In March 2013 the company released a new product: Audience Propensities. A propensity is a prediction, hedged by a probability score, about a specific consumer behavior, such as how a customer will respond to a particular offer. For instance, let’s say you have a discounted insurance product and you want to reach only those potential customers who would be extremely unlikely to buy that product at full price. Acxiom executive vice president Phil Mui and his team showed how Acxiom identifies the people with this propensity in a live demo at a product briefing in March 2013. In a manner of minutes the Acxiom system crunched 700 million rows of data and outputted a number. Mui revealed to the audience that if they were looking for someone with that propensity, “there are 275,012 people that you can reach out to.” Mui was careful to point out, “This is live. You can buy this audience today.” There are three thousand such propensities Acxiom can model.24
In September 2013, after a spate of unfavorable press and inquiries from the U.S. House of Representatives, Acxiom took a bold move in the right direction and opened a Web site called AbouttheData.com, to give consumers a partial window into the sort of data the company had on them in its databases, afford consumers the power to make amendments to their profile (these are tracked), and even opt out of being in the Acxiom database. The move was not without risk for Acxiom. As company CEO Scott Howe told New York Times reporter Natasha Singer, “What happens if 20 percent of the American population decides to opt out? It would be devastating for our business.”25 The database is a good first stop for anyone looking to better understand how she looks to marketers.
Careful record keeping and a bit of calculation can give you an idea of what companies often refer to as your “lifetime customer value.” But this is only an idea. Past behavior doesn’t dictate future results. You may have a cat today, but what if you fall in love with a dog person, or a ferret person? (Don’t do that.)
The big data present can give retailers a good understanding of your future buying as your future exists right now, but the naked future is one that’s always moving. To gain an understanding of what that movement looks like, you need more than a snapshot; you need to understand how the subject you’re observing is evolving, where she goes, what she does, what she encounters. You need to know who she talks to, who can influence her, and whom she can influence.
The days of planting RFID tags in cosmetics are long gone. Such cheap tricks are no longer necessary. Today, consumers give that information away eagerly.
How Facebook Turned You into an Advertisement
If you were on Facebook sometime between August 14 and October 4, 2010, you probably played a role in an experiment. Facebook turned 253 million users into test subjects to study contagion. No, Mark Zuckerberg didn’t release a hostile virus into the New York water supply (yet). The contamination event that the Facebook Data Science Team was monitoring was related to information, specifically URLs and how they spread between people.26
Here’s how this experiment worked: You were randomly placed into one of two groups. If you were in the first group, then when one of your friends posted a story you saw it in your News Feed as you normally would. If you were in the second group, the same story would appear far lower down in your News Feed where it was much less likely to be seen.
The objective of the experiment was to examine the probability of a user’s sharing a news article, video, or link even if that user didn’t know anyone else who had shared it. Facebook’s interest in information contagiousness goes beyond curiosity. A customer’s friends’ Facebook posts are an indicator of—among other things—how likely a customer is to abandon a company or brand or pick up a new habit.
The head of the experiment for Facebook was Eytan Bakshy, one of the star players on the Data Science Team. In person Bakshy seems very young to have such a coveted job, with access to a user base of hundreds of millions of people to experiment on. He bears a strong resemblance to the character of Leonard Hofstadter, the experimental physicist character played by Johnny Galecki on the geek-beloved television show The Big Bang Theory, but Bakshy comes off as a bit more serious . . . and a bit smarter.
The Facebook team knew that if they could show the likelihood of a user’s sharing an item when none of her friends had shared it, then the team could show how much more likely a user is to share a link that comes to her from someone in her network. And getting people to share information within a growing network is the entire value of Facebook. The ability to prove that the Facebook News Feed and the information shared in it can cause a behavior change is an extremely important aspect of the Facebook business model. Bakshy and his team found that you’re 7.37 times more likely to share a link that one of your friends has shared than to share that same article with no social signal.
The experiment also gave Facebook insight into a far more difficult question, one with a more potentially lucrative answer: how your relationship with different people influences the likelihood that you will like what they like.
When it comes to purchasing behavior, understanding who is influencing whom is a murky question because of a phenomenon called homophily, which is the tendency of similar people to exhibit similar behavior. If you and I both attended a liberal arts college, are of the same income, work in similar professions, and share some other overlapping demographic characteristics, there’s a good chance that we’ll both post a big article that appears in the New York Times to our Facebook page independently of each other. If the article shows up in my News Feed before it shows up in yours, it’s not clear that I influenced you to post it. You might have just happened to see the same article that I did later in the day. The Data Science Team’s experiment provided a formula for determining who in a network is inspiring who to share what. But the study’s most surprising revelation was that the people you’re closest to don’t necessarily influence your online behavior more than the people you’re only nominally friends with, folks with whom you have only a casual off-line relationship if any at all.
