets of rules that would predict customer behavior. They confirmed the wisdom of delivering the right offer at the right time, which meant talking directly to customers as well as sending customized direct mail. In order to succeed, they needed several coordinated teams to work together. Based on these actions, the company reduced customer churn by one-third in the contact group and slashed the direct mail budget by more than 60 percent. In addition, they increased the monthly minutes and associated revenue from subscribers as well as increased the number of customers switching to digital service.

This type of success story has been repeated at numerous companies. A large European bank had been using an older data mining algorithm (CHAID) to cross- market personal loans to customers of another banking service. CHAID enabled them to send solicitations to only 50 percent of their customers yet reach 85 percent of the expected responders. (Without a predictive model, you'd have to mail to a random 85 percent of customers to reach 85 percent of expected responders.) However, they gained the ability to explore additional predictive models when they started using SAS's Enterprise Miner, which has a variety of algorithms. The result was that now they needed to mail to only 40 percent of their customers to reach 85 percent of the responders. The 20-percent improvement resulted in a savings of hundreds of thousands of dollars each year, easily justifying the cost of data mining technology.

These results and returns are certainly impressive. Consequently, data mining is spreading out of its traditional areas of strength into other businesses such as the pharmaceutical industry where some cutting-edge applications may help in the research and clinical use of drugs. Dr. Robert Small, Vice President of Data Mining at GlaxoSmithKline (GSK), reports that not only are they using data mining in traditional marketing applications, but they are also using data mining for analysis of call center data and in research and development to analyze genomics data.

Dr. David Memel, director of Medical Outcomes Research and Economics at Roche Labs, Inc., is investigating using data mining to get additional value from clinical trial data. In particular, they are trying to determine for whom certain drugs work best and which people are more susceptible to adverse reactions to a drug. Like GSK, Roche is also exploring the use of data mining in drug discovery, research and development, and marketing.

Sometimes organizations worry needlessly about getting into data mining. They worry that they're too small to mine data, that they'll need a Ph.D. statistician to use data mining tools and that they'll have to build a data warehouse first. Happily, none of this is true. You can find good entry- level data mining tools for less than $15,000. With some training in both the tool and how to use data mining, a competent business analyst can accomplish quite a bit; and while having an appropriate data warehouse is certainly advantageous, it's not required for data mining. Vendors are competing for your business and want to lower the barriers to trying this new technology.

Data mining investments are already yielding large measurable returns. This technology is an essential tool in fulfilling the promise of data warehouses.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access