The Corporate Household
Defining the business leads to better data quality.

A. Blanton Godfrey
agodfrey@qualitydigest.com

Last fall I participated in the seventh annual Conference on Data Quality at MIT. It had been some time since I'd thought seriously about the many ramifications of data and information quality. As I struggled to put my thoughts on paper, especially tying the past eight years of working in Six Sigma quality to data quality issues, I experienced several blinding flashes of understanding. The first, and perhaps most obvious, was that the structure of problem solving used widely in Six Sigma (i.e., define, measure, analyze, improve, control) fits data quality issues perfectly.

One of the subsequent flashes concerned the old axiom, "Defining the problem is half the solution." This was clearly demonstrated at the conference during both the opening talk and the lunchtime keynote address. The opening speaker noted that many of the previous years' papers and research had focused on defining the problems of data and information quality. He challenged this year's speakers, and all of us engaged in research in this area, to start presenting--or at least exploring--solutions to the problems already defined. Although many of the presentations continued in the simple problem-definition vein, others described organizations' actual efforts to define data and information quality issues before--or at least while--they implemented enterprise resource planning systems.

The most stimulating talk for me was the keynote address, "Defining Corporate Households," by Stuart Madnick, a professor at MIT. Using as his analogy the U.S. Census Bureau's difficulty in defining a household, he expanded on this problem in terms of corporations and other organizations. He used IBM and MIT as examples and challenged us to answer the questions, "How much did IBM sell to MIT last year," or conversely, "How much did MIT buy from IBM?"

First, we must define MIT. Does it include Lincoln Labs? Do we mean simply the university or all the students as well? Does this include part-time students, adjunct faculty, visiting faculty and others? He pointed out the numerous definitions of MIT as they appeared in published reports. Then he demonstrated how many different answers we might get if we searched IBM's sales databases, or any other supplier to MIT, depending on the way we searched. He cited more than 20 ways MIT might appear in the sales database (e.g., MIT, M.I.T., Massachusetts Institute of Technology, Mass. Inst. Tech.). Of course, sales records might also list purchases from individuals, labs and centers that were sent to offices or home addresses without any mention of MIT. So, related questions are: Which databases do we actually search? Do we search only IBM's databases for MIT sales? Do we also include third parties such as stores and distributors that sold IBM equipment to MIT? If an MIT faculty member buys an IBM laptop from a computer store, does that count as an IBM sale to MIT?

We must also look at the IBM side of this interesting problem. What is IBM? Do we count all IBMs? What about recent acquisitions? Do we count only the corporation's sales after these companies were acquired? What about companies only partially owned by IBM, or those that are joint ventures? Do we weigh their sales by fraction ownership? Is our original question simply about products sold to MIT by IBM, or are we also counting service sales?

During his presentation, Madnick asked us to think about our own problems in defining corporate households and asked for examples. I realized this was an issue I'd been addressing for years without labeling it. Since I've been appointed dean of the College of Textiles at North Carolina State University, I've struggled with the question, "What are textiles?" Weekly, sometimes daily, different articles appear in newspapers, business magazines and trade journals discussing international competition, job losses, growth or decline of manufacturing in the United States, sales growth or decline in different market segments, trade surpluses or deficits, and other issues. Why are the numbers so different, depending on the source? The naive answer is that the different trade organizations have their own reasons to present numbers in the way that makes their points.

These difficulties in defining the corporate household help explain why we so often disagree about numbers. What do we really mean when we talk of sales? How much of a company's revenues come from a certain product? How much of the profits come from other product lines? How much is exported and imported?

Years ago I served on a National Academy of Science panel on international trade statistics. I chaired the committee's quality subgroup. We were trying to define the quality of U.S. trade statistics. The real questions were, "How big is our trade deficit?" and "Do we even have a deficit?" The truth was, we didn't know. The data were so bad, the definitions so flawed, the processes so poorly defined and managed that we really didn't know how big the deficit was for goods. We decided not to even try to measure service trade or capital flows. And as far as this column is concerned, that's a topic for another day.

About the author

A. Blanton Godfrey, Ph.D., is dean and Joseph D. Moore professor at the College of Textiles, North Carolina State University. He is the co-author of the recently published Modern Methods for Quality Control and Improvement, Second Edition (John Wiley & Sons, 2001). Letters to the editor regarding this column can be sent to letters@qualitydigest.com.