Process Improvement Economics

Although this report contains general industry data, each company needs to create an individualized plan and budget for its improvement strategy. Table 18.1 presents information based on the size of companies in terms of software personnel. The cost data in Table 18.1 is expressed in terms of "cost per capita" or the approximate costs for each employee in software departments. The cost elements include training, consulting fees, capital equipment, software licenses, and improvements in office conditions.

Table 18.1. Process Improvement Expenses per Capita

The sizes in Table 18.1 refer to the software populations, and divide organizations into four rough size domains: fewer than 100 software personnel, fewer than 1,000 personnel, fewer than 10,000 personnel, and more than 10,000 which implies giant software organizations such as IBM, Accenture Consulting, and Electronic Data Systems (EDS), all of which have more than 50,000 software personnel corporatewide.

As Table 18.1 shows, software process assessments are fairly inexpensive. But improving software processes and tool suites after a software process assessment can be very expensive indeed.

It sometimes happens that the expenses of process improvement would be high enough so that companies prefer to bring in an outsource vendor. This option is used most often by companies that are below average. If a company is far behind similar companies, then turning over software development and maintenance to an outside company that already uses state-of-the-art processes and tool suites may make good business sense.

Another important topic is the time it will take to move through each stage of the process improvement sequence. Table 18.2 illustrates the approximate number of calendar months devoted to moving from stage to stage. Smaller companies can move much more rapidly than large corporations and government agencies. Large companies often have entrenched bureaucracies with many levels of approval. Thus, change in large companies is often slow and sometimes very slow.

For large companies process improvement is of necessity a multiyear undertaking. Corporations and government agencies seldom move quickly, even if everyone is moving in the same direction. When there is polarization of opinion or political opposition , progress can be very slow or nonexistent.

Table 18.2. Process Improvement Stages in Calendar Months

An important question is, what kind of value or return on investment will occur from software process improvements? Table 18.3 shows only the approximate improvements for schedules, costs, and quality (here defined as software defect levels). The results are expressed as percentage improvements compared to the initial baseline at the start of the improvement process.

The best projects in the best companies can deploy software with only about 5% of the latent defects of similar projects in lagging companies. Productivity rates are higher by more than 300%, and schedules are only about one-fourth as long. These notable differences can be used to justify investments in software process improvement activities.

As can be seen from this rough analysis, the maximum benefits do not occur until stage 5, when full software reusability programs are implemented. Since reusability has the best return and greatest results, our clients often ask why it is not the first stage.

The reason that software reuse is delayed until stage 5 is that a successful reusability program depends on mastering software quality. Effective software quality control implies deploying a host of precursor technologies such as formal inspections, formal test plans, formal quality assurance groups, and formal development processes. Unless software quality is at state-of-the-art levels, any attempt to reuse materials can be hazardous. Reusing materials that contain serious errors will result in longer schedules and higher costs than having no reusable artifacts.

Table 18.3. Improvements in Software Defect Levels, Productivity, and Schedules