Properties of Data Quality and Metrics for Measuring It

Abstract

Metrics for measuring data quality (or lack of it) are valuable tools in giving us some quantitative objective measure of the problems we may be dealing with. In this chapter, we first discuss a few key properties of high-quality databases/lists. This is followed by a number of typical examples in which lists might be merged. Finally, we present some additional metrics for use in assessing the quality of lists produced by merging two or more lists. Although quantification and the use of appropriate metrics are needed for the quality process, most current quantification approaches are created in an ad hoc fashion that is specific to a given database and its use. If there are several uses, then a number of use-specific quantifications are often created. For example, if a sampling procedure determines that certain proportions of current customer addresses are out of date or some telephone numbers are incorrect, then a straightforward effort may be needed to obtain more current, correct information. A follow-up sample may then be needed to determine if further corrections are needed (i.e., if the database still lacks quality in some respect). If no further corrections are needed, then the database may be assumed to have an acceptable quality for a particular use