Predictive analytics is the technology of wringing predictions and profits from the seemingly irrelevant. It silently churns data to guess locations of meth labs and carjackings; to put a price tag on the value of customers, employees, and managers; and, above all, to predict who will buy what and pay how much. Big Data makes the most of its ninja invisibility. The consumer rarely suspects how many of his actions have been tracked and outguessed. The predictions are not perfect (yet). And just as Claude Shannon could outguess his own prediction machine, you can outguess Big Data. This chapter will describe a few of the more widely applicable tricks.
You’ve probably gotten weird calls from your cell phone carrier, cable company, or health club. The caller will ask if there’s anything she can do to serve you better. She’s not being neighborly. The call means that an algorithm has predicted that you’re likely to “churn” (cancel your service).
Customers are hard to dissuade once they’ve made up their mind to cancel. So predictive analytics is used to guess which customers will cancel before they know it themselves. Every time you go to the gym, you swipe your card. The algorithm knows whether your gym visits have been trailing off of late. It also knows how often people in your demographic group, with similar declines in attendance, have canceled.
Or maybe it’s your cell phone company and they know, based on your data usage, that you’d get a much better deal with a competitor’s plan. They also know that the competitor is launching a big ad campaign.
Since the predictions are not a sure thing, the conversation starts out chatty. The caller, following a script, is stalling to see whether you volunteer any complaints. Should the caller determine that the prediction was wrong and you’re a satisfied customer, the call ends there.
Otherwise, you’ll be presented with the so-called primary offer—a discount or freebie to accept a new contract. Never accept a primary offer. Once you reject it, the caller will bring up the secondary offer. It’s all in the script. Sometimes the secondary offer is better; other times it’s just different. You might as well hear both offers. You can always reverse yourself and ask for the first after hearing the second.
A still better strategy is to reject all offers. Wait a few days, then call to cancel the service. (Do this even if you intend to keep it.) You’ll find that scripts sometimes offer sweeter deals for customers who initiate the call to cancel. Once again, you want to reject the first offer and listen to the second. Accept the best deal—assuming you want to continue the service.
Here’s a grim thought for the next time you’re waiting to speak to a customer service representative. Whether you’re put on hold is sometimes determined by a prediction of how profitable a customer you are. The jargon is ARPU, for “average revenue per user.” To Big Data, we’re all slabs of consumer meat.
The call center’s software uses caller ID, just as some psychic hotlines do. In case you’ve never tried it, Googling a phone number usually returns a name and address. Another search or two, and you’ve got plenty. Boiler-room psychics play out this information to convince the gullible of their powers. Big Data has other methods.
Neustar Information Services, a data firm based in suburban Virginia, specializes in “real-time consumer insights.” When people call Jenny Craig, one of Neustar’s clients, an algorithm “pinpoints a caller’s location within feet.” It knows whether the person is calling from a cell phone or landline and has the mailing address of the account. Neustar is able to rate customer profitability instantly. Its software can predict the odds that a credit card applicant will qualify. When the chance is high, the customer gets to speak to a human being immediately. Presumed deadbeats are shunted to an overflow call center far, far away.
The next time you have trouble getting through to a business, you might want to hang up and call back using an Internet phone service like Google Voice. This time the software will see your Internet phone number, not your regular one. That won’t always work in your favor, but sometimes it will. Phone numbers are more useful in weeding out “unprofitable” customers than in identifying good ones. “Knowing the bottom is more important than knowing the top,” explained Gordy Meyer, founder of the consumer analytics firm eBureau. “If we can find twenty-five percent who have zero chance of [buying], we can say ‘don’t waste your money on them.’ ” Your Internet phone number is likely to have less data attached to it than your main number. Companies want new customers, so they are unlikely to lump blank slate numbers in with the bad apples.
Why not just dial *67 on your regular phone to block caller ID? It’s a little-known fact that *67 doesn’t work with toll-free numbers. Not only can the company read your phone number, but they also know you blocked caller ID for that call. That does look suspicious.
Go into Starbucks and you’ll find three sizes of coffee with confusing names: Tall, Grande, and Venti. A Starbucks newbie might imagine that Tall is the large size… until she notices that it’s cheapest and is actually the smallest. Since new customers don’t know quite what they’re getting, they tend to order the middle choice, Grande. Guess what? The three sizes are twelve, sixteen, and twenty ounces. The Grande is sixteen ounces, and you’ve just ordered two full cups of expensive coffee.
Analytics of sales data reveals just how contingent our choices are. The customer does not necessarily walk in the door knowing exactly how much caffeinated beverage she wants. That is a detail invented on the spur of the moment while checking e-mail or checking out the barista. When you give a customer three choices—say, small, medium, and large—and when that customer has no overriding reason to choose one over another, there is a tendency for the customer to go with the middle choice. In the marketing literature, this is known as extremeness aversion. It’s comparable to the way a magician’s volunteer picks a card or cup in the middle, not too close to either end.
In this case it’s the retailer who is the magician. The retailer engineers choices in such a way as to maximize profits. That typically means making the middle choice a little bigger or more expensive than what the median customer would otherwise choose.
