Chapter 6: Methods for Cost Analysis

5.6 Software Effort

Software effort is the number of hours required for software development. It the number of hours needed for analysis, design, coding, and testing. Software effort is a unit of time, usually staff hours or staff months, which are an estimate of software development time. Software effort is the time required to analyze, design, develop, and test a software product. It does not usually include software maintenance effort, which must be estimated separately. For instance, the software effort based on Boehm's model is 39.55 staff months for 10,000 lines of code. Likewise, the software effort for the Walston/Felix, Bailey/Basili, and Doty models is 42.27, 15.81, and 58.82 staff months, respectively. The average effort of these models is 39.14 staff months or 6,783.85 staff hours to develop 10,000 lines of code. The effort includes analysis, design, code, and test hours. Once again, 6,783.85 hours does not include maintenance. These models are old and were probably designed or calibrated using software projects with millions of lines of code. Table 1 displays the formulas for software effort.

Table 1: Formulas for Software Effort

Source	Model [*]
Boehm	Months = 3 — KSLOC 1.12
Walston/Felix	Months = 5.2 — KSLOC 0.91
Bailey/Basili	Months = 5.5 + 0.73 — KSLOC 1.15
Doty	Months = 5.288 — KSLOC 1.047
[*] KSLOC = thousand source lines of code.

Software effort plays an important role in the analysis of costs and benefits and ultimately the ROI of SPI. Software effort models are used to estimate a portion of SPI costs, namely software development effort. There are other costs, such as training costs, maintenance costs, and the costs associated with using a new software process, and there are costs for special tools, special methods , and appraisal costs. What is important is that you use a software effort model that is right for your situation. It may be a commercial off-the-shelf software effort model, a bottom-up software effort estimate, or a parametric equation such as a linear or log-linear model designed especially for your data and environment. Let's assume that 25% of the software effort from these models is software testing. Deduct that in order to estimate software testing costs separately. Therefore, the average effort is 5,087.89 or 5,088 staff hours instead of 6,783.85. We will use this as a basis in future software effort estimates.