Effective Measurement of DQ/IQ for BI

Introduction

There has been a growing interest among researchers and industry practitioners in DQ, DQ assessment and improvement. However, the basic problem of DQ assessment and improvement still remains largely open. For example, Massachusetts Institute of Technology (MIT) has instituted Total Data Quality Management (TDQM) research effort which is grown from industry needs for high quality data. The objective of this program is to establish a solid theoretical foundation and devise practical methods for business and industry to improve DQ. The research scope of this program covers 3 areas i.e. 1) definition of DQ that addresses issues of data quality definition, measurement, and derivations; 2) Analysis of DQ Impact on Business that addresses the value chain relationship between DQ and business outcome and 3) improvement of DQ that addresses various methods for improving DQ.

Thus research interest in DQ remains an enduring subject, which is reiterated even by very recent literature. For example, Fehrenbacher et al., 2012 state that the importance of DQ is ever increasing and that research in this field focuses mainly on 2 aspects i.e. criteria and assessment. This recent work observes that while researchers have developed a number of frameworks, criteria lists and approaches for assessing and measuring DQ, still research in this discipline indicates that assessing DQ remains to be challenging. This work argues that although DQ is subjective, most of the existing frameworks and assessment methodologies do not often consider the context in which the assessment is performed. Through empirical data research this cited work suggests that the perceived importance of DQ criteria has changed over the last decade.

Heinrich et. al., 2009 observe that many DQ metrics are designed on an ad hoc basis and that they are often highly subjective and propose six normative requirements in order to enable a scientific approach and an evaluation of the metrics.

Direct references to DQ in BI Systems in published literature are limited. The same may be explored through references to related types of Information Systems such as DW, DSS, EIS or data mining. Similarly, the terms DQ and IQ have been used interchangeably in literature. There has been no consensus about the distinction between DQ and IQ – DQ may be referred to technical issues or quality in sources or ‘raw data’ and IQ may be referred to non-technical issues or quality in processed data stores or insights presented from data for decision making.

The objectives of this chapter are as follows:

• •

  Identify gaps that exist in measuring DQ for BI based on published frameworks;

• •

  Analyze existing DQ/IQ measurement frameworks in light of their applicability for BI;

• •

  Examine DQ requirements (normative requirements) applicable to DQ for BI;

• •

  Discuss at length a proposed framework and how it seeks to address the gaps;

• •

  Consolidate findings and directions for future work in the study of measurement of DQ for BI.

In summary, past literature shows that poor quality of data in a warehouse adversely impacts the usability of the warehouse and managing DQ in a warehouse is very important (Shankaranarayanan, 2005). The needs of DQ definition and approach for DW and/or Decision Support Systems are unique and different. However, most of the research work focused on this specialized type of information system, have approached with metadata model (integrating / extending existing metadata in a warehouse with quality-related metadata, which has associated practical problems in implementation) and have not focused on implementation aspects of the framework.

Background

There has been no consensus about the distinction between data quality and information quality. In this chapter, no distinction has been made and term data quality has been used to refer to the full range of issues.