Why: The technique is used when: 1) many variables are present and their interrelationships are unclear, 2) the system can't be analyzed by direct and formal methods; 3) building analytical models would be time consuming, complex, and just too hard, 4) you cannot do direct experiments, 5) when the input details such as equipment life and repair times are not discrete and they vary over time according to a distribution, and 6) you need to do some tweaking of the system to understand where opportunities lie for improving uptime, reliability, and costs.
When: Build models before you commit systems to bricks and mortar so you know their performance on paper. Revise the models after they are in operation to help improve the unknown weaknesses an improve costs for future cases.
Where: Monte Carlo models are used for gaining insight about how things work and data collected from the model is done at an accelerated rate compared to real life.
These definitions are written by H. Paul Barringer and are also posted on his web site at www.barringer1.com