DATA QUALITY: AN INTERVIEW WITH TOM REDMAN, SENIOR CONSULTANT, CUTTER CONSORTIUM

Q: You made the point in last year's Business Intelligence Executive Report, "Data Quality for E-Business" (Vol. 1, No. 4) that "there are literally hundreds of potential dimensions of data quality. However, there is a core set of about a dozen dimensions that most customers consider most important most of the time." Are there any particular areas that you suggest leaders focus on today?

TR: First, leaders should focus on data accuracy. Data that is erred just gets in the way too much. The data values must be created correctly the first time. And organizations, with reasonable effort, can improve by an order of magnitude or two.

The second area of focus is clear definition of terms. Think about a common business term, like customer. "Customer" gets realized in databases and data models all the time. But different people may think about this term differently. A customer to a salesperson might be someone they need to call on to close the deal, while a customer to someone in manufacturing might be the location that they need to deliver a product. It's natural that these terms take on slightly different meanings to support different areas within the business. It's virtually impossible to get people to agree on a common definition; what must be done is that all terms must be carefully defined so that people at least have a chance of understanding when they use someone else's data. Then they can say, "Oh, he's talking about this in a slightly different way." This requires discipline in data modeling.

The third area of focus is the relevancy of data. I've made this statement many times: "Half the data in a typical organization is never used by anybody, for anything, ever." People will contradict me and say, "No, Tom, you're wrong. In this organization, it's 75%." We're creating a whole lot of data that we simply don't need. Getting rid of that data and not creating it anymore helps create a pool of money that you can spend to make more fundamental improvements.

I also want to answer the question, not from a technical point of view, but from a business point of view. What do senior leaders need to do? The first thing is to recognize the enormously high cost of poor data quality. We find that [this cost] is at least 10%-20% of revenue in the typical company. It's an enormous expense. Most of this cost is hidden in the way people work. But, it can be reduced dramatically.

The second thing that managers have to do once they recognize the problem is to make management accountability for data clear. In particular, a policy that says, "If you create the data, then you are responsible for their quality." This is extremely beneficial; it gets organizations into the realm of preventing new errors at the sources of quality problems rather than finding and fixing errors downstream.

--Tom Redman, Senior Consultant, Cutter Consortium