The Naked Future (15 page)

The Las Vegas casino business model before Gary Loveman was built entirely around the concept of bigger: bigger signs, bigger fountains, bigger volcanoes and attractions visible from the Strip, bigger names on the marquee to pull people in, bigger lobbies, more of everything. In the fight for flash Loveman saw a war not worth winning. He turned his focus to the job of remaking Harrah's utterly unglamorous customer loyalty program, transforming it from a simple thanks-for-coming voucher scheme into a massive, data-run, telemetric decision engine.

Loveman didn't invent the customer loyalty program; the airlines did. In 1978 the U.S. airline industry was deregulated, big carriers expanded their routes and slashed their prices, and a vast frontier was suddenly open. American Airlines relocated to the Dallas–Fort Worth area the following year, where an executive quickly realized that the company could offer lower-price fares for the customers that used the airline most often. But finding these people was no easy task. Fortunately for American Airlines, they had one of the world's biggest computerized databases, the Semi-Automated Business Research Environment (SABRE). SABRE allowed travel agents and American Airlines' clerks dispersed around the world to book passengers on quickly filling flights in something like real time. In what might be considered the first case of a major company using a computerized database for customer profiling (outside the insurance industry), American Airlines scanned their database to figure out who were their 150,000 best customers.
^7
,
8
These people became members of the world's first computerized customer loyalty program, AAdvantage, in 1981. (Frank Lorenzo, Texas International Airlines CEO, came up with the first frequent-flyer club in 1979, but this small airline lacked the computer resources of its larger competitors.) After the official launch, other big airlines developed their own programs within a matter of days. These soon included not just deals on airfare but also rental cars and hotels in such trendy spots as Las Vegas, Nevada.

Loveman took what the airlines had been doing for decades, and was in place already at Harrah's, and perfected it. In 1998 he
created the Harrah's Total Gold program, today called Total Rewards. Here's how it works: when club members book their hotel or restaurant reservations, when they swipe their Total Rewards club card in one of Harrah's video slot machines or use their Harrah's account to place table bets, when they win, when they lose, when they hesitate, when they cash out, feel the itch, and come back to hit the one-armed bandit one last time on their way out of town, the system knows . . . and remembers. But the database is more than just a play-by-play record of plundered 401Ks. Customer service reps both in the casinos and at call centers around the world look to learn everything they can about Harrah's Total Rewards members to tailor very specific offers to them.

It's not cheap to collect, keep, and utilize all this data. Harrah's reportedly spends more than $100 million a year on IT, but Total Rewards has more than earned its investment back. What started with 12 million subscribers in 1990 hit 26 million in 2003, more people than the combined populations of Greece and Portugal. By 2010, 40 million people were in Harrah's system. This gives the company access to the lives of the people in its casinos. Instead of rigging the table games, the system works to rig the customers.
⁹

“Let's say we have a sixty-year-old woman who lives in a comfortable suburb of Memphis,” Loveman told journalist Robert Shook in 2003. “She visits our Tunica property on a Friday night, briefly plays a dollar slot machine, and goes home. Based on traditional casino methods, she'd have a low theoretical worth, perhaps a few dollars. Consequently, she wouldn't be a likely prospect to pursue, and little effort would be made to get her to come back. And in all likelihood, she wouldn't respond to it. Our present system draws distinctions between the observed worth of a customer and what we predict their worth is.”
¹⁰

Your predicted worth is a number representing how much money the Harrah's system has calculated you can be persuaded to lose—er, gamble with—when you go to one of their locations. It's based on your ZIP code, what you play, how you play, and other indicators of wealth and willingness to gamble. Reportedly, the system is 90
percent accurate at predicting how much a customer can be persuaded to drop at a casino.
¹¹

These customer-worth scores serve a greater function, predicting exactly what offer potential customers respond to and when to issue it. If you're in the casino and you're losing heavily (as measured by the account activity on your card), Harrah's will dispense a “luck ambassador” with a coupon for something the system has calculated you'll like, anything from a free drink to show tickets. But these offers don't just come to you while you're on the floor. If you go to Harrah's on your birthday, on your annual vacation, on the seventh day of the seventh month of the year, the company will send you an offer timed to encourage you to do that sort of thing again.

