Scaling Up and Scaling Down

Post-gold rush software engineering practices have unequivocally proved their worth on large projects. (Doubters can turn to Chapters 13 and 14.) They also have a lot to offer smaller projects. Larry Constantine describes an Australian Computer Society Software Challenge in which three-person teams had to develop and deliver a 200 function-point application in six hours.^[2] This is a significant challenge, equivalent to writing about 20,000 lines of code in a traditional third-generation language or about 5,000 lines of code in a visual programming language.

A team from Ernst and Young decided to follow a formal development methodology a scaled-down version of their regular methodology complete with staged activities and intermediate deliverables. Their approach included careful requirements analysis and design. Many of their competitors dived straight into coding, and for the first couple of hours the team from Ernst and Young lagged behind.

But by mid-day they had developed a commanding lead. At the end of the day the team from Ernst and Young lost, but not because their systematic approach had failed. They lost because they accidentally overwrote some of their working files, squandering their afternoon work and delivering less functionality at the end of the day than they had demonstrated at lunchtime.

Would the team from Ernst and Young have won without the configuration-management snafu? The answer is yes. They reappeared a few months later at another rapid-development face-off this time with version control and backup and they won.^[3] Their success was achieved not by stripping down their earlier approach but by identifying weaknesses in their old process and making improvements.

This general value of applying systematic process improvement within small organizations has been confirmed by a Software Engineering Institute study.^[4] The success rate of process improvement programs in organizations with fewer than 50 software developers has been just as good as it has been in larger organizations. Moreover, small organizations have fewer of the problems that inhibited success in larger organizations, such as organizational politics and turf guarding.