Abstract
We introduce a portfolio management method that uses effort estimates to build sets of feasible deadlines for software projects at the bidding stage. Effort estimates can involve considerable error, and this must be taken into account when selecting deadlines. We show how a simple probability model can allow for possible errors. The model is built using a single effort estimate for each current project, together with historical data on estimated and actual effort for former projects.
The probability model is used in two ways: firstly to find the probability of successfully meeting a set of proposed deadlines; and secondly to select deadlines that deliver a fixed probability of success. Rather than treating projects in isolation, we work with the full company portfolio, enabling setbacks in some projects to be balanced by gain in others. Our method is implemented with demonstrations in the tool PROJMAN.