BR4DQ: A methodology for grouping business rules for data quality evaluation

Abstract

Data quality evaluation is built upon data quality measurement results. “Data quality evaluation” uses the “data quality rules” representing the risk appetite of the organization to decide on the usability of the data; “data quality measurement” uses the business rules describing the “data requirements” or “data specifications” to determine the validity of the data. Consequently, to conduct meaningful and useful data quality evaluations, business rules must be first completely identified and captured at the beginning of the evaluation to perform sound measurements. We propose that the evaluation leads to better and more interpretable and useful results when the potential contribution of these business rules to the measurement of the data quality characteristics is first evaluated, avoiding the inclusion in the evaluation of those not having potential contribution and the resulting waste of resources. Considering this, we feel that for a better management of business rules for data quality evaluation, it makes sense to group all business rules having an important contribution to the evaluation of data quality characteristics, something that other business rules management methodologies have not covered yet. Through our experiences in conducting industrial projects of data quality evaluations we identified six problems when collecting and grouping the business rules. These problems make data quality evaluation processes less efficient and more costly. The main contribution of this paper is a methodology to systematically collect, group and validate the business rules to avoid or to alleviate these problems. For the sake of generalization, comparability, and reusability, we propose to do the grouping for data quality characteristics and properties defined in ISO/IEC 25012 and ISO/IEC 25024, respectively. Lastly, we validate the methodology in three case studies of real projects. From this validation, it is possible to raise the conclusion that the methodology is useful, applicable in the real world, and valid to capture and group the business rules used as a basis for data quality evaluation.