By Elliot King
While assessment is obviously the first step, it should be just as obvious that it can't be the last. Data assessment indicates that problems exist in the data. The goal should be to consolidate the data generated through the data assessment process into issues that can be addressed. The data assessment process may show that 30 percent of purchase orders lack a customer ID and 15 percent have incorrect contact information. The quality issue is that your purchase order records are flawed.
Moving from issues to action requires a series of systematic steps and the first one is not to fix the flawed data. Simply fixing data errors is like bailing water on a leaky ship. No matter how fast you fix errors, more will be coming in.
Once data issues are identified, the first step is to determine their impact on the bottom line. It could be that the errors do little or no harm to corporate processes and need not be rectified. Unfortunately, that generally is not the case, but determining the impact of data quality problems is essential for guiding investment and priorities in fixing them.
The next step is to understand why these errors are occurring. Do front-end data entry screens have to be altered? Does the data incorporated from third-party databases fail to conform to your company standards? Do certain divisions of your organization ignore specific business rules? Are certain data transformations being executed inaccurately? Is there significant data decay?
Once the root cause of the data problems has been pinpointed, a suitable remedy can be constructed and then implemented. The final step is to monitor the remedy.
While this process seems straightforward, each step has several different directions from which it can be approached. The key, however, is to work systematically. If not, you will find yourself applying a Band-Aid when you really should be doing major surgery or vice versa.
The first step in a data quality program is to assess your data. Whether you opt for data profiling or some other assessment mechanism, this part of the process consists of systematically identifying exactly where the problems can be found in your data sets.
While assessment is obviously the first step, it should be just as obvious that it can't be the last. Data assessment indicates that problems exist in the data. The goal should be to consolidate the data generated through the data assessment process into issues that can be addressed. The data assessment process may show that 30 percent of purchase orders lack a customer ID and 15 percent have incorrect contact information. The quality issue is that your purchase order records are flawed.
Moving from issues to action requires a series of systematic steps and the first one is not to fix the flawed data. Simply fixing data errors is like bailing water on a leaky ship. No matter how fast you fix errors, more will be coming in.
Once data issues are identified, the first step is to determine their impact on the bottom line. It could be that the errors do little or no harm to corporate processes and need not be rectified. Unfortunately, that generally is not the case, but determining the impact of data quality problems is essential for guiding investment and priorities in fixing them.
The next step is to understand why these errors are occurring. Do front-end data entry screens have to be altered? Does the data incorporated from third-party databases fail to conform to your company standards? Do certain divisions of your organization ignore specific business rules? Are certain data transformations being executed inaccurately? Is there significant data decay?
Once the root cause of the data problems has been pinpointed, a suitable remedy can be constructed and then implemented. The final step is to monitor the remedy.
While this process seems straightforward, each step has several different directions from which it can be approached. The key, however, is to work systematically. If not, you will find yourself applying a Band-Aid when you really should be doing major surgery or vice versa.
Leave a comment