Data under the Scope

Mining corporate databases means even the smallest detail can turn up gold.

Burke Campbell and Murray Conron
Financial Post
 

UP CLOSE AND PERSONAL: Data mining software enables companies to get a more precise interpretation of their customer information.

Northern Telephone had a business challenge. To boost its profitability, it needed to draw upon its market strengths, identify its most profitable services and customer groups, and capitalize on emerging customer needs. The monthly service reports and billing summaries did not provide intuitive pointers as to where to develop the business. Yet Northern Telephone knew it had the means at hand, by mining the information buried in the mountains of data it had collected from all of its customers' calls.

Northern Telephone is an example of how enterprising companies are using data mining tools to wring profits from their databases. This Bell Canada subsidiary, based in New Liskeard, Ont., provides telephone, cellular, paging, mobile satellite and high-speed Internet service to more than 66,000 customers in Northeastern Ontario.

Data mining is ideal for businesses such as phone companies, because it can help them sort through the reams of customer data to fine-tune their products and services. Data that itemizes daily household usage, who calls whom, time and duration of calls, and whether a voice line is also for fax can lead to new products to benefit the customer and increase income for the business.

For example, households that make many lengthy calls between 3 and 6 p.m. probably have teenagers who would co-exist better with their own private lines. Extensive telephone usage between 9 a.m. and 5 p.m. characterized by voice, fax and modem signatures suggests a customer with business activity.

Northern deployed data mining software from Cognos Business Intelligence tools. "Our marketing group has gained new insights into the customer base. The bonus came when we could deploy these tools over the Web for the sales group," says Doug Clark, data warehouse manager at Northern Telephone.

The managers now view business performance indicators at a glance while market analysts can drill down to target new groups of subscribers. Staff can generate the service reports in a fraction of the time and are empowered to perform analysis on payroll and frequency of repair.

Target marketing to newly identified customer groups with special service lines has boosted Northern's sales of lines, functions and equipment.

A similar pilot program is just wrapping up successfully at Northern's Quebec counterpart, Telebec.

Having weathered these two projects involving the stockpiles of telephonic data, Mr. Clark says: "You can never be too rich, too famous, or have too much disk space for a data mining project."

During tight economic times, more businesses are looking at mining their richest resource --the corporate data collected on internal operations and transactions with customers. The data mining software market has been growing at 16% annually and will exceed US$1-billion for the year 2002. Convergence of technologies accounts for a lot of this acceptance. Computing power and data storage have become affordable and widespread. A decade of data warehousing makes massive amounts of data accessible for analysis (measured in gigabytes or billions of bytes, and terabytes or trillions of bytes).

Also called "knowledge discovery," data mining unearths relationships within the data that may not even have been suspected. For years, analysts and managers have plumbed databases and data warehouses, using standard query and analysis tools, including OLAP (Online Analytical Processing). They have asked questions such as, "Did sales of Product X increase as we expected just before Christmas?" or "Do sales of Product X decrease when there is a promotion on Product Y?"
 

In contrast, the data miner probes with more precise queries, such as, "What are the significant factors that affect the sales of Product X?" or "What is the profile of people who are most likely to respond to our next mailing?" or "Which existing customers are likely to buy our new product?" or even "How can we improve production and reduce waste?"

Out of all possible indicators, exhaustive modelling and analysis combined with human expertise isolates the few patterns or relationships in data that have true business significance. Before launching into this investment in technology and data exploration, corporations naturally look for the return on investment. Data mining projects, properly designed and executed, deliver lower operating costs and increased efficiency, by identifying the winners and losers.

For example, compared with mass campaigns, the target mail or phone campaign cuts solicitation and distribution costs, boosts the customer response rate and reduces annoyance of uninterested customers.

Project costs depend on the availability and readiness of suitable data, computing power and knowledgeable staff, and the returns from data mining work may not be immediately apparent.

Banks are always seeking ways to increase revenues from credit card operations, mainly through more interest earned and reduced administrative costs and liabilities. Before reducing the minimum required payment as an incentive, for example, a bank may mine the history of credit card balances, payment timeliness and credit limits, and design different strategies for targeted customer groups.

Attracting and retaining credit card subscribers effectively has helped CIBC stay ahead of its competitors. The bank has found SAS data mining products helpful. "In today's competitive market, the window of opportunity with any customer is very small. Our target marketing and campaign management methods help us get through that window more quickly," says Julio Tavares, director of database marketing for CIBC Card Services. "Advanced analytics have allowed us to predict what our customers will look for and want to do in the future. "The Royal Bank has recently gained a comprehensive view of each of its customers' behaviour. The focus is on client profiling from all transactions at the branches, bank machines and Web portals, and servicing requests at the call centres. Software from NCR Teradata integrates the data warehousing and mining.

"Measuring client responses and profitability is vital to the design of our marketing programs," says Kevin Purkiss, senior manager of customer analytics for the Royal Bank.

"Providing our customers with the right products and services at the right time is a key contributor to our business success."

New mining tools for marketing and other research continue to emerge. For e-businesses, Web mining studies the traffic pattern of each visitor to a Web site. By categorizing customers from the selections they make, the e-business can use this intelligence to prepare visitors for a more enticing reception on their return.

Text data mining links the context in masses of free-form documents. A promising example, the University of California at Berkeley's Lindi system helps geneticists search all biomedical research literature to determine how newly discovered genes might react in treating a particular disease.

Video mining can index video news archives as vast as any CNN would have for efficient computer searches. "With competition so strong in the information-rich industries such as financial services, insurance and telecommunications, organizations that master data mining to shape investment and management strategies will be the leaders," says Gordon Linoff, a founder of the consulting firm Data Miners Inc. in Boston.

For the rest of us, it is comforting that business continues to refine data mining techniques in ways that will increase customer service and satisfaction.

~~~