Support to effectively manage software development

Managing large software development projects using existing management tools can be a little imprecise. An altogether softer approach has been developed as a commercial tool, enabling project managers to get inside the development process and optimise resources. The drawback with conventional software project management tools is that they are unforgiving. There isn’t a means of ascribing levels of confidence or uncertainty and, as a consequence, the resulting answers may or may not be believable. The MODIST project management tool incorporates Bayes’ Theorem on Probabilities and the notion of conditional probabilities, and enables managers to input data in the form of probability distributions and histograms.

“The four key parameters when managing a large software development project are timescale, effort available, quality level and functionality,” says David Milledge of QinetiQ and MODIST coordinator. “These parameters can be determined to varying degrees of certainty. The variation in productivity of programmers, for example, measured in lines of debugged code per day, can vary by a factor of 20:1.”

The use of Bayesian Networks has many benefits compared to conventional statistical modelling. Statistical modelling, for example, produces a causal assessment on likely project duration and effort required. The Bayesian Network supports both deduction and adduction, and allows managers to explore the development process by posing ’what if’ questions, working back from the desired outcome, and observing the dependencies and risk factors.

“With MODIST, you can set up real-life situations,” says Milledge. “You might have to deliver software of a certain level of complexity with a development team of fixed size, and may need to know how long it would take to produce the software to a certain level of quality. Once the programme gets going, you can refine estimates and inputs, and the associated levels of confidence, and the outputs become progressively more precise. It makes risk assessments visible, and auditable, to the project management team.”

The prototype was available fairly early on in the project lifetime, which enabled it to be trialled in fifteen different software development projects. The tool has now been commercialised and is available from one of the partners, Agena, under the name of Agena Risk. A key market is for projects that need to make optimal use of resources spread across sites, time zones, organisations, and market sectors.