Pay Serious Attention to Data Quality

The heightened interest in business analytics is quite encouraging. It indicates that organizations recognize they need to make informed decisions based on facts -- not just opinion. Being better armed with facts means organizations can gain greater insight into what's going on and how to better manage risk. This is perfectly aligned with what has been advocated by effective risk management practices and assurance activities such as compliance and internal auditing. So while being reassured there is a resurgent interest in analytics, this in itself can expose an organization to additional risk -- the risk of relying on inconsistent, incorrect or incomplete data.

Data quality is a critical prerequisite to effective business analytics. Poor data quality jeopardizes the performance and efficiency of operational systems. It undermines the value of analytic and business intelligence systems upon which organizations rely to make key decisions. Decisions based on poor data can result in direct financial loss, mar customer relations and damage an organization's credibility in the marketplace. So as organizations move to embrace business analytics -- or even more acutely, predictive analytics -- business leaders need to pay serious attention to the accuracy, quality and reliability of their data.

So what exactly is meant by data quality and what are the causes of data "corruption" that jeopardize the quality of data? Unfortunately, the causes are numerous and take many forms -- ranging from simple data entry errors and poor data element definitions to faulty programming logic for data entry or migration between systems.

The most obvious cause for poor data quality is data entry. If an organization has no standards or IT controls for how data is entered into a system, the data will quickly reflect their lack. It is in this way that duplicate entries are made in master data, such as the Vendor Master file. Incomplete information can also result, for instance, where no vendor contact information exists. Simple data entry errors can occur; behold the simple typo. Invalid data can creep into key information systems such as the Employee Master File: Don't know the social insurance number for your new employee? Why not enter a series of "9's" until you get around to correcting it later (or not)?

Data decay is another such source. List brokers know data in simple name and address decays at a rate of 25 percent per year -- or around 2 percent per month. By not keeping data up to date, is your organization sending invoices to an incorrect address? Are corporate credit cards still active for personnel no longer employed with your organization?

By way of example, we can look at one health care organization that knew they had a problem with their Vendor Master data and their whole Purchase-to-Pay business process. They learned they were frequently making duplicate payments and often made payment errors. When they took a closer look at their internal control environment, they realized that it was weak and one of the primary causes of their poor payment performance.

They turned to their newly minted Internal Audit Manager who understood the risks associated with poor data quality -- especially in master data files, since so many key business decisions are made around information gleaned from them.

The process the audit manager undertook used audit analytics to prove out the integrity of the organization's master data to seek out errors, anomalies and indicators of potential risk and fraud. He considered such master data as "ground zero" for the risk of fraud, error and abuse.

Initial investigations into the data proved out the following from their Vendor Master file:

• 23,778 active vendors
• 135 identical vendor names appearing 314 times
• 571 instances of a single address appearing multiple times
• 430 duplicate tax ID numbers for 145 vendors
• 5,941 vendors with no Taxpayer Identification Number (TIN)
• 11,497 vendor records with no vendor contact details

If this organization hadn't turned toward its internal audit department, they would have assumed they had almost 24,000 vendors. In reality, once the duplicate, triplicates and inactive vendors had been removed, it became apparent the organization had significantly fewer vendors than thought. In fact, they had only 5,495 active vendors. This insight resulted in a 75 percent reduction in the overall size of the Vendor Master file. This "data cleansing" activity resulted in significantly improved practices in their AP department and in the overall timeliness and accuracy of their procurement activities.

The main point here is that business analytics can provide amazing insights into how an organization is operating -- in hindsight, with insight and with foresight. But one must be attentive to the quality of the data being analyzed and put first things first. Step one is to check the validity of the data, ensure its quality and completeness. Step two is to ask those key questions that help provide the information needed to make informed decisions. Internal auditors, armed with analytic technologies of their own, can provide a huge amount of assistance in determining data quality and addressing the risk of drawing incorrect conclusions based on bad data.