No cure no pay: How to contract for software services

of all software projects are total failures and another 40% are partial failures according to widely quoted surveys in UK, USA and Norway. Large government projects in all 3 countries have been reported with spectacular failure and expense to taxpayers (Royal Academy of Engineering and British Computer Society 2004). What is the problem? Most discussions have centered on improving the software engineering process itself: better estimation, better requirements, better reuse and better testing. No doubt all those can be improved. However, I suggest the motivation to improve them needs to be put in place first. Think about it. Most of these failures have been fully paid for! We not only pay well for failure, but the bigger the failure, the more people get paid! My suggestion is simple. Pay only when defined results are provably delivered. This requires several things: ∞ Contracts that release payment only for meaningful results; ∞ The ability to define those results, particularly qualitative ones, and particularly the organizational ones; ∞ The ability to deliver those results incrementally, thus proving capability at early stages and continuously. Note: This paper specifically addresses the software problem, but I am sure that the ideas here apply to the wider systems engineering problem to some interesting degree as well.