Facebook, it turns out, is particularly useful for researchers looking to separate fake friends from real ones. In sociology these two groups sometimes go by names that were originated in 1973 by sociologist Mark Granovetter: weak ties and strong ties. Our strong ties are the people with whom we interact often. These are family, friends, people we have a lot in common with, and with whom there is much homophily. Our weak ties are the people we add to our network without quite knowing why. Off-line, that category includes people you’ve had sparse interactions with, the girl from that cocktail party whom you remember as interesting, even if you can’t remember her name. Before social networks these weak ties would orbit for a while and then be lost to oblivion. On Facebook your weak ties remain visible through the News Feed even if you interact with them very rarely.
The experiment showed that weak ties in sum exercise more effect on your sharing than even your friends and family. They serve an important role introducing us to news and stimuli outside our normal circle. “These people share information from sources we might not frequent. As a result of seeing content from these users, you are many times more likely to share it. Surprisingly, while strong ties are individually more influential, weak ties are collectively more influential,” Bakshy told the crowd. Despite the reputation of Facebook as a place where people post mostly personal pictures of their kids and cats, in fact most of the links, articles, and other content that people share come from weak ties. It’s not personal, but news, petitions, meme photos of a grouchy cat, stuff from outside. But it has allowed Facebook to better anticipate what sort of content, what memes, provocative blog posts, and other material its individual users will respond to, Bakshy says. The evidence suggests that the approach is working. For all the recent talk of Facebook being the passé social network, it continues to dominate in a number of key metrics, perhaps the most important of which is time on the Web site. The average Facebook user spends almost an hour a day on Facebook, far more tha
n users spend on any other social network. No, our relationship with Facebook is not as exciting as it was. But it is stable, even marital.
It’s also allowed Facebook to better predict which of the people in your network, whether real friends or weak ties, can move you closer to buying something, and by how much.
Skip ahead to a second experiment that Bakshy spearheaded in 2011. This one involved a far more modest subject pool of just 23 million Facebook users. It worked like this: the subjects saw a story in their News Feed, perhaps for a History Channel broadcast or the Tough Mudder decathlon sporting event. Unlike a provocative New York Times piece, a funny George Takei photo, or some bit of information that you may share naturally, the purpose of this story was clearly mercantile. It probably didn’t look like a commercial but it was still a product endorsement, or, as Facebook calls it, a sponsored story.
The subjects of the experiment were placed into three groups. The first group saw the story and the identity of one friend who liked the associated product. The second two groups saw the story and the identities of more friends who liked the product. The stated goal was to measure the role of “social influence in social advertising.”
In advertising, well-paid celebrities have long been a proxy for the familiar. Psychologists Carl Senior and Baldeesh Gakhal have shown that we’re more likely to buy something from a famous person than from someone who is merely beautiful, in part because we trust familiarity over physical attractiveness. No one is more familiar to you than your friends. That’s why they’re better pitchmen than virtually any celebrity.27
What Bakshy and his team found is that even the slightest hint of weak-tie friend affiliation with a product or brand can increase the probability that a Facebook user will “like” a product story by 10 percent. The effect is much more robust when users are informed by Facebook that one of their strong ties likes something. Now here’s where contagion comes in: once you click “Like,” a seemingly innocuous action that the entire Facebook platform and all of its affiliates prompt you to do all the time, you become an advertisement to your friends. They see the sponsored story about the product you “endorsed.” A number of them see the link and also like it; next thing you know everyone is taking Nike+ challenges and has no idea why.
Bakshy and his team can also measure the dose-response function of each sponsored story to which you’re exposed, in effect predicting your tolerance rate for this sort of marketing or even, one day in the not so distant future, how quickly certain friends exhaust their influence on you. Bakshy has indicated that this is a possible future direction for the research. After all, there are thousands of different ways people can be tied to one another and, within the context of a social network where every interaction can be seen and scored, many different ways to measure those interactions—including, in the words of Bakshy, “trust and intimacy.”
Facebook has already begun making use of these insights. A program called Facebook Offers lets businesses extend coupons and deals directly to fans through the News Feed. When Facebook first announced offers, users couldn’t control whether their friends saw if they accepted the offer; your acceptance was a tacit endorsement that was broadcast to your friends.28 In the summer of 2013, Facebook second-guessed this approach. Turns out when users agree to share the fact that they accepted an offer, they have a much bigger influence on their friends to accept that same offer. 29
If you find yourself in a strange city, a program called Facebook Local Search, currently part of the company’s mobile app, can give you a list of local places to visit. Today, it works a bit like a less fun Foursquare. But Facebook collects a lot more personal and connection data than its competitors. In the future Facebook Local Search and Facebook Home suggestions will be based on your habits, your likes and dislikes, and which friend recommendations are going to be most influential for you. You, too, will be making recommendations, perhaps without realizing it.30
Facebook assumes that you’ll come to appreciate the additional personally relevant context in the ads to which you’re exposed, the additional convenience of knowing that someone in your network liked a product that you’re considering. They may be right. We may not yet feel at ease with the direction that advertising is taking but few of us have a strong attachment to its current manifestation in which we’re constantly bombarded by images and sounds from people we do not know, selling us items we do not want, and incapable of hearing us when we voice our refusal.
