Insights from Software Estimation

The accuracy of an organization's estimates is a good indicator of how well it manages and executes its projects. A Standish Group survey of more than 26,000 business systems projects found that the average project overran its planned budget by more than 100 percent.^[11] The level of estimation error reported in this survey is consistent with other industry findings.^[12]

Figure 13-3 shows one finding from a study of United States Air Force projects at different levels of software practices. (This analysis is based on the SW-CMM, which I'll discuss more in Chapter 14.) Each point below 100 percent represents a project that finished under budget. Each point above 100 percent represents a project that overran its estimate.

Figure 13-3. As organizations improve their software practices, they gain more control over their project estimates, which is generally indicative of better project control.^[13]

As you can see from Figure 13-3, the least sophisticated projects (SW-CMM Level 1) routinely overran their projects' budgets in other words, they routinely underestimated their projects' costs. More sophisticated organizations (at SW-CMM Level 2) spread their estimation error more evenly between overestimation and underestimation, but the relative error of the estimates was still commonly 100 percent or more. For the most sophisticated organizations (those at SW-CMM Level 3), overruns and underruns become equally common, and the estimation accuracy became much improved.