By Elliot King
In the history of computing, no single mantra has been bandied about so much and with so little effect. It is intuitively obvious that if you are working with poor quality data, the results generated by using that data will suffer accordingly. And many people have an intuitive sense of the quality of their data.
The problem is most people have no more than an intuitive sense about their data quality. And while it is easy to make the argument that organizations should work to improve the quality of their data, it is more difficult to determine how much should be appropriately invested to that end, particularly if you can’t measure what poor data quality costs the organization.
Frankly, it’s difficult to determine how much time and money should be used to fix a problem if you don’t know its scope. The short answer is this–if you want to fix it, you had better be able to put a number on what it costs to be broken.
In the data quality arena, putting numbers to the problem seems difficult, but it doesn’t have to be.
The first step is to determine what hurts the most in your organization. Poor data quality can have an impact in many different ways. It can lead to inefficiencies in business processes. It can drive people to build redundant systems as safeguards and work-arounds so people do not have to rely on data they don’t trust. It can increase an organization’s risk profile. Or it can lead to bad decision-making.
The first step in measuring the cost of data quality is to identify where the organization is most vulnerable to being hurt if data is wrong. The second step is to determine a baseline. How good is your data now? Direct response is an easy, but always important, example.
What percentage of your messages does not reach the right recipients and what is the cost of the waste in that process?
The third step is to determine a reasonable goal. What is an acceptable amount of waste in your direct response campaign? Finally, what would you have to do to get from here to there? Once you have done the math, management will be able to do the math as well.
When calculating the cost of poor data quality, you should not look at data holistically. You have to understand the context within which it is used and move forward from there.