Precision and Accuracy: What's Sufficient?

Project performance (i.e., on schedule, on budget, and meeting the technical specifications) is important throughout the project's life cycle. Collecting and measuring performance information are critical for knowing how well the project is progressing. Tracking, reporting, and responding to project performance is critical to controlling the project's progress. The information needed to report on performance areas are typically collected at various times during project execution. However, the collection of information is based on a plan, not a random selection of data as would be done in a sampling technique. Building this collection plan is typically accomplished in a reverse order: 1.Identifying the results desired. 2.Selecting the collection point. 3.Select the type and quantity of information to be collected. There is a difference between "data accuracy" and "data precision." "Accuracy" refers to the data and how closely it represents the actual measurement. "Precision," however, relates to the degree of granularity of a measurement. Data may be: -accurate but not precise -precise but not accurate, or -precise and accurate. During data-collection planning, you must understand your real needs for precision (the fineness of measurement) and accuracy (the opportunity for collecting valid data). Measuring projects with more granularity than is needed can cost more money without giving adequate return. To achieve the right level of granularity, strive for the precision and accuracy that allow you to identify early indicators of problems so that you can quickly correct or better yet prevent the problems. Also consider the consequences of being wrong. For example, think of a time when you (or someone you knew) spent an hour or more trying find the source of an error in your bank account. If the error was minor (such as the cost of bus fare), your time probably was misspent. You would have had a better return for your time if you had "corrected" the error by adding an entry to your account, "Adjustment for minor, untraceable error." But a significant error such as the cost of a car payment would have serious consequences. Finally, also remember that establishing tolerance ranges for precision and accuracy can be as useful or more useful than actual data points.