Let’s play 20 Questions. I’m thinking of a mathematical rule that applies to the sequence of numbers 3, 6, 9. You will win our game if you can guess the rule I’m thinking of solely by suggesting other three-number sequences and asking whether they fit my rule, yes or no.
The vast majority of people, when confronted with this problem, will quickly guess that my rule must be increasing multiples of 3. (It’s obvious, right?) To confirm this they might ask whether the series 12, 15, 18 would work (yes, it does fit my rule). How about the series 6, 9, 12? (Yes again). It only takes two or three questions to confirm for most people that this is indeed the rule—increasing multiples of 3.
But most people would be wrong, because the rule I was actually thinking of was simply that the three numbers in a series must be positive, increasing whole numbers. That’s all. So yes, the sequence 12, 15, 18 works, but so does 2, 3, 7. So does the series 25, 390, 4563, for that matter. Now before you dismiss this as a trivial problem, consider the fact that there are many, many number sequences that wouldn’t fit the rule I was thinking of at all. 9, 6, 3 would not work, for instance, nor would 1.5, 4.5, 7.5, nor -3, 0, +3. Suggest one of these sequences and you’d immediately learn that your hunch about “multiples of 3” is wrong.
A more productive way of thinking, at least in this kind of situation, would be to look specifically for examples that might disprove your theory. Technically, this is called “testing the null hypothesis,” but another way to think about it is to remember that genuine insight comes fastest when you look for information entropy—or, data that is unexpected or not predictable. Either way, what we are talking about is integral to the scientific method.
And the scientific method is an increasingly important idea for businesses as they try to make sense of a rapidly expanding volume of big data. Consider how Germany’s TeamBank, for example, uses the “null hypothesis” technique to improve its profitability. Part of the Volksbanken Raiffeisenbanken (VR) Group, each year TeamBank intentionally makes several thousand consumer loans that do not conform to its own detailed lending policies. They do this for the sole purpose of testing the assumptions that underlie these policies. By analyzing how well or poorly such loans actually do perform, statistically, the bank can continuously tweak and improve its credit rules.
Interestingly, however, as described in a recent Harvard Business Review article, “Finding the Profit in Fairness,” TeamBank also provides evidence for the central idea in Martha Rogers’ and my most recent book Extreme Trust: Honesty as a Competitive Advantage. We argue in Extreme Trust that the transparency generated by ever more cost-efficient and ubiquitous interactivity (including social media platforms) will compel businesses to compete by being more proactively trustworthy. To remain competitive, businesses will increasingly implement policies and practices protecting a customer’s interest even when the customer himself might not realize his interest is at stake. “Extreme trust” is demonstrated, for example, when Amazon reminds you that you already bought a book you are about to order, or when USAA counsels you that you probably don’t need the full amount of property insurance you are asking them for. Or now when TeamBank refuses to tack on additional (unadvertised) fees to a consumer loan.
So not only does TeamBank manage its business scientifically, it also provides a real-world example of a consumer bank gaining a competitive advantage by using “extreme trust.” It’s a sad commentary on the current state of the financial services industry, but most large consumer banks today earn a substantial portion of their profit by being untrustable—burying their fees, and even (in the U.S. at least) tricking customers into inadvertent overdrafts. Perhaps the success of TeamBank’s marketing strategy will serve as a persuasive “null hypothesis” test itself, which would help other financial services companies find their way to a more scientifically valid—and trustable–-business model.
[Image: Flickr user Wendell]