The big-box retailers are trying to shrink themselves down to like size. In 2011 Walmart purchased Silicon Valley data-mining company Kosmix to help Walmart Labs get a handle on social network data from Facebook and Twitter. They deepened their investment in 2013 with the acquisition of predictive analytics start-up Inkiru. They’re competing with dozens of other outlets and corporations including Target and Amazon to make the most out of whatever data about you they can get. In the hyper-personalized retail environment, your likes, dislikes, ZIP code, income, habits, gullibility, and friendships will one day affect not just your impulse buys and online shopping, but every purchase you make at the gas station, the coffee shop, and the grocery store, even including the price you pay for tuna.
This is one of the areas of predictive analytics where Gryc sees opportunity and change. “Right now, the corporations can afford these analytics. They can afford the data. But what’s going to happen in the next few years is that it will become a lot easier for consumers to calculate a lot of these metrics.”
The naked future envisioned by Gryc is one in which consumers and retailers are locked in an information arms race and he’s the arms dealer. Today, one side has a clear and seemingly insurmountable advantage. But consumers have more information at their disposal than many realize. Any bank or wallet app can monitor in real time the amount of money we spend on things we don’t really need and may not even want. A number of other apps can help you discover the consequences that a given purchase will have on your waistline, bank balance, or your goals. An app called Oroeco can even help you track and predict the effects of your purchases on the environment. Technology helps companies better predict where we’ll be, what we’ll buy, what we’ll want, but it also helps consumers consume smarter, and sometimes even consume less. The big data present is one where companies use our data against us, to trick, coerce, and make inferences that benefit them at our expense. That behavior won’t change in the future, but with a better awareness of what’s going on and a willingness to experiment with the right tools we can make the fight a bit fairer. With enough personal record keeping, it’s possible to turn the tables on the ever more coercive advertisers. For instance, using a QS system such as the earlier mentioned Tictrac, you can see how the media you expose yourself to affects your purchasing behavior, your ability to meet your own savings goals. Indeed, you can see how your exposure to Facebook changes your happiness and your financial security.
You have all the information that you need to help you resist ever more coercive mobile messaging; you give it away to your phone all the time. The next step is to start using it, to become smarter about you. Imagine answering a push notification on your mobile device and seeing the following message:
There is an 80 percent probability you will regret this purchase.
The answer to highly customized, context-aware advertisements is the strategic, personal use of personal information. The war is just beginning. Both sides will experience victories and perhaps moments of true partnership.
But not every party will emerge from this war as a winner.
Back at Saatchi & Saatchi, there’s a clamber of electric saws outside. Down the hall, long slabs of drywall sit against bare beams. Carpenter stations are set up. I ask Becky Wang if the company is expanding. She tells me that the offices are shrinking. “We lost some business,” she explains. One of the workmen stops by her office. “Hi!” she says. “I’ve been looking for you. I’ve been trying to use the computer in the
hallway, but someone keeps stealing the keyboard. Can we move it?”
As we walk to the elevator, Becky gives me the names of people in New York and Silicon Valley whom she considers to be leaders in retail analytics. They work out of tiny start-ups that I’ve never heard of. The conversation feels a bit melancholy. The company where she works, with its carefully constructed facade of cool invincibility, is vanishing piece by piece. Before the elevator opens, I ask where I can follow up.
“Use my Gmail,” she says and then, quietly, as though not to alarm the people around her that the vessel they are on is flooding and she is stepping into the last lifeboat, “I’ll be leaving here, too.”
CHAPTER 7
Relearning How to Learn
THE year is 2020. You’re at a parent-teacher conference on the eve of the first day of a new school year. Your daughter is going into freshmen algebra tomorrow and you’re at this conference to meet her new teacher. You show up armed with every math quiz, every math problem that your child has attempted throughout elementary and middle school, as well as a breakdown of how long she took on each and at what time during the day—after breakfast, before dinner—she performed best. The profile may even reveal whom your daughter talks to online, whom she studies with, and how those supposed friends influence her homework performance. This is a lot of information to carry around. If you were to print all this material, you would be dragging boxes along behind you. But this information is already stored on the cloud. All you have to do is give your child’s teacher a link.
You have a request: “Would you mind taking all this data and creating an individual learning program for my daughter to make positively sure she finishes this year with an understanding of algebra? By the way, she’s very shy, won’t ask any questions in class, and probably can’t devote more than an hour to algebra a night. Thank you.”
The Naked Future Page 15