Scenario based planning

Scenario based planning is a very popular tool today, much used and much misused. It is a very valuable tool when correctly applied. First, to properly understand when it can and when it cannot be used, we must distinguish between two situations.

The problem of finding good decision alternatives.
The problem of evaluating existing decision alternatives.

It is the first use that is dangerous, but still very popular. The basis is the incorrect assumption that if you wish to find the decision that maximizes expected profit (or expected utility or whatever you wish to do), you can learn about this decision by studying individual scenarios. Almost all textbooks in operations research tell us to use sensitivity analysis (what-if-analysis) to study optimality when some parameters are not fully known. Sensitivity analysis is just another word for scenario analysis, and there is no theoretical basis for using this approach. There is no clear connection between what is best on average, and what is good in individual future situations. In my article on sensitivity analysis, I explain in detail why scenario analysis cannot be used to find good candidate decisions. The result of such an approach can be arbitrarily bad. So the use of scenarios (or sensitivity analysis) to see what happens under uncertainty is just operations research folklore, with no theoretical basis. Good candidates cannot generally be found by studying individual scenarios.

Evaluating existing alternative decisions is a different matter. Using scenarios to see how the different possible decisions behave under different assumptions about the future is indeed the correct way of comparing alternatives. This is in principal the same as simulation, a term less popular than scenario analysis these days. Note that scenario analysis (simulation) can be used to say which of a number of candidate decisions (strategies) is best, but not how good it is. There may be other decisions that are arbitrarily much better, we cannot know from scenario analysis.

For scenario analysis to make any sense, it is important that scenarios are defined such that they are independent of our decisions. The purpose of scenario analyses is to find out what we ought to do under varying environments. If we start to define scenarios that are dependent on our own decisions, we are very likely to get lost in a very advanced way.

A major issue in strategic planning is to identify scenarios, that is, to identify which random events are most important for us (or our company). Picking the right ones and dropping the unimportant ones is what makes the basis for a good strategic plan. And when you observe the major uncertain events, remember that they may all be (statistically) dependent. Forgetting dependencies may lead to disasters. Alternatively, we may simply require random variables used in scenario analyses to be independent. If two random events are dependent, most likely, there are some more basic random variables that are independent. They should then be used.