Do you go to the casino when your wife is out of town? As Christina Binkley of the
Wall Street Journal
discovered, Harrah's knows that, too. She observed a Harrah's-hired telemarketer named Mr. Salvador go through the process of contacting repeat customers with special offers and saw firsthand the wealth of data Harrah's can use to turn any customer encounter to the company's favor. “On a recent list was a thirty-four-year-old man who hadn't been to a Harrah's Entertainment Inc. casino since November 2003. Before then, according to the data, he had made trips to the Rio in Las Vegas, as well as casinos in Tunica, Miss., and East Chicago. ‘This is a customer who can only play when his wife is on vacation or when he's on a trip,' says Mr. Salvador.'”
¹²

Where did that information come from? There are several ways to infer it. This thirty-four-year old man might have indicated on a customer satisfaction survey form his marital status and that he was traveling alone. Alternatively, perhaps when he entered that Harrah's hotel in November 2003 a friendly desk clerk or casino cashier asked if he was in town on a special occasion and he mentioned that the special occasion was being away from his wife. She simply complied with the company guidelines and typed this answer into the form in front of her. Either way, what was a very forgettable exchange for this thirty-four-year-old man has become
a piece of data that now follows him everywhere and influences how and when Harrah's contacts him. Harrah's has become a vast sense organ. As an institution, it's constantly detecting and responding to new information on its customers. It remembers everything, weighs every interaction. It knows your limits better than you know them yourself, and it wants you to keep playing.

Writers in the business press credit Loveman with completely changing the culture at Harrah's. Read that to mean he fired a lot of people, mostly in marketing. In his interview with Greenfeld, Loveman explains his impatience with traditional marketing methods. “Testing and measuring is important to us. When our employees use the words ‘I think,' the hair stands up on the back of my neck. We have the capacity to know rather than guess at something.”

What does the history of the casino loyalty programs mean for the future of shopping, marketing, and advertising? Simply put, the retail world of 2013 has become a Harrah's casino.

The date is April 25, 2012. I'm at an upscale gourmet grocery chain in a tony Toronto neighborhood. With me is Wojciech Gryc, the twenty-five-year-old creator of a software platform that automates customer analytics. His company, Canopy Labs, is looking to put the predictive analytics-crunching capabilities of Walmart and Target into the hands of small- and medium-size businesses. The Canopy Labs platform “predicts which customers are most likely to accept the offer you are going to be suggesting,” he explains. Though only four months old, Gryc's company has already attracted funding from Silicon Valley angel investors. He's riding the big data hype wave but his tech experience is deeply ingrained. He got his start in big data at IBM but his dad had him at the computer at the age of ten. Point to a product and he can tell you how to use data to sell it. “You can optimize any shopping experience,” he says. In this instance he means optimizing the experience for the seller, the clients he works for.

We're here to see what that looks like.

Grocery stores are laid out according to a marketing science that's actually been around for decades. Thousands of recorded customer decisions help such outfits as the middle-class gourmet grocery that I'm standing in to nudge customers where the store wants them to go. Past the cash registers we encounter cakes and cupcakes. These are prominently displayed because they're potential impulse buys. They have a higher profit margin and longer shelf life compared to the fruits and veggies. They're also the sort of thing people don't put on a regular shopping list. The store wants to make sure we see them. This is an act of priming. Even if you demure the sweets, you still had to think about it. You'll be less able to fight off the next impulse.

