Customer Data Integration: Reaching a Single Version of the Truth

DATA QUALITY: JOB NUMBER TWO

The problem with data quality used to be that no one knew it was a problem. Those days are long gone. The new problem with data quality is that everyone now wants to roll up their sleeves and start fixing the data. Usually, this means writing code or talking to vendors.

Not so fast. Though we’re big fans of data quality automation, we often see companies rush toward automation without first recognizing the flaws in their data. After all, data quality is a business problem and it’s not enough to know that data is missing, contradictory, or doesn’t make sense. Businesses need to understand the “desired outcome” of good data. They should ask themselves the following questions:

• What can’t we do as a business because our data is bad?

• Does the lack of standards for customer data impede our ability to sell additional products and services?

• How will improving our customer data improve revenues?

• What are our data acceptance criteria?

• How will improving our customer data increase operational efficiencies?

• Who’s responsible for fixing the data?

• How “good” does the data have to be? What are the metrics of good data?

This last point is an important one. Since operational systems are rarely responsible for their data quality, understanding how data is used will inform its quality metrics. Typically, operational data is “good enough” to support operational processing. However, data used for strategic purposes—like analyzing customer behaviors—requires ...