Monday, June 15, 2015

Do You Look Backwards at Data Quality?

Image Credit: Dilbert.com

Perhaps you should.

If data is neither correctly entered nor maintained, the data output will prove to be useless, misrepresenting, and perhaps even risky.  When an organization uses bad data for external agency reporting or to allocate advertising dollars it creates legal risk, as well as potentially wasting precious resources such as staff time and budget.  One of the common definitions of data quality makes this point clear...data are of high quality if they are "fit for their intended uses in operations, decision-making, and planning" (Juran, 1995).  When organizational data quality is low it affects everything from strategic planning to constituent relationship management (CRM).

Implementing a CRM solution is a perfect opportunity for organizations and their business units to reexamine and streamline internal business processes for efficiency gains.  In any CRM system, the quality of data is established at three discrete points: data entry, validation, and ongoing data maintenance.

Entering correct data into the CRM is not as easy as it may sound.  According to a metastudy on errors in spreadsheet modeling, 11% - 91% of spreadsheet models include data or calculation errors at the time of creation (Panko, 2009). To address these initial data entry errors, organizations must implement robust data validation and review processes.  But even the best validation process won't catch errors that emerge due to changes in constituent data.  Think about how regularly people move or change jobs, mobile numbers, and email addresses. If this data is not maintained, we will quickly find our CRM system with data quality issues.

How then do you assure quality of data? You look backwards.

If you are experiencing data quality issues, look backwards and determine where the problem originates. At what point is data quality affected? At the very beginning when new data is imported into the system? Somewhere in the middle when data is transferred between systems?  Or both? If you never go back to the point of origin, you will never fully address and resolve the problem. 

Here are some key questions worth asking -

1. Is data reviewed / validated upon data entry? Data entry can mean the actual process a person takes when manually entering data directly into a system or it can mean the process in which a person imports data from an external source (e.g. purchased list).

2. Is a current business process contributing to data quality issues? How easy or hard is it for users to update their contact information? What are the roadblocks?   

3. Are there standards to how data is entered and maintained? Are these standards followed uniformly across the organization and business units?

Some data producers may argue that there is insufficient time or resources for data validation. This is true especially when resources are scarce.  The counter argument, however, would be to think about how much more time and effort will be necessary on the tail end to identify and correct bad data. Often times, such cleanup effort is much greater than the initial data input effort.

Co-authored by:

Jonathan See (CIO) and Dr. Michael L. Williams

Dr. Michael L. Williams serves as both the Associate Dean and professor of Information Systems at Pepperdine University's Graziadio School of Business and Management.

Juran, J. M. (1985). Managerial Breakthrough: The Classic Book on Improving Management Performance. New York: McGraw-Hill.

Panko, R. (2009).  Revising the Panko-Halverson Taxonomy of Spreadsheet Risks. Proceedings of the 42 HICSS. 
http://www.computer.org/csdl/proceedings/hicss/2009/3450/00/10-09-06.pdf

No comments:

Post a Comment