Organizations and governments worldwide will spend about $1 trillion this year on IT projects. Recent data suggested only about 35 percent of those projects are likely to be completed on time and on budget, with all their originally specified features and functions. Many projects, perhaps 20 percent, will be abandoned, often after multimillion-dollar investments—and the biggest projects will fail most often.

One well-documented $170 million software failure was blamed on a lack of defined requirements in the original contract; a lack of software engineering, program, and contract management skills; and underestimates of the complexity of interfacing the new system with legacy systems, addressing security needs, and establishing an enterprise architecture.

Other software-development failures have brought down entire companies, such as the $5 billion drug-distribution firm in Texas that declared bankruptcy as a result of a poorly implemented resource planning system. 

It doesn't have to be this way. The quality of a system is influenced by the quality of the process used to acquire, develop, and maintain it, the analysis and forethought that goes into an architecture that supports business goals and requirements, and the training provided to teams involved in the project. Using proven methods for process and product quality, software success is predictable and achievable, and failure is avoidable.

Once coding starts, teams trained in mature software engineering processes can remove defects early, when defect removal is 10 to 100 times less costly than it is during test. This dramatically reduces test costs and only marginally increases costs upstream.