Over the past decade no chain amassed more data to figure out these sorts of placement decisions than Walmart, which experiences more than 1 million customer transactions every hour.
¹³
In 2004, in what is probably the most famous instance of a retailer using customer data to effect product placement, Walmart stacked the shelves of its stores standing in the path of Hurricane Francis with Pop-Tarts and beer. Their terabytes of customer data from a previous mega storm, Hurricane Charley, showed that these were the items most sought after in Walmart stores before hurricanes. Hurricane survivors, it seems, like treats that can be microwaved or ingested straight out of the package (only the latter is advisable in the case of beer).

Discovering the reasons why customers pick up one item and discard another is the central challenge of predictive analytics in retail. This is what I call the Bear problem. My cat, Bear, rejects food randomly. I open a can of a type of food he liked yesterday, or a week ago, and he'll either turn his nose up at it or eat it for reasons that aren't obvious. He also spends part of the day running around outside. We have an inside cat who never rejects food. If we assume that Bear's fickleness is caused by something he discovers outside (which is indeed just an assumption of causation based on
an observed correlation), then determining why my cat rejects what I give him involves tracking a huge number of variables, such as where he goes on his excursions and what he encounters there.

In figuring out why people pick up one thing and put down another, Walmart is basically faced with the same problem that I am. And they've employed some controversial tricks to get at it. In 2003 Walmart and Procter & Gamble ignited a firestorm of controversy when the
Chicago Sun-Times
revealed that the two companies were outfitting cosmetics with RFID tags. When customers picked up one of the chipped cosmetics items like Lipfinity lipstick, a surveillance system would follow the customer around the store to see what other items she considered buying but did not.
¹⁴

Both Walmart and P&G maintain that they used the tags solely for the purpose of tracking how the consumers handled the products
inside the store
and, it should be said, no one has found any evidence to the contrary. But from a retail perspective, data about how customers use products outside of a store are far more useful than knowing what they do inside the store. Does the average Lipfinity owner leave her lipstick on her dresser at home or carry it in her purse? Where does that purse go?

Not of all Walmart's tactics to triangulate customer behavior have been so controversial. In the fall of 2006 Walmart stepped away from the stereotype of the impersonal big-box retailer with the creation of its “store of the community” program, which gave managers much more leeway in terms of stocking their shelves and laying out their stores. The company took careful note of what worked in what market, to better understand why some strategies succeeded and others failed. This enabled Walmart to expand the number of planograms, or acceptable merchandise-layout configurations, from five to more than two hundred by 2010, an important step in customizing the retail experience to the individual. What was an impersonal big-box outlet began to metamorphose into a village store, better reflecting the purchasing habits of the community.
¹⁵

But altering the layout of particular outlets and the displays
therein for
every
shopper will never be practical, even in the future. The challenge today is to re-optimize an experience designed for a statistically generic person into an experience for living individuals.

How do you personalize something that was designed for the aggregate?

“If you were the management of this store, how would you optimize tuna steaks to me?” I ask Gryc. We determine that one of the key variables of tuna steak pricing should be freshness. Tuna steaks are highly perishable. From the store's perspective, it makes better economic sense to offer me a discount on the tuna if the alternative is throwing it out at the end of the day. In the vast majority of grocery stores around the world, the current method for accomplishing this is to hang a
SALE
sign on the tuna's glass casing. A better method, Gryc explains, is notifying your customers who buy a lot of tuna and then offering them a time-sensitive coupon. These e-coupons can easily be delivered to a user's phone on the basis of context, meaning where that user is and even what activity that user is engaged in, easily determined by other app usage or simply location. (Someone in a bowling alley is unlikely to be windsurfing.)

The store can then track the number of those coupons that are redeemed. This is sometimes called one-to-one marketing at scale. It allows the store to order tuna with greater confidence that it'll be able to move it at different price points. It's less likely to have excess tuna and won't need to offer as many deep discounts. A restocking decision that, perhaps, was once made at the district manager level can be made at the department manager level; instead of being made in a distant office, it's made by the guy behind the counter. When people redeem the coupons the store gets data on which sorts of customers respond to which sorts of offers.