Optimizing the Contribution of Testing to Project Success

摘要

Let's define the Goal of development projects as: Providing the customer with what he needs, at the time he needs it, to be more successful than he was without it, constrained by what we can deliver in a reasonable period of time. Furthermore, let's define a defect as the cause of a problem experienced by the users of our software. If there are no defects, we will have achieved our goal. If there are defects, we failed. We know all the stories about failed and partly failed projects, only about one third of the projects delivering according to the original goal. Apparently, despite all the efforts for doing a good job, too many defects are generated by developers, and too many remain undiscovered by testers, causing still too many problems to be experienced by users. It seems that people are taking this state of affairs for granted, accepting it as a nature of software development. A solution is mostly sought in technical means, like process descriptions, metrics and tools. If this really would have helped, it should have shown by now. Oddly enough, there is a lot of knowledge about how to significantly reduce the generation and proliferation of defects and deliver the right solution quicker. Still, this knowledge is ignored in the practice of many software development organizations. Last year at the PNSQC, I presented a paper: How Quality is Assured by Evolutionary Methods, describing practical implementation details of how to organize projects using this knowledge, making the project a success. In this paper we'll extend these Evolutionary methods to the testing process, in order to optimize the contribution of testing to project success. Important ingredients for success are: a change in attitude, taking the Goal seriously, which includes working towards defect-free results, focusing on prevention rather than repair, and constantly learning how to do things better.