Should We Apply Bayes’ Theorem to Scenario Planning?


What Is Scenario Planning?

Scenario planning is a method that many organizations use to envision alternative futures and then strategize around how they could adapt to each one. Most scenario planning (sometimes known as scenario thinking or analysis) results in the creation of a handful of different scenarios.

Scenario plans are targeted toward specific subjects that help organizations become more forward-thinking in regard to their products and services. “A good set of scenarios should always be customized to a particular context,” writes Jay Ogilvy, cofounder of the Global Business Network. “The scenarios that Royal Dutch/Shell used to anticipate the drop in oil prices in 1986 were far different from the scenarios a major computer manufacturer used to navigate its transition from products to services.”

Why Use Scenario Planning?

The idea behind scenarios is less to predict the future than to change people’s thinking about it, freeing them up to see alternatives to conventional thinking. This helps them plan for and react to changes in their business environment, which can include social, political, demographic and technological changes as well as business and industry changes.

Not only can good scenarios expand people’s thinking beyond the here and now, they can help them spot trends or issues they’re currently neglecting. And they can better identify new areas of opportunity.

How Does It Work?

Depending on your process, scenario planning can take anywhere from six to ten steps. Ogilvy advocates eight:

  1. Discuss the organization’s key concerns and then pick an issue to focus on. It could be something as specific as new technologies in the industry to something as large as the geopolitics of the nations crucial to supply chains.
  2. Use brainstorming to identify key factors, many of which will be internal to the company or industry.
  3. Use brainstorming again to look at the external forces—such socioeconomic or environmental issues—that could have a large impact on the future.
  4. Now start boiling things down in order to select the most critical uncertainties that have arisen via the previous two steps.
  5. Use a method to choose which scenarios to develop. My favorite method—and the one outlined by Ogilvy in Forbes—is selecting two major uncertainties and then using them as the axis in a 2-by-2 grid. I usually begin by creating a 3-by-3 table in Microsoft Word and wind up with something like what you see here:
Scenarios for Energy and Climate Change
Temperatures Rise Quickly Temperatures Level Out
Climate Agreements Grow Stricter/Stronger The Rush to Renewables The Global Reg Blowback
Climate Agreements Lack Teeth Complacency In the Face of Rising Dangers The Status Quo Rules


6: Let’s say we’ve decided to focus on energy trends and have determined that the two largest uncertainties revolve around the rise of global temperatures and the strictness of global climate agreements. Once the basic framework is established, the next step is to create narratives for each of these quadrants. The stories can be complex with twists and turns, if that’s what the group envisions.

7: Consider the implications of each scenario and the strategic options it warrants. This stage may take considerable time or be done relatively slowly. The main goal is to find a single strategic plan that will work in any of these scenarios, though that might not be possible. If not, then the firm may need to develop multiple contingency strategies.

8: The organization should not just file away the scenarios. Instead, strategists should look for early indicators and keep their ears to the ground. If things start moving in the direction of one of the scenarios (or in a direction not foreseen), then strategic decisions should be revisited as needed.


An Unauthenticated But Widely Used Portrait of Thomas Bayes

What Is Bayes’ Theorem?

Bayesian statistics are based on the theorem put forth by Rev. Thomas Bayes in the 1700s. Despite its venerable age, the theorem is embraced more widely today than at any other point in history, used everywhere from baseball predictions to artificial intelligence programs.

In essence, Bayes’ theorem tells us how to calculate conditional probabilities. That is, it’s used to gauge the probability that something is or will be true, given a larger context. It often boils down to a way of fine-tuning your understanding of how the world works.

Writing in Scientific American, John Hogan defines it as “a method for calculating the validity of beliefs (hypotheses, claims, propositions) based on the best available evidence (observations, data, information).”

Why Use Bayes’ Theorem for Scenarios?

One possible reason is to gain a better idea of which scenarios are most probable so that companies have a better idea of which strategies to emphasize and which resources to allocate. The danger is that such probabilities could do more to harm than to help the scenario planning process by giving business leaders a false sense that they understand the future better than they do.

How Does Bayes’ Theorem Work?

One typical example used to explain the theorem is medical testing. Let’s say that lab results show you test positive for a specific disease. Your doctor explains that the testing methods are 99.9% accurate for those who are ill. That means that there’s a 99.9% chance you actually have it, right?  Well, no, not according to Bayes.

Drilling down, we discover that the “99.9% correct” figure indicates that, assuming you do actually have the disease, the test will reveal that fact with 99.9% probability. But the test is, in this case, prone to false positives for 1 in 10 people who do not have the disease. That’s a little more comforting, but you assume there’s still a very high chance you have the disease.

Well, let’s find out by breaking it down even further.

Bayes’ theorem puts things in context via just one short equation:

P(A|B) = P(B|A)P(A)/ P(B)

The first part — P(A|B) — refers to the probability that A is true when it’s put in context of B. In this case, A refers to actually having the disease and B refers to the fact you tested positive for it. P, of course, stands for “probability.” The bar between A and B—that is, the |– basically means “given the fact that.” So, it all adds up to “the probability that you really have disease given the fact that you tested positive for it.”

Now let’s look at the rest of the equation:

  • The P(A) portion refers to the probability that any given person will come down with this disease. That chance happens to be just one in a thousand, or .001.
  • The P(B|A) portion refers to the probability that someone will test positive given that they have the disease. We know that’s .999. In other words, if you have the disease, the test will almost always catch that fact.
  • The P(B) is the probability of getting a positive test in the first place. You need two pieces of information to get this: 1) the probability of getting a positive test when you really do have the disease (.999) and, 2) the probability of getting a positive test when you’re NOT sick (otherwise known as a false positive). We’ve said there’s a 1 in 10 chance of that happening. So, you take .999 and multiply it by the chances of getting sick in the first place in order to get the chance of a true positive. Then add that resulting number to .1 multiplied by .999 (which is the probability of a false positive if you’re healthy).

So, here’s how it all breaks out:  P(A|B) = .999 * .001 / (.999 * .001) + (.1 * .999)

The top portion: .999 * .001 = .000999

The bottom portion: (.999 * .001) + (.1 * .999) = .000999 + .0999 = .100899

The final answer: .000999 / .100899 = .0099 which is a little less than .01 or one in a hundred.

And so, despite the positive test, there’s only about a one in a hundred chance you have the disease in question.

Can We Combine Scenario Planning with Bayes’ Theorem?

It depends. We could apply it if our understanding of the conditional probabilities is pretty good, but that will not always be the case. Let’s take the global warming scenario above. Maybe a great research team would be able to determine solid statistical relationships between renewable energy usage, climate change agreements and average global temperature, but it seems unlikely. So Bayesian analysis will probably not be of much use in that scenario.

On the other hand, if a major computer company wants to create scenarios that incorporate factors such as future profits, market share, and global demand, it could probably locate enough historical data to make reasonably good probability assessments.

But even if such a company could apply Bayesian analysis, should it? If it finds that one scenario quadrant is significantly more likely than another, will its leaders be tempted to bet on just one very targeted and specific scenario as opposed to developing a more adaptable strategy that could allow the firm to thrive in multiple scenarios?

Perhaps. And that’s a danger: Too much certainty can be deadly in an uncertain world.

On the other hand, as long as decision makers are wise enough to hedge their bets and expect the unexpected, a Bayesian analysis might add valuable perspective.

So, I recommend using Bayes’ theorem when probabilities are well understood, but be careful not to over-interpret the results.  The future is, after all, forever uncertain.


Note: The feature image for this article is by Mattbuck, courtesy of Wikipedia

Leave a Reply