Business Intelligence’s Fatal Flaw

 

In an article from Readers’ Digest Treasury for Young Readers, you are shown how to construct an Hexapawn robot. Hexapawn is a game played on a nine-square board with, as one might expect, six chess pawns. The pawns move as they do in chess, and start on rows 1 and 3. The object of the game is to advance a pawn to the last row, capture all of your opponent’s pawns, or else put him in a position where he cannot move. The robot part of it has to do with twenty-four matchboxes, some maps, and colored beads. Little maps of every possible position are drawn up and placed on the tops of the matchbooks. Colored arrows indicate each possible move from that position, and corresponding colored beads are placed in the matchboxes. You then “teach” the robot to play by playing game after game of Hexapawn, and removing the colored bead from the appropriate matchbox that corresponds to the last move of all losing games. After about eleven or so games, the robot becomes perfect, and cannot be beat.

Before I go on to challenge outright the tons (literally) of research and writing that have gone into modern quantitative analysis in business, I want to discuss another game: the Ultimatum Game. This game has the game manager approach two subjects, and makes the following offer: to give them $100 (USD) on the condition that Subject B agrees to the first plan that Subject A articulates to split the money. If Subject B does not agree to Subject A’s plan, then neither person receives anything.

Game theorists attempting to determine Subject A’s best strategy for maximizing their payout calculated that the best proffered plan would be for A to receive $99, and B to receive $1, on the theory that B would rather receive $1 than nothing at all. But a funny thing happened to Subject A as he was preparing to deposit his $99: that plan was almost always rejected in actual experiments of the Ultimatum Game. There were actually instances where a 50/50 split, or even splits where Subject B received more than Subject A, were rejected. After having reviewed the data from the experiments, game theorists tended to chalk up the dramatic differences between their theoretical expectations and real-world results as owing to “cultural” factors, or else Subjects B not acting in a rational manner. Nothing could be further from the truth.

Consider the calculated/expected outcome’s implications. If a stranger approached you and, say, a friend with whom you just happened to be walking down a sidewalk, and presented the Ultimatum Game’s rules, and your friend offered up the 99 – to – 1 split, does that not imply that your friend was 99 times more worthy of unearned largesse than you? And – the value of a single dollar bill being what it is – wouldn’t it be worth it to forgo the $1 in order to reject the implication? We haven’t even touched on Subject B’s willingness to punish Subject A for being greedy, or arrogant, or dozens of other reasons why the experimental data was so at odds with the theoretical projections.

Which brings us to the problems with quantitative analysis in business as it is currently taught in the nation’s universities. The free marketplace is an extremely complex environment (it may even qualify as chaotic – there’s really no way of knowing). And yet, the most basic analysis tactics put forth in the current literature treat it as if it’s relatively simple, and can be captured mathematically. For example, the decision on whether or not you should close your business when it is losing money is supposed to be predicated on whether or not your revenues exceed your fixed costs, rather than just your total costs. Umm …yeah, what if next week you are to learn of the award of a contract that you bid, where you estimate a 50% chance of winning, and that work would put you back into the black, and in a big way? Or of three such proposals? Of course, the kind of information that your general ledger can offer up can’t possibly capture that, and is, really, comically incapable of making available the definitive quantitative analysis that would support that decision, one way or the other. The asset managers are simply turning to their version of the Hexapawn robot, and retrieving the colored bead that tells them what to do, not realizing that the game they are playing is no where near confined to a nine-square board. And, when their so-called quantitative analysis is proven wrong, they can simply deflect blame on to cultural factors, or players acting irrationally. Hey, guys – it’s the free marketplace! Nobody acts in a way that you can predict, or calculate – in other words, the world is, by your definition, irrational, and will always be that way.

Must I say it? The notion that the general ledger can possibly inform the decision of whether or not to stay in business is pseudo-intellectualism of Cecil B. DeMill proportions (I’m actually hoping that Cameron uses this quote in his teaser on the web site). And that is business intelligence’s  fatal flaw – the arrogant premises from which the quants proceed.