Aug 08, 2016 Walt Wojciechowski
Businesses participating in any sector, whether fintech or manufacturing, endure the burden of collecting, managing and analyzing data. For those with data quality assurance plans, this responsibility produces a return in the form of cost reductions, quicker decision-making and the ability to quickly respond to market changes.
According to a study by Gartner, poor data quality costs the average organization $14.2 million every year. Addressing this type of loss entails implementing an enterprise-wide data governance plan. Such strategies are inherently designed to ensure the quality of all information organizations collect, store and process. There are three steps to achieving this goal:
"Settling on a format entails introducing rules and ensuring data adheres to them."
1. Define data formats
The first step in developing a data QA plan is to settle on which formats are appropriate for certain information. Most importantly, analysts process these data in distinct ways, so they cannot be set up in an identical manner.
Settling on a format entails introducing rules. If a piece of information meets the specifications dictated by these laws, then it is adequate for analysis. There are many tools available that enable businesses to apply and ensure data adheres to formats.
For example, credit reporting agencies (CRAs) use e-OSCAR, an internet-based, Metro 2-compliant automated system that helps CRAs address consumer credit history issues. The solution features intuitive error reduction functions, and can edit illogical entries. MicroBilt, which specializes in alternative credit data and develops business risk tools, uses this system when collecting and processing information.
2. Establish flow control
People change data all the time. While some may be aware of best practices, others may make apparently minor adjustments that hinder later analysis projects. To address this problem, Oracle advised that companies introduce flow control.
The software company outlined the purpose of this process, which is to verify incoming data according to the rules enforced by formatting standards.
For instance, say a business analyzes its customers' alternative credit data to determine the risk of lending to particular individuals. In order to make accurate decisions, the company specifies that all customer profiles must include any outstanding evictions, suits and Liens. If those items are listed "N/A" it signifies that the person in question is not facing any such penalties, but if left blank, it may leave assessors questioning.
3. Dictate accountability
Humans aren't perfect. Bitpipe ranked people as the No. 2 cause of data quality issues. People may enter information incorrectly, fail to safeguard their credentials or make some other kind of error.
The best way to address this problem is to determine which individuals can access and change specific data. It's best practice to develop these policies around individuals' roles: accountants can handle financial data, but not information in a customer relationship management system, for example.
Another factor to consider when developing a data QA plan is the third-parties you work with. Ensure they have experience collecting and managing large volumes of data - a sign that they will handle information in a manner that improves your operations.