A Few Words About Data

st-louis-airport-300x225I was going to title this article “How Target Knows You’re Pregnant Before You Do”, but my mom mentioned to me that my titles never truly match what the heck I’m talking about.  This article is going to be a rare exception to that rule.  Thanks Momma.

The predictive power of computers–once they have enough data–is fascinating.  My preoccupation with data began when I learned that supermarkets were getting in trouble for  sending out ‘suggestive’ coupons after crunching a customer’s shopping habit data.   The prime example of this, I believe, was Target.  Target was sending women coupons for baby stuff before these women knew they were pregnant!  Imagine the father of a high school daughter getting maternity clothing coupons in the mail?  [for a great article on this, check out How Target Figured Out A Teen Girl Was Pregnant Before Her Father Did, Kashmir Hill, Forbes, Feb 16, 2012].   Similar problems are created at home when criminal defense attorneys start sending “Arrested? Hire Me” letters to that same daughter–somebody’s going to start asking questions.  Sure, kids think that they can hide their shenanigans by intercepting the mail before Mom & Dad get home, but “kids helping out” can be even more suspicious than overly suggestive attorney solicitations.

Anyway.  How does Target know a woman is pregnant?  Well, we humans are rather predictable.  No, I’m not talking about you, dearest reader–you are unique, different, and “nobody understands you.”   Target’s computers were able to anticipate the shopping needs of women based upon changes in their shopping history.  Once a computer knows you’re buying luggage, you’ll be buying “travel size” deodorant and toothpaste at any moment.  Same goes for pregnant women.  The data points were, apparently, more sophisticated than the old myths regarding pickles & ice cream.  It was important that Target figure out when a woman was pregnant before it is made public, because, “birth records are usually public, the moment a couple have a new baby, they are almost instantaneously barraged with offers and incentives and advertisements from all sorts of companies.  Which means that the key is to reach them earlier, before any other retailers know a baby is on the way.” [from an excellent article entitled “How Companies Learn Your Secrets“, by Charles Duhigg, New York Times, Feb 16, 2012]  Duhigg’s NYT article provides some clues on how this works, noting that: “One Target employee I spoke to provided a hypothetical example.  Take a fictional Target shopper named Jenny Ward, who is 23, lives in Atlanta and in March bought cocoa-butter lotion, a purse large enough to double as a diaper bag, zinc and magnesium supplements and a bright blue rug.  There’s say, an 87 percent chance that she’s pregnant and that her delivery date is sometime in late August.” 

At times, people strive to be unpredictable.  In soccer matches, a penalty shootout may determine the winner of the match, so goalies will use past shootout data to predict where the ball is heading.  But the soccer player on other side of the penalty kick will attempt to be unpredictable–knowing that the goalie knows where he has kicked the ball in previous shootouts.  Computers also try to be unpredictable.  Think about gambling machines.  Think about the lottery.  If you could use data to predict how a computer would spit out numbers, it would be over for machine based gambling. Can we predict “random” computer behavior?  To dig deeper, let’s start with this recent NPR Report:

“A few years ago, a rumor started going around the casino world.  There was a crew of Russians hitting up casinos across the U.S.  They’d roll up, find their favorite slot machine, play for a couple hours, and walk out with thousands of dollars.  They didn’t lose.  . . All of it was caught on camera, but there was no evidence that these men ever physically tampered with the slot machines.”  NPR Planet Money, Episode 773: Slot Flaw Scofflaws

A Wired article by Brendan Koerner put it this way: “In early June 2014, accountants at the Lumiere Place Casino in St. Louis noticed that several of their slot machines had–just for a couple of days–gone haywire.  The government-approved software that powers such machines give the house a fixed mathematical edge, so that casinos can be certain of how much they’ll earn over the long haul–say, 7.129 cents for every dollar played.  But on June 2 and 3, a number of Lumiere’s machines had spit out far more money than they’d consumed, despite not awarding any major jackpots, an aberration known in industry parlance as a negative hold.  Since code isn’t prone to sudden fits of madness, the only plausible explanation was that someone was cheating. “Russians Engineer a Brilliant Slot Machine Cheat–And Casinos Have No Fix“, B. Koerner, Wired, May 6, 2017.

How did the Russians Cheat?  Data.  Simple slot machine data.

How could data help beat a slot machine, if the machine is supposedly random? According to Koerner’s article, “Slot machine outcomes are controlled by programs called pseudorandom number generators that produce baffling results by design. . . But as the “pseudo” in the name suggests, the numbers aren’t truly random.”

The Russians used their phones to video record a couple dozen slot plays.  This video (data from 24 spins) was sent back to Russia for processing.  To you and I,  this data would look something like “Cherry – BAR – BAR – Cherry”, but the computers in Russia would crunch this information and tell the boots on the ground when it was time to bet big and press the spin button.  Having video recordings of 24 spins on a slot machine was enough data to help these folks win $250,000 a week!

Data is everywhere, but most folks just don’t want to use it.

And, this brings me to what happened Monday night.  I’m sitting at a nice airport bar at St. Louis Lambert International Airport, near Gate E24 (Yes, I took that picture, and the sunset looked better in person, they always do).   My flight back to Orlando left at 9:00 pm, and the bar was empty.  Several of the bar’s waitresses decided to have a pow-wow next to me, begging their boss to let them cut out of their shift early because it was so slow.  Their boss rejected them, explaining that he didn’t know what other flights were going to be near Gate E24 later that evening.  Maybe a Boeing 777 jumbo jet would have a flight out later, and that many passengers would require all hands on deck.

At this point, I injected myself into negotiations on behalf of the waitresses, defending their early departure request (pro bono, I might add).  I asked the boss if he had any data to suggest that more flights were heading out.  He claimed to have no access to that data.  Hum.  There’s “DEPARTURE” monitors all over the airport.  The bar’s section of the airport is all Southwest Airlines, so, its public knowledge as to which planes are taking off, what gate they’re departing from, what time they’re leaving, and where they’re going.  All of this is important data to an airport bar.

The boss explained that certain flights are better for business than others.  His bar loves departures to Las Vegas.  The 6:30 am flights to Las Vegas will pack his bar.  Yes, standing room only at 5:45 am, Bloody Mary’s for everyone.  Contrast these bar revenues to the early morning flights to Wyoming, and you’ll find there’s little need to open the bar early (no offense to you party people in Wyoming…did I just say “party people”?  Wow, I’m getting old.).  The point is, this bar could utilize publicly available flight data to better schedule waitress shifts.  I know what you’re thinking, I’m just some old white guy sitting behind a desk (second time I’ve claimed to be ‘old’ in this paragraph) .  What do I know about running a bar?  Nothing.  I’ll get back to writing legal stuff shortly, I promise.