Safe and Simple Software Cost Analysis

Barry Boehm, USC

“Everything should be as simple as possible, but no simpler. ”

- Albert Einstein

Overview

There are a number of simple software cost analysis methods, but they may not always be safe. The simplest is to base your cost estimate on the typical costs or productivity rates of your previous projects. This will work well if your new project doesn’t have any cost - critical differences from your previous projects. But it won’t be safe if some critical cost - driver has changed for the worse.

Simple history - based software cost analysis methods would be safer if you could identify which cost driver factors were likely to cause critical cost differences, and if you could estimate how much cost difference it would make if a critical cost driver changes by a given degree. In this column, I’ll provide a safe and simple method for doing this by using the recently - published cost estimating relationships in the book Software Cost Estimation with COCOMO II (by Barry Boehm, Chris Abts, A. Winsor Brown, Sunita Chulani, Bradford Clark, Ellis Horowitz, Ray Madachy, Don Reifer, and Bert Steece, Prentice Hall, 2000). COCOMO II is an updated and recalibrated version of the Constructive Cost Model (COCOMO) originally published in Software Engineering Economics (by Barry Boehm, Prentice Hall, 1981). I’ll also show how the COCOMO II cost drivers can be used to perform cost sensitivity and tradeoff analyses, and discuss how similar methods can be used with other software cost estimation models.

COCOMO II Productivity Ranges

Figure 1 shows the relative productivity ranges (PR’s) of the major COCOMO II cost driver factors, as compared to those in the original COCOMO 81 model.   The productivity range for each factor is the ratio of the project’s productivity for the best possible factor rating to the project’s productivity for the worst possible factor rating, assuming that the ratings for all of the other factors remain constant. Here, relative “productivity” is defined in either source lines of code (SLOC) or function points per person-month. The term “part” in Figure 1 reflects the fact that the COCOMO 81 development mode involved a combination of Development Flexibility and Precedentedness.

For example, suppose that your business software group has been developing similar applications for 10 years on mainframe computers, and their next application is to run on a micro - based client - server platform. In this case, the only cost driver factor that is likely to cause significant cost differences is the group’s “Platform experience” (assuming that such factors as “Platform volatility” and “Use of software tools” do not change much). From Figure 1, we see that the productivity range for the “Platform experience” factor is 1.40. This means that changing from the best level of platform experience (over 6 years) to the worst level (less than 2 months) will increase the amount of effort required for the project by a factor of 1.40, or by 40%.