Correlation versus causation in the world of business

201602-gratisography-icecreamIn science, we take great effort to avoid confusing correlation with causation. Two things can happen simultaneously, but not be the cause of one another. Confusing the two is a well-known logical fallacy.

How do you overcome potential confusion? Test. Set up experiments that isolate variables. Show that you’re able to effect change in one variable, but not the other.

Sometimes you have to search, and search, and search for hidden variables. The more exhaustive your search, and the larger the volume of robust experiments, the better chance you have of suggesting a causal relationship.

What does this have to do with business? Everything!

Business is full of suppositions about cause and effect. We launched a new research and development project, and twelve months later, our sales went up. Bada big, bada boom…our R&D project drove increased sales.

You don’t have much room for experimentation in business. It’s too resource-intensive to establish properly controlled experiments. You often have to rely on finding natural experiments. Or of conducting such small-scale experiments that results aren’t interesting.

The most common approach is to rely on the narrative of leadership. High-profile decision-makers have a worldview. They make decisions that are consistent with that worldview.

At the executive level, you rely either on the worldview of the most charismatic leader, or you rely on the consensus worldview of the team. Regardless, a narrative structure drives the decision-making.

Does leadership believe that new technologies will fuel revenue growth and open new markets? Then expect an expanded R&D budget.

Does leadership believe that economies of scale will drive down costs and unlock hidden profitability? Then expect mergers and acquisitions.

Does leadership believe that the distractions of a large product and service portfolio are limiting performance? Then expect divestitures and portfolio rationalization.

Any of these beliefs make sense in a vacuum. You won’t be able to prove, in a scientific sense, that any one factor is dominant. Poll equally competent leaders on the question, and you’ll get different answers.

It doesn’t mean the problem is hopeless. You need narrative. You need a worldview. If you wait for scientific certainty, you’ll never make a decision, much less act on your conviction.

Your narrative is informed by experience. It’s informed by what you’ve read, who’ve you spoken with, what your values are. Sometimes you’ll see correlation, and you’ll have to make a judgment about causation. It’s unavoidable.

You can still speak precisely about correlation, and not come off as a coward. “If historical correlations hold, I expect to see [X].” You can comment on whether you expect correlations to hold, or to weaken, or to strengthen. You don’t have to speak in absolutes, confusing correlation with causation.

But sometimes, you won’t be able to avoid a prediction. Serving your customers, or satisfying a request from your boss, will demand you make a call. That’s fine. When you see correlation, rely on your scientific intuition to speak to the likelihood of underlying causation. It’s better than pretending the distinction doesn’t exist.

Leave a Reply

Your email address will not be published. Required fields are marked *