The prediction and control of project duration: A recursive model

Most large projects overrun. This indicates that current planning and control systems are inadequate. This paper suggests that the problem with planning systems such as PERT is that they assume that activities are independent, and that project activities occur in a linear sequence. In practice, projects behave as recursive networks, with reversions to earlier stages for rectification or to incorporate changes. The probability of a reversion to an earlier state is claimed in this paper to be the main factor determining project length. This kind of recursive network can be modelled as a Markov process. This paper uses a simple model to show that Markov analysis can represent the behaviour of real projects better than PERT.