An experiment to assess the cost-benefits of code inspections in large scale software development

摘要: We conducted a long term experiment to compare the costs and benefits of several different software inspection methods. These methods were applied by professional developers commercial product they creating. Because laboratory for this was live development effort, we took special care minimize cost risk project, while maximizing our ability gather useful data. The article has goals: (1) describe experiment's design show how used simulation techniques optimize it; (2) present results discuss their implications both practitioners researchers; (3) new questions raised findings. For each inspection, randomly assigned three independent variables: number reviewers on team (1, 2, or 4); teams inspecting code unit (1 2); requirement that defects be repaired between first second team's inspections. selected without replacement from pool 11 experienced developers. dependent variables included interval (elapsed time), total defect detection rate. Our showed these treatments did not significantly influence effectiveness, but certain combinations changes dramatically increased interval.