Making predictions was always part of the art of management. Now that we can readily predict multiple views of future performance it is easy to become concerned that the results will simply complicate decision making. Clearly the physical plant installed has a specific design and capacity but during the planning process good use can be made of an understanding of the range of possible future behaviour. Widespread access to detailed physical models and low cost high speed computing makes it possible to harness measures of uncertainty to the task of improving field development decisions of all types. Divestment or acquisition decisions are strongly affected, well location decisions can be improved and plant modifications or well workovers better chosen in scope and timing. 

As oil and gas resources are increasingly recovered from reservoirs whose behaviour has considerable uncertainty, the need to measure that uncertainty and educate ourselves in ways to make use of the measurements is increasing. It is now possible to combine new approaches to history matching and multiple predictions into measures of uncertainty for less than the cost of obtaining a simple single history match five years ago. When uncertainty is quantified, both understanding and economic outcomes can be improved, eventually making more hydrocarbons accessible by lowering cost. 

The talk will illustrate the difficulties and pitfalls that can arise when history matching and prediction are performed deterministically using practical examples. An approach that avoids some of the major difficulties of current history matching practice will be discussed and illustrated, showing how it integrates into the process of corporate decision making under uncertainty.