By Michael E. Raynor
The application of logic to data in the pursuit of answers can be very effective. Aristotle illustrated this in demonstrating how one would determine whether or not Socrates is mortal. We start with the observations that all men are mortal and that Socrates is a man. The application of deductive reasoning to those two premises yields the inescapable conclusion that Socrates is mortal.
If such deductive reasoning were all it ever took to reach a correct conclusion, there would be far fewer bad decisions. The problem is, far too often the facts are either ambiguous or incomplete in ways we cannot see until it is too late. When we apply reason to unwittingly incorrect or unknowingly under-specified premises, we end up with precise, convincing, and completely wrong conclusions. It’s a distinction that logicians have long appreciated: A valid argument is one in which the conclusion follows from the premises; a sound argument is one that has the added benefit of being based on true premises.
Christopher Cerf and Victor Navasky’s classic book The Experts Speak compiles mistaken predictions; while some are truly ridiculous, many more simply illustrate the distinction between valid and sound—and should inspire humility rather than provoke hilarity. For example, in 1962, a Decca recording executive had to make a call on a new four-member guitar group. Noting that similar-looking bands had been popular a few years ago but that most were failing miserably, he concluded that “guitar groups are on their way out” and that, therefore, the Beatles should seek their fortunes elsewhere.
We should be similarly sympathetic to the Western Union executive who turned down Bell’s patent for the telephone, which he could have acquired for his employer for $100,000 (about $2.5 million today). While we don’t know what his reasoning was, it might have been something like this:
Premise 1: To be successful, technologies must have customers.
Premise 2: There are no customers for the telephone.
Conclusion: If brought to market, the telephone would be unsuccessful.
It’s easy enough to gin up a counter-argument today, but from the perspective of 1876, I can see his point.
Finally, consider the iPhone, a smash hit by any standard. When Apple launched it, in 2007, I assumed it would fail. Why would I think such a fool thing? Well, the iPhone was a sustaining innovation—that is, it seemed poised not to create a new market so much as perhaps incrementally improve the existing mobile-phone business. Disruption theory—still my preferred candidate for a means of predicting an innovation’s success—states that entrants fail when they show up with sustaining innovations. Ergo, the iPhone will flop. Oops.
What’s remarkable is that rats actually do much better than people: We try to figure out the pattern of colors and keep guessing some of each.
The problem is not, I think, that I’m an idiot. Rather, it is that we lack both perfectly accurate theories and complete and accurate data to feed into them. Consequently, people of good will and intelligence can look at the same data and reach opposite conclusions, because in many circumstances, we are forced to use something beyond mere data and logic to make our choices. But what?
Outcomes that are unpredictable and chaotic at one level are often predictable and stable at another. We don’t know if or how much it will rain in Granada today, but we have a pretty good sense of the precipitation we can expect in Andalusia over the course of a year. This allows us to state expectations for any given instance in terms of probabilities, which serves to quantify our ignorance and so temper our hubris.
Careful research can also reveal what sorts of activities are systematically and strongly associated with the outcomes we desire. There is increasing evidence, for example, that strategic positions built on unique non-price dimensions of value are systematically more profitable than those based on low-price—perhaps as much as 80 percent of the time. Now, there are plenty of counterexamples of companies that compete on low price that do just fine, thank you very much, just as there are periods of drought or deluge along Spain’s Mediterranean coast. But since we can’t predict specific outcomes, the most reasonable response is simply to play the odds and prepare for what typically happens.
Going with the grain of the wood in these matters can be more difficult than you might think. In an experiment, when researchers showed rats two colors with relative frequency, say, 80 percent red and 20 percent green, but in random sequence, rats pretty quickly figured out that red appeared more frequently, so they picked red four out of five times.
What’s remarkable is that rats actually do much better than people: We try to figure out the pattern of colors and keep guessing some of each. The best of us manage to infer the ratio of red to green and then match the frequency of our guesses of each color to the frequency with which each turns up. This results in a long-run success rate of, at best, 68 percent.
With this in mind, consider the travails that Abercrombie & Fitch has gone through. From its debut as a public company in 1996 through to 2007, it was terrifically successful, building a brand and customer experience that allowed it to command significant price premiums over its competitors and resist the sort of discounting that can erode margins and profitability. When the recession hit in 2008, A&F stuck to its guns and didn’t discount while its competitors (American Eagle, Aeropostale, and others) did. This made it the object of some ridicule by analysts and commentators, who pointed to the company’s falling revenue, store closures, and declining profitability.
With the recovery, however, A&F might well get the last laugh. Unlike its industry peers, A&F does not face the daunting—and sometime insurmountable—challenge of curing its customers of an addiction to sales and clearances. Success is, of course, not guaranteed, and although I don’t know how CEO Mike Jeffries and his management team reached their decision, I can say that what they chose to do was the odds-on favorite to work in the long run.
The point of the story is not that Jeffries and his team are smart and that analysts and leadership at other retailers are dumb. The data was sufficiently ambiguous that no matter how insightful, experienced, and expert one might be, the conclusion to be drawn was simply under-determined, as is likely to be the case with many choices of moment. However, from strategy to risk management to human resources, it is possible, through the careful evaluation of large-scale research, to get a pretty good probabilistic handle on an increasing number of important questions. The key to making the most of these insights is having the discipline to trust the numbers and not give in to our intuitions. If non-price positions are typically better than price-based positions, don’t sometimes pick one and sometimes pick the other. Instead, however difficult it might seem, always go with what works most of the time.
Hold on, though. There are at least two reasons that a great deal of latitude for individual choice remains. First, when it comes to complex questions, our research is still rudimentary, often conflicting, and incomplete. Second, the particulars of how you play the odds matters in ways that we have yet to fully unpack. That still requires an ineffable mixture of intuition and judgment.
We have something to learn from rats when it comes to playing the odds, but I don’t think they’ll win the race just yet.
From the Archives