Operational measurement of data quality

Abstract

In this paper, an alternative view on measurement of data quality is proposed. Current procedures for data quality measurement provide information about the extent to which data misrepresent reality. These procedures are descriptive in the sense that they provide us numerical information about the state of data. In many cases, this information is not sufficient to know whether data is fit for the task it was meant for. To bridge that gap, we propose a procedure that measures the operational characteristics of data. In this paper, we devise such a procedure by measuring the cost it takes to make data fit for use. We lay out the basics of this procedure and then provide more details on two essential components: tasks and transformation functions.