Dirty data—inaccurate, redundant or outdated information—has been the bane of many organizations struggling to use information for higher performance. Yet corporate-wide data cleansing seems so overwhelming that many businesses delay or defer even trying. Can a company live with dirty data? Yes, and no. By Royce Bell and Frank Dravis
Co-Authored with Business Objects Download this article [PDF, 121K]
PDF Help New perspective on an old problem Of course data quality matters. But dirty data isn't always bad—at least not equally bad. Stated another way, not all information has to be equally clean or has to be cleansed at the same time. What matters most is data that is central to achieving high performance. It's the data that, if managed incorrectly, will drag down business performance but, if managed correctly, has a powerful upside. Just which data is crucial depends on the industry, even the specific company. Consider capital markets, where reference data on customers, security and transactions makes up about 40 percent of the information used in every trade. This data typically comes from numerous internal and third-party sources, is rarely updated and is not generated as part of the transactions that a firm's operational systems are designed to handle. Yet poor quality reference data is an enormous cost that can drag down performance in the securities industry.¹ Setting priorities Knowing what data really matters, a business can set priorities for data quality improvements. Two examples: A telecommunications provider analyzed its revenue stream and decided initially to cleanse the data from the top 20 percent of its customer base, which generated more than 80 percent of its revenue. In doing so, the data quality project staff was able to reduce the scope of cleanup from 20,000 businesses to just 800. A manufacturer of high-tech equipment sells primarily through a direct mail catalog, which was printed and mailed four times per year, at a cost of $3 million for each mailing. Concerned with the low response rate, it focused on cleansing its customer file and found 50 percent redundancy in customer records. Eliminating that duplication saved $6 million per year—which it applied to price cuts that gave it a competitive advantage. With clear business priorities, a company can break the data cleansing task into manageable pieces. An initiative that might have seemed too daunting can instead gain traction, build momentum and have specific measurable impact. Three questions for the C suite Three key sets of questions can shape the nature, scope and timing of a data quality initiative. - What data is critical for our specific strategic goals? Suppose the broad objective is to increase sales and revenue. Does that mean more effective customer targeting? If so, the business will need an integrated data profile of the customer. Or if the goal is lower production costs, the need is for an integrated view of suppliers.
- What changes in data quality do we need to make—and measure? To create an integrated view of the customer, for example, the key might be reducing the percentage of duplicate or incomplete customer records. For an integrated view of suppliers, a key metric might be to reduce the instances of different supplier information from one system to the other.
- Can we undertake a comprehensive effort at this time? Some businesses may be ready to immediately address the most critical data. In other businesses it may be better to begin with limited initiatives or self-contained pilot projects that give a data quality team experience, build confidence and verify economic assumptions in the business case.
Who answers these questions? The top leaders. They manage the critical data collection processes. They understand the business risk of basing decisions on inaccurate data. And now, in the newly stringent regulatory environment, they are being held accountable for the information being reported. Getting to better data The process of answering those questions should focus a leadership team on their real data quality priorities and help them decide how, and how quickly, to proceed. Regardless of a company's specific plan and timetable, though, certain conditions are essential for success. Executive sponsorship Having an executive sponsor for data stewardship across core business functions is essential. The executive sponsor oversees the development of the data quality programs or plans; defining their scope; assigning roles and responsibilities; setting priorities; facilitating strategic decisions; establishing measurable and realistic goals; and developing a business case. Mobilization and collaboration CIOs and IT organizations understand the importance of data quality, and in some organizations have inherited the responsibility given their role in managing enterprise data. The executive sponsor needs to mobilize data champions from IT and elsewhere in the business, encouraging them to collaborate in defining an approach to address data quality problems for a given business process. Ideally they develop a series of data quality projects, which they review in detail with executive sponsors. Sustained support A crucial and often forgotten final success factor is ongoing assessment of data quality and support for the data management program. Without a long-term commitment to maintaining data quality and appropriate usage, a company faces covering the same ground again and again. Clean data does matter Improving data quality produces both tangible and intangible benefits, as executives reported in a data quality survey² . Most often cited were increased customer satisfaction, a "single version of the truth," greater confidence in analytical systems, reduced costs and less time required to reconcile data. Our own experience with clients confirms the real business value of a well-conceived data quality initiative. For example: A communications company sought to improve call center efficiency, enhance customer service and create a better one-to-one customer approach in downstream activities. Call center efficiencies alone created $800,000 to $1 million in annual savings. A health and life sciences client found that data cleansing generated value estimated at $2 million to $4 million per year. A United States telecommunications company was striving to meet the billing requirements specified by a major state. The solution encompassed a new electronic billing system and technology to standardize names and addresses. Annual savings due to reduced data rework activities, efficiency in billing operations and postal discounts exceed $150,000 per year. Data quality doesn't need to be a daunting, unreachable objective. By honing in on the truly critical data, then creating a focused and paced program for data improvement, companies can realize the value of cleaner data and higher performance. About the Authors Royce Bell
Royce is CEO of Accenture Information Management Services. Bell has 23 years of experience in management consulting, technology and outsourcing services. He has worked with a number of large multinational companies, notably Shell and BP, and several government institutions and has a wealth of international experience in the United States, Europe and Asia. Bell has held various responsibilities within Accenture, including leading national and regional industry programs and service lines as well as managing global client accounts. Frank Dravis
Frank is the vice president of EIM Strategies at Business Objects. He has 23 years of experience in information technology and software development, with the majority of his experience centered on solutions design and implementation. His background in primary research of information quality markets and technology trends has made him a frequent speaker at industry events on a broad range of technology topics. He publishes many articles and is often quoted in leading industry publications on the topic of data quality and data profiling.
¹ "Data Management—The Quest for Quality," Accenture White Paper, 2004.
² "Data Quality and the Bottom Line: Achieving Business Success Through a Commitment to High Quality Data," Wayne W. Eckerson, The Data Warehousing Institute, p. 10. Survey was conducted in 2001, and queried 647 individuals in a range of positions, industries and countries. To Top |