Make the Case for Better Quality Data

"It costs ten times as much to complete a unit of simple work when the data are flawed in any way as it does when they're perfect."

This simple rule of thumb provides a point of light on the topic that those of us in the data profession have not done well. We struggle to provide powerful business cases for improving data quality that command attention and focus the effort. Most senior executives are well aware that their data are not up to snuff. But absent a business case, data quality doesn't make the cut among competing priorities.

Of course, true leaders know this is no excuse. Given the growing importance of data to every aspect of their business, they know they must develop an understanding of the costs of poor data quality, shortcomings in the methodology notwithstanding.

These costs include direct but previously hidden costs, more important but intangible costs such as a loss of trust, and lost opportunity costs. This post aims to help leaders assemble an overall picture, starting with costs they can estimate and using these as a platform for understanding the unknown and unknowable costs.

First, develop a keen eye for non-value-added work. Customer billing is a great example. When a customer complains about a billing error, you spend time to research the correction and make good for the customer.

You spend a lot of time and effort, but you are no better off than you would be if you'd billed the customer correctly the first time. You've done a lot of work, but you can't charge extra for your troubles.

Working to find and fix errors is part and parcel of most operational processes, and once they turn a harsh eye on non-value-added work, most leaders can root it out. These costs are estimable. You simply isolate the non-value-added work, count up the people doing it, and apply the right load factors. Alternatively, you can measure the error rate and use the rule of thumb to get a first estimate.

Either way, the results are usually stunning. Even a very low overall error rate of 3% adds nearly 30% non-value-added costs. Numbers such as these make clear that the best way to reduce costs may well be to improve quality.

There is a lot of non-valued-added managerial work as well. It stems from the cold, brutal reality that most managers simply do not use data they don't trust: "These numbers don't look right. Let's verify them before we make this decision." They then work to verify or justify all the numbers.

In principle, one could measure the associated costs, but they pale in comparison to the costs of trying to manage when you don't know what's going on. You can't act quickly; you can't make good decisions; and you can't align people to the work. The lost opportunity costs are unknown and are almost certainly unknowable. Thus, the second key is to acknowledge that Dr. Deming's famous observation — "the most important costs (of quality in manufacturing) are unknown and unknowable" — extend to data as well. And though difficult, you must get your arms around these costs.

Let's go back to the billing example once more, as it also provides good examples of lost opportunity. Customers only complain about billing errors that hurt them, primarily overbilling. But if you're overbilling, you're almost certainly underbilling as well. Since customers don't complain about that, you've lost the opportunity to collect revenue you're due. Worse, these billing errors make it more difficult to build a long-term, trusted relationship with your customers; these lost opportunity costs are probably unknowable.

For many organizations, the most important impact of bad data will hit home with their big data efforts. Quite simply, bad data make everything about big data — from discovering something truly novel in the data, to building a product or service around that discovery, to acquiring needed support, and finally, to monetizing the new product or service — much more difficult. And no one knows how to quantify the costs associated with an industry-changing discovery that you could have made but didn't.

As you're making the case for better quality data in your organization, learn to apply the developing discipline of infonomics. While in everyday language, it is plain enough that poor-quality data are liabilities and high-quality data are assets, they don't appear on the balance sheet. Doug Laney, of Gartner, and others are building a body of practice that supplements today's accounting methods to help address this shortcoming. They're developing valuation models that help quantify both the economic and noneconomic values for data from a variety of perspectives. Pick the one that best suits your circumstance and apply it.

Lastly, recognize that not all data are created equal. Cut the effort down to size by focusing on your most important data, customers, operations, and strategies. Engage the head with solid estimates of the non-value-added work. Engage the heart with stories that bring the unknown and unknowable costs to life.

Mere managers find all of this daunting indeed. True leaders will acknowledge the difficulties but will also see opportunity — opportunity to understand what's really happening; to build support up, down, and sideways; and to launch the data quality program they know their organizations need.

Make the Case for Better Quality Data
Thomas C. Redman