For those of us without a functioning crystal ball, prediction (of events) and forecasting (of magnitudes) are pretty much a waste of time. Experts are particularly inept in this regard, as we have seen especially over the past several years. A blindfolded guess may work out better on average.
Dealing with uncertainty cannot therefore be based on a prediction or forecast. The best we can do in most cases is to try to figure out what might happen to us if a certain event occurs or an important variable exceeds a critical threshold.
This approach to uncertainty means that we don't have to worry about prediction reliability or forecast accuracy. What we do have to worry about is making a solid connection between the uncertainty and our organization.
We need make a short list of major uncertainties — what might come along — and determine each ones impact on our organization. Many may generate financial crises. Others may impact supply of essential materials. Some may disrupt critical communications and systems. Transportation may be disrupted.
Each of these stresses and damages the organization. Our planning job is to assess the potential damage, figure out how to minimize it, and sketch out plans for recovery.
We generally do this using specially designed systems models in which the event or magnitude is an input. The output is a dynamic view of how the organization might respond.
Once we have a good model that plausibly encompasses uncertainties that we have selected, our main job lies in making sure that the organization is robust and resilient should any of these come to pass. It must be able to absorb the impact and recover from whatever comes along.
The final step is tracking the identified uncertainties routinely to watch for indications that an uncertainty may be moving toward materializing.
These terms are commonly regarded as synonyms and used interchangeably. We use them differently:
Prediction refers to events. They happen or do not.
Forecast refers to levels or magnitudes.
For example, we may predict "serious" inflation, or dollar devaluation, or a terrorist attack, with some time frame. Events are binary — they either occur or not occur. They don't (or shouldn't) sort of occur.
On the other hand, we forecast next year's sales, interest rates, and materials prices. They are not usually associated with events.
So we can have both a prediction and a forecast. We might predict a dollar devaluation and forecast its magnitude at 10%. We can be right in our prediction but way off in our forecast.
This distinction may not appear important but it turns out to be quite useful in scenario modeling.