Predict with Precision
2 min Read
Probabilistic Thinking in Financial Modeling
Financial professionals can gain valuable insight through probabilistic scenario analysis. As a powerful analytical tool, it helps inform the decision-making process by identifying and quantifying the full field of future possibilities. However, sometimes the “spread” of what is possible is too wide, and it is necessary to identify the outcome that best actualizes an organization’s goals.
One way to accomplish creating an ideal prediction is by understanding your best, worst, and expected case scenarios. The best and worst cases create predictive bounds to work within, and the expected case is often the median of the two. The ideal prediction can be determined through probabilistic thinking.
Charles Duhigg’s book on the science of productivity, “Smarter Faster Better,” discusses ways to make better decisions. He argues:

In its simplest form, determining a probabilistic prediction can be accomplished by first assigning odds to each scenario, and then using a weighted average calculation to combine each possibility.
As an example, let’s say that a business is determining scenarios for the launch of a new product. The CFO expects a steady growth in product sales after launch. However, the CFO also needs to set the worst case of a terrible product launch and the best case of an explosive product launch.

To combine and create a single projection, consider the likelihood of each scenario occurring. If there is no basis for assigning odds, make them equal to one another. If, however, one scenario is more likely than the others, assign probabilities accordingly.

A fourth probabilistic scenario, the combined case illustrated in the following graph, can be set up and graphed simultaneously with the expected, best and worst case scenarios. At the end of year, evaluate which assumption was most accurate.

Incorporating probabilistic strategies into everyday decision-making is relatively easy to do and, with practice, business officers can better anticipate the future, which will lead to more thoughtful planning at an institutional level. This allows modelers to move beyond arbitrary forecasting assumptions and focus on realistic drivers that will move the needle in a meaningful way.