A New, Rigorous Taxonomy of Data Behavior and Quality

Data quality practitioners are often ambiguous in their definitions of “dimensions” or “characteristics” of data behavior and its quality. This presentation introduces a more rigorous and granular taxonomy of data behavior. We consider the behavior or attributes of data at the cell, row, column, table, and database level in our evaluation of “raw” data or facts (at the grain as it was captured or created). We then look at the process of deriving information from raw data through aggregation, normalization, and integration.

This presentation will include numerous practical examples of each characteristic, including anomalies and errors. In this session you will learn:

New contextual terminology for labeling data errors
DQ characteristics specific to each kind of data
The timely delivery of decision-able information is not a DQ issue
How to describe observed DQ problems with greater clarity

Michael Scofield has been an Assistant Professor in Health Information Management at Loma Linda University in the Department of Health Information Management. He is a frequent speaker to a variety of professional and general audiences on topics of data management, data quality, data warehouse design, and data visualization.

His career has included education and private industries in the areas of data quality, decision-support systems, data warehousing, and data management. His articles appear in DM Review, the B-Eye Newsletter, InformationWeek magazine, the IBI Systems Journal, and other professional journals. He also has humor published in the L.A. Times and other journals.