Wednesday, August 3, 2011

Brad Pitt stars in a movie about data...and some other stuff

Sure, women swoon over him and men envy him, but otherwise, Brad Pitt lives an ordinary life just as the rest of us do.  When he’s not collecting African children or riding motorcycles with Clooney, he likes to re-think data.  In fact, he made a movie about it, scheduled to hit screens on 23rd September.



What, you thought “Moneyball” covered baseball?  You thought author Michael Lewis skillfully wove a personal story about the can’t-miss prospect who missed with a team story about taking chances?  Read it again.  (And buy it here; I get a free set of steak knives if you buy enough).




No, “Moneyball” the book and, presumably the movie, offers an object lesson to marketers about overcoming limitations of data, a pillar of truly translinear marketing.  Every marketer who cares about data should heed the lesson that just because data have existed forever doesn’t mean they actually mean what he or she might think they mean.



WARNING: I’m going to talk about baseball.  But it won’t bore you one bit because really, I’ll be talking about the data.  See?  Much better.



Let’s start with a central motif of the book: the batting average.  Baseball enthusiasts have used the batting average as a benchmark for offensive prowess since Henry Chadwick invented it in the late 1800s.  To calculate a batting average, you must divide the number of hits a player has by the number of at-bats and present the results out to the thousandth place, or three places to the right of the decimal point.  So batting average really tells you how much the player contributes to the offense, right?



Wrong.



Any baseball fan knows that a player can reach base without getting a hit.  He may get hit by a pitch, he may advance on an error, he may beat a catcher’s throw to first base after striking out (unlikely) or, most frequently, he may draw a walk.  Moreover, “at bats” do not measure all the times that a player steps up to the plate.  At bats represent a subset of plate appearances.  At bats do not, for instance, include plate appearances that resulted in errors or walks.



It helps to mention at this point that Chadwick came from England and started with cricket, which makes no intrinsic sense to anyone outside the Commonwealth.



The hero of “Moneyball,” Oakland A’s general manager Billy Beane, spoke with statisticians who convinced him to think differently.  First, they said, it doesn’t matter whether a player gets a hit.  Rather, it matters whether they make an out.  Each inning has three outs--no more, no less--so the longer it takes a team to make those outs, the more likely it is that they will score.  



These same statisticians found a better offensive benchmark in on-base percentage, which measures the number of times a batter reaches based divided by his total plate appearances.  (Despite being called a “percentage,” the OBP is rendered to three decimal places just as the batting average is.  Apparently, they couldn’t completely de-Chadwick-ify it.)  The difference matters because the market--major-league teams--overvalues batting average in comparison with on-base percentage.  



So Beane quickly amassed a team of players with low batting averages and high on-base percentages.  And his teams made it to three straight playoffs despite featuring one of the lowest payrolls in the leagues.  Real baseball fans will note here that much better-paying teams bounced the A’s from the playoffs every year and that Beane himself admitted “my shit doesn’t work in the playoffs.”



However, these critics miss the point.  Beane didn’t set out to win championships; he set out to win ballgames because winning teams fill seats and thus bring revenue to the team.  Ultimately, Beane’s job came down to creating revenue.  So he succeeded.



Marketers, moreover, should pay mind to the larger point: challenge your metrics.



Beane and his merry band of statisticians asked themselves whether the measurements they used really told them what they wanted to know.  When they determined that they didn’t tell them what they wanted to know, they found measurements that did.



By the same token, marketers should:

  • Look at the numbers they depend upon the most
  • Ask themselves what these numbers really measure and whether this reality matches their actual needs
  • Create your own damn measurements if they don’t


Do it right and maybe Brad Pitt will play you in a movie, too.

No comments:

Post a Comment