The Many Paths to Data Quality

September 2006 AMR Research Report Data quality is an elusive goal for most companies because it is treated as a one-time event. But The Many Paths to since it is critical to getting value from enterprise Data Quality applications, companies need to pick a strategy for measuring data quality, cleansing data on an ongoing basis, and possibly enhancing it with outside information to meet their business objectives. by Bill Swanton © Copyright 2006 by AMR Research, Inc. ® AMR Research is a registered trademark of AMR Research, Inc. No portion of this report may be reproduced in whole or in part without the prior written permission of AMR Research. Any written materials are protected by United States copyright laws and international treaty provisions. AMR Research offers no specific guarantee regarding the accuracy or completeness of the information presented, but the professional staff of AMR Research makes every reasonable effort to present the most reliable information available to it and to meet or exceed any applicable industry standards. AMR Research is not a registered investment advisor, and it is not the intent of this document to recommend specific companies for investment, acquisition, or other financial considerations. The Many Paths to Data Quality by Bill Swanton Companies need to pick a strategy for measuring data quality, cleansing data The on an ongoing basis, and possibly enhancing it with outside information to Bottom meet their business objectives. Line Data quality is an elusive goal for most companies because it is Executive treated as a one-time event. No matter how well they do, the data Summary begins to decay immediately. Quality data is important to getting value from enterprise applications, since up to half of the trouble tickets logged by help desks are traced to master data errors. A strong data quality pro- gram is also critical to success for strategic sourcing, customer management, and global data synchronization (GDS). As part of our ongoing research on master data management (MDM), AMR Research has interviewed dozens of global companies on how they cleanse, enrich, and measure the quality of data. Here’s what we found: • Global ERP rollouts and emerging implementations of service-oriented architectures (SOAs) are forcing consolidation of data quality initiatives. • Measuring data quality requires business rules. • There are six major architectural approaches to using data quality software and services. • Data quality and enrichment often require a combination of vendors. The right data quality solution for your company is highly dependent not only on your needs, but on the existing data architecture it is meant to serve. Approach data quality as a part of your overall MDM strategy, starting with documenting your data architecture. AMR Research Report | September 2006 © AMR Research, Inc. 1 Global ERP rollouts and emerging implementations of SOA are forcing consolidation of data quality initiatives Past data quality initiatives were often one-time or annual events, but that doesn’t work anymore: • The one-time data cleanup done for an ERP implementation or consolidation is recurring more frequently as the pace of mergers and acquisitions increases. A new company’s data needs the same cleansing and deduplication. • Spending analysis for strategic sourcing has shifted from a yearly event to continuous tracking of vendors and categories to enforce spending policies. • More rapid new product introductions mean more new materials, suppliers, and products. • Quarterly customer analytics are giving way to immediately identifying existing customers at point of interaction (web, store, or call center). • Data quality decays rapidly. Dun & Bradstreet (D&B) reports its database of businesses experiences annual changes of 20% for addresses, 17% for business names, and 18% for phone numbers. Data quality also becomes more critical as companies move to single global instances of ERP and other enterprise applications. Global scope means more individuals are responsible for entering master data, and companies are tracing the root cause of approximately 50% of help desk calls for these applications to master data errors. Even worse, the business intelligence or data warehouse group is fighting these same data quality problems, resulting in duplication of effort. Compliance is data driven As companies rely on enterprise applications as the cornerstone of their compliance strategies, the software relies on accurate attribute or demographic data about custom- ers, vendors, and materials. Many of these attributes, such as a vendor’s minority busi- ness status or a hazardous material classification, need to come from or be verified with external data sources. Costs for information services from companies like D&B can skyrocket if multiple business units each pay for the same data. SOA bringing convergence of EAI and data integration The next looming trend affecting data quality is the emergence of SOAs. As companies build extended, cross-functional business processes with SOA, data consistency and quality across linked applications become more critical. In addition, these processes often require data quality to be checked in line. For example, an order management process may need to verify immediately that a new customer’s order address is deliver- able, making periodic batch data quality processes much less effective. 2 © 2006 AMR Research, Inc. AMR Research Report | September 2006 Measuring data quality requires business rules When we ask a group of clients how they measure data quality, the most common response is “emotionally” (translation: it stinks). Like all process improvement efforts, data quality requires a quantitative methodology, such as the Six Sigma DMAIC (Design, Measure, Analyze, Improve, and Control), to take the emotion out of the process. This not only allows a rational analysis of the source of quality problems, but makes it easier to assign accountability to organizations and individuals for steward- ship of their data. A crude indicator of data quality is to count the number of adverse incidents caused by master data issues. These might include the following: • Number of help desk tickets with master data as root cause • Number of application exceptions requiring manual intervention (for example, an MRP run that failed) • Number of customer complaints (such as noncompliance letters for data synchronization) The issue with these measures is that a costly problem has already occurred. A better approach to measuring data quality is to see how well it fits the definition of master data management originally published in “Master Data Management Framework: Begin With an End in Mind”: Master data management is a system of business processes and technology components that ensures information about business objects, such as materials, products, employees, customers, suppliers, and assets, is current, consistent, and accurate wherever it is used inside or exchanged outside the enterprise. This approach requires that companies create a master data architecture, explicitly doc- umenting the business rules for data and stewardship. Data quality is then measured by how many of these rules are followed or violated. AMR Research Report | September 2006 © AMR Research, Inc. 3 Business-rules-based approach Figure 1 shows the rules-based approach, embodied in several of the data quality and data profiling tools available today. Business rules are established, and data from multi- ple sources is checked for consistency with the rules, which might include the following: • Consistency across applications • Validation against a list of legal values or ranges • Checks for required attributes • Complex calculations for reasonableness for dimensional or weight attributes These checks are made as part of data integration processes or as a background process.