Better software reliability by getting the requirements right

For too long software has been produced using processes that result in a product that is completed late, over-budget, and below the quality expectations of end users. Over the last 25 years a great many incremental, evolutionary, and iterative software development process models have been introduced with the intention of improving software development; yet a recent study by the Standish Group reported that fewer than 28% of software projects are completed on schedule within budget and with most of the originally required features. We believe that getting the requirements right is the key to building successful and reliable software products. In the requirements process the product features are defined, the users are identified, and the user interactions with the system are designed. In addition, usage profiles are developed and estimates of the anticipated software reliability based on those profiles and on formal models of the system are obtained. Thus, reliability becomes an integral part of the requirement development process rather than an after thought of testing at the end of the development process. Although, defining the requirements may be an iterative process, a complete and correct set of requirements must be completed before starting to build the product