The shortest distance between two points is not always a straight line. How often do we see policy statements from ministers and others which are self-evidently simplistic and unrealisable yet clearly seemed to make great sense to them when announced? Moreover, the unintended consequences of decisions can linger on as poisonous legacies for generations and become the basis for worse conditions than the ones they were intended to alleviate.

Everyday examples abound: quantitative easing (to help growth and reduce unemployment) vs. inflation; welfare systems that can remove the incentive and habit of work across whole communities; ‘light touch’ regulation (to stimulate entrepreneurship and investment) vs. excessive leverage and risk taking.

Being able to explore the consequences of decisions before they are made and sharing the debate with other stakeholders dramatically improves the chances of a good sustainable decision that, once made, stays and improves. If people can reach agreement on objectives and accept a proposed set of priorities as reasonable, the debate then moves forward to impact, consequences, options and trade-offs. Many of our most pressing problems fall into this category and seem stuck as the various parties adopt intransigent positions when a feasible set of outcomes that benefit all may well be within reach.

The advantage of mathematical modelling systems lies in the speed with which policies can be examined, the vast number of possible conditions which can be explored and the power of the visualisation that provides the key to greater insight and understanding of the ‘landscape’. Catastrophic but rare events can be foreseen and the combination of unusual circumstances that led to them can be isolated and dealt with as part of the initial planning thus overcoming a human tendency to assume that as a dangerous event has still not happened then it is increasingly unlikely!

Powerful statistical analysis of large data sets can also be a very fruitful source of insight into such areas as: which types of individual are likely to perform best (or worst) in certain roles e.g. as sales people or as security threats. With a preferred profile, better decisions can be taken on hiring, role fitting, and selection for further development.

As seen very recently, however, reliance on models without proper understanding of the assumptions underlying their safe use can easily lead to quite unacceptable decisions being made by overconfident ‘experts’ on whom others have come to rely. Stress testing, regulations on capital adequacy, assumptions on sovereign risk using statistical correlations and the assumption of normally distributed behaviour – possibly acceptable within an individual firm – led to wildly inaccurate signals when large numbers of firms began to behave in a coordinated way.