A View of the Best

Further examination of estimation practices holds out an interesting process-improvement carrot. Most organizations find that as their projects become larger, each team member becomes less productive. In contrast with the economies of scale experienced in other kinds of work, software projects usually experience diseconomies of scale.

Organizations that use systematic estimation practices use formulas like this one to estimate their software project effort:^[14]

Effort = 2.94 x KSLOC ^1.10

Effort is the number of staff-months and KSLOC is the estimated number of lines of code in thousands. The numbers 2.94 and 1.10 are derived by calibration using data from projects that the organization has already completed. The values of 2.94 and 1.10 apply to typical organizations. The specific value of the exponent (1.10) is significant because it means that larger projects require disproportionately more effort than smaller projects.

NASA's Software Engineering Laboratory (SEL) is a notable exception. The SEL was the first organization to receive the IEEE's award for software process achievement and is one of the most sophisticated software development organizations in the world. The SEL uses the following formula to estimate effort on its projects:^[15]

Effort = 1.27 x KSLOC ^0.986

In spite of its small type, the exponent of 0.986 points to a momentous difference between the SEL's formula and the formulas used by every other software organization. Every other published estimation model uses an exponent greater than 1.0. The fact that the SEL's exponent is less than 1.0 suggests that the SEL is achieving a slight economy of scale. The process improvement carrot this represents is that sufficiently mature organizations might be able to turn the large-project problem on its head. They might actually be able to improve per-person productivity slightly as their projects grow larger. Although rare, this is indeed a logical consequence of the specialization I discussed in Chapter 10.