The How-Likely Case

The How-Likely Case: When The Most-Likely Scenario Isn’t Likely At All, by Mark Chussil

“Everything is 50/50. Either it will happen or it won’t.” — Unknown

Strategists commonly analyze best-case, worst-case, and most-likely scenarios before making a significant strategy decision. That covers about 0.000007618% of the possibilities. I’m not making up that number.

Here we’ll talk about why there are so many scenarios, why the “most-likely” scenario is hardly likely at all, and why we won’t get better predictions by getting better precision. We’ll end with some ways to make better decisions anyway.

So many scenarios

I ran a business war game with a company in which we took into account the potential actions of several organizations: their business, key competitors, government agencies, and others with vested interests in the industry’s evolution. Teams of executives role-played the organizations in accelerated real time. I designed the war game to… well, we won’t go into that.

Before the war game began I asked each team to spend 15 minutes listing the potential actions they could take. The shortest list had 8. The longest had 17. All told, there were 39,382,200 possible combinations of potential actions. If we were to look at three scenarios — best, worst, most-likely — we would be looking at 0.000007618% of the possibilities. (I told you I didn’t make up that number.) That is, assuming we could actually identify which were best, worst, and most-likely.

To put it another way, hoping that those three scenarios would tell us something meaningful about the competitive landscape would be like clipping a tiny part of a single letter in this essay and using it to infer what the essay was about.

Sidebar. The purpose of that war game wasn’t to evaluate their best, worst, most-likely (BWML) scenarios. I brought up the war game because I have enough information to calculate the number of scenarios and because the company is hardly unique in the number of possible futures it faced, which tells us that a conventional BWML analysis generally looks at a minuscule sample of what could happen. Even if a BWML analysis could assess a scenario with perfect accuracy, the odds that it would perform its perfection on the right three scenarios are vanishingly small. End of sidebar.

Of course the 39,382,200 possible scenarios were not equally likely. If a government agency issues a Thou Shalt regulation, it would presumably be much more likely that businesses would obey the rule than defy it. The combination of actions in some scenarios might be silly or even impossible. But still, however we cut it we’ve got 1) a large number of scenarios and 2) little confidence that the BWML scenarios will give us a representative picture.

Is there any hope? Let’s see how far we get with heroic assumptions. Let’s say we could:

• Efficiently and correctly eliminate 99.9% of those scenarios as being impossible, silly, redundant, or trivial. That’s probably too optimistic but we’re just thought-experimenting here.
• And then somehow view the remaining 39,382 scenarios. It’d take roughly 600 pages if we choose to print them.
• And then efficiently and correctly spot the most-likely one.
• And then achieve consensus that we’ve spotted the right one.

So, we’ve chosen the most-likely scenario. How likely would it be? Remember, we’ve already eliminated impossibilities, redundancies, and trivialities. Would it really be likely enough for us to call it “most” likely in any meaningful sense?

Precision not to the rescue

Precision doesn’t help much. That’s because scenarios can differ not only in degree but also in kind.

A business might change its price by +2.0%, +2.5%, or +3.0%. Those are differences in degree. A business might change its price, exit the market, launch a new product, merges with a competitor, vertically integrate, and more. Those are differences in kind.

It is reasonable to assume that the outcome of its 2.5% price increase would be somewhere between the outcome of the 2.0% increase and the 3.0% increase. It might not always work out that way but it’s a defensible shortcut.

It is not reasonable to assume that the outcome of merging with a competitor is somewhere between the outcome of launching a new product and vertically integrating.

The differences-in-kind issue means we don’t get to reduce the conceptual or computational load by looking at a few representative scenarios. (Representative of what?) It means too that we won’t know which scenarios are “best” or “worst” without evaluating them all. (Which isn’t impossible. That’s what ACS strategy decision tests do. Evaluating 39,382,200 could take as little as 30 minutes.) And that’s why we won’t solve the strategy-decision problem with a BWML analysis even if we deploy a microscope through which to peer more precisely at those three scenarios.

We don’t need a microscope to dissect a few scenarios. We need a wide-angle lens to explore a lot of scenarios.

Preparing versus predicting

Predicting the future is tremendously difficult partly because of the number-of-scenarios and differences-in-kind problems. It’s difficult also because it is hard to know what other organizations will do. They themselves may not know yet. They may be trying to make decisions at the same time you are. They may even be waiting and seeing what you do.

Still, you must make decisions and those decisions will affect your future. Fortunately, preparing can help you make decisions with much better odds of success, and fortunately too, preparing is easier than predicting.

In a future essay I’ll discuss how cool analysis can help you prepare. For now we’ll end with some shortcuts that don’t require any technology more advanced than thinking caps.

Scope out the terrain. You and I might not be able to assess 39,382 scenarios in our heads — actually, there’s no “might” about it — but merely knowing something about the possibilities is useful. In the war game it was strategically stimulating to know how many options each organization had. That knowledge prevented the executives from falling into overconfidence and tunnel-vision traps.

Apply competitive intelligence. CI can hint at or even reveal which actions the other organizations might adopt. If you can eliminate a few of the potential options, you can focus your attention on those that are left.

Keep asking what if. You are considering actions A, B, and C. How would each be affected if other organizations do X, Y, and Z? Your new product might not be materially affected if another company vertically integrates, but it might be affected a great deal if that company merges with a competitor. Are some of your options less vulnerable than others to what other organizations might do? I’ve seen exactly that, many times: an initially preferred strategy was discarded when a simulation showed it was highly vulnerable to competitive response.

Remember that your actions affect theirs. As you contemplate your moves, ask yourself not only what wonderful results you hope to achieve but also how your moves might influence others’ actions. You might not get those wonderful results if you provoke them to respond. Nobody intends to start a price war but price wars get started. On the happier hand, well-chosen action may lead them to behave as you want.

Talk it through. Say you want to do X. What happens after you do it? (“What happens” is about them as well as about you.) What happens after that? After that? After that? After that? Business war games can do a great job here. Rigorous, contrarian-rich conversation can too.

Notice that those approaches start by broadening the range of what you think about; in other words, they consider multiple scenarios. Contrast that with techniques that quickly and/or invisibly encourage you to narrow your field of vision. Best, worst, most-likely analysis does the latter: it defines the future in terms of three outcomes in which many factors — such as your competitors’ actions — are necessarily assumed to be known or irrelevant.

Look through your wide-angle lens first, the microscope later.