THE STRATEGIC IMPACT OF POOR DATA QUALITY (Part 2)

By Eugene Reagan
Senior Consultant

This is part two of a two-part article...

Overheard Conversation No. 2:
“Weren’t we going to start marketing annuities to our loan customers?” “We were, but pulling the data together turned out to be too much work.”

The Approach to Data Quality
How does data quality have this kind of impact? Why does it continue to bedevil major strategic initiatives in company after company? Most importantly, what can be done to address this ongoing and growing problem?

•  Accessibility
•  Accuracy
•  Completeness
•  Ease of manipulation
•  Objectivity
•  Relevance
•  Security
•  Timing

1. Define the dataset. Everyone involved in the analysis must have a clear understanding of the sources of the existing data and its uses in the proposed new environment. This may include all the data produced in support of a particular process or task, all the data produced from a single system or set of programs, or selected data that will be pulled from various systems into a single database or data warehouse.
2. Identify the customer audiences. Who will represent the information collectors, custodians, and consumers? These individuals will provide subjective and objective evaluations of the data quality, both now and in the future.
3. Develop analysis information. Objective measures of quality—such as error rates, incident reports, and response time statistics—should be used when possible. It is important to remember at all times that data quality problems are not necessarily systems issues. High error rates may be related to employee training or turnover, and data quality may never be acceptable until those problems are addressed. Frequently, objective assessments of data quality will be difficult to obtain.

   Therefore, subjective assessments must be developed through interviews and surveys of the different customer groups. This can be tricky, even when objective data is available. For instance, a 2 percent error rate may be satisfactory for the information collectors but absolutely unsatisfactory for the information consumers, who hope to use the data for specific, targeted purposes.

4. Identify and evaluate discrepancies. Identify the particular quality dimensions that are perceived as problematic by one or more of the survey audiences. Address these discrepancies as symptoms during the quality improvement process.

5. Understand the “metadata.” When the data is to be used in a new database or data warehouse, there must be an understanding of the metadata, the data about the data. These are the conventions or standards related to format and content that will be used in the new application. For each data element, these characteristics include:
   
   •  Name
   •  Definition
   •  Description
   •  Field layout
   •  Processing restrictions
   •  Source system or program
   •  Target system destination
   

6. Perform root cause analysis. Attack the major categories first. Focus on specific data elements that seem related to the symptoms previously identified. These problems may not be related to the format or content of the data elements themselves; they may be the result of training issues or process flaws that had little impact in the existing environment but cannot be tolerated in future activities.
7. Develop and implement solutions. Specific implementation plans must be developed to address all of the data quality issues that are identified. These plans should outline definite responsibilities and detailed timetables.

It is vital that insurance companies address data quality issues in an organized, aggressive fashion. Such an approach maintains the credibility of all system development efforts and will reduce the time and cost required to deploy a system. Most importantly, it enables the implementation of key business strategies, which leads to opportunities for revenue growth.