This devious tactic works best with customers who aren’t that price sensitive—as at Starbucks or the Apple Store. Apple’s iPads currently come with 16, 32, 64, and 128 gigabytes of storage. The second choice, 32 GB, has been the most popular. Is it because the customer knows he needs exactly 32 GB? No, most couldn’t tell you how many gigabytes they’ve got on their laptop. They pick a choice in the middle precisely because they don’t know what they need.
Whenever you’re not sure which of several options is right for you, take that as a cue to think seriously about picking the smallest or cheapest one.
The most ambitious goal of Big Data may be differential pricing. This means tailoring prices to the individual, based on predictions of how much that consumer is willing to pay. It sounds unfair and un-American and Blade Runner–ish. Well, that ship has sailed. We’ve long had differential pricing in certain markets. Vacationers who plan ahead get lower airfares than businesspeople who can’t. Coupons, rebates, and loyalty cards are other examples. The shoppers who bother with them save pennies over those who can’t be bothered.
There is industry talk of a not-too-distant day when posted prices will be obsolete. Each consumer will, by technology to be determined, see a price customized to him or her alone. We’re already starting to see the outlines of that. The cashier at my supermarket recently invited me to apply online for a new program offering personalized discounts. There are no coupons to bother with; you just download an app to your smartphone. It gives you a list of discounts based on your shopping history.
What history? Everything. You’ve been using a loyalty card, right? The algorithm knows everything you’ve ever bought, as long as you swiped that card. The program is called Just for U and the supermarket’s FAQ page has a pertinent question.
Why don’t you just give me these prices in the store?
If every shopper were exactly like you, we could! But the reality is every Vons customer is different. This program is personalized just for U!
Uh-huh. A more complete answer is that they’re trying to learn how much of a discount it takes to make you switch brands. Personal discount apps engage in price spoofing, offering consumers discounts on brands or products they don’t normally buy, to see whether they react. Maybe your family likes Rice Krispies. You might get a good price on a store-brand rice cereal, or Cocoa Krispies, or Kashi granola. If the algorithm learns that many customers of a particular product are brand loyal, the company may be more inclined to raise prices (or to offer discounts only to those who would otherwise switch to a cheaper brand).
A Colorado blogger, Emily Vanek, discovered that she could milk her Safeway store’s discounts by switching between Starbucks and Dunkin’ Donuts ground coffee. Starbucks is more expensive. When Vanek switched to Dunkin’ Donuts, she looked like a price-sensitive shopper. This got Vanek a discount on Starbucks. Next time she bought Starbucks, and that ultimately triggered a special deal on Dunkin’ Donuts. This can apparently be continued indefinitely. Just as political campaigns lavish attention on those few voters who are on the fence, digital discounts flow to likely buyers who switch brands a lot.
Browser cookies customize your Web experience. Sometimes they customize your price. Navigate to the page of a product you intend to buy and note the price and shipping charges. Then clear your browser’s cookies and check the price again. You may find it’s lower (or there’s free shipping or a coupon for your next purchase).
As a general rule, repeat buyers care less about price. The first time you sent your aunt chocolates on her birthday, you researched the many online chocolatiers and their prices. The second time, you probably skipped the comparison shopping and reordered from the same company. It had the best value before, so why reinvent the wheel? This is the sort of pattern that analytics zeros in on. Some sites offer a better deal to new customers.
It’s not hard to clear cookies. Go into your browser’s settings, choose “Privacy” or some such heading, and click “Remove all website data.” The drawback is that you’ll be starting from zero with all websites. It’s much easier to use two browsers. I use Safari for everyday use (with cookies enabled) and Firefox (with cookies turned off). When I use Firefox, I’m a new customer to every site.
Then there’s the abandoned shopping cart ploy. Put whatever you want to buy in a retail site’s shopping cart. Click “check out.” Begin filling out the form. Make sure you enter your e-mail address but don’t enter any payment information. Leave the purchase in limbo and wait for the discounts to roll in.
Within a few days, you may get an e-mail reminding you about your abandoned shopping cart. Much of the time it will offer a discount, free shipping, or other inducements. A 2012 Reuters article noted that a litany of Web businesses, from Best Buy and Home Depot to Lands’ End and Zappos, were doing this.
According to one study, about 65 percent of online shopping carts are abandoned. Some buyers are interrupted by the boss walking in, while others experience sticker shock at seeing the grand total with shipping and tax. A follow-up e-mail gives retailers a second chance to close the deal. Analytics shows that follow-up e-mails are more likely to be opened, and that discounts do indeed work. Some companies are willing to discount their profit to gain a new customer, knowing that repeat customers might not care so much about price (see above).
From the consumer’s perspective, abandoning your shopping cart is like walking away from the car salesman. It may be necessary to get the best deal on the table.
• When a company that bills you regularly calls to ask how they can serve you better, it’s usually because their software has predicted that you’re about to discontinue service. Seize the moment to negotiate a better deal.
• Retailers offer a confusing range of options to sell customers up—to persuade them to buy coffee or megabytes they don’t need. When in doubt, consider the cheapest option.
• Supermarket discount apps and cards offer special deals to consumers who switch back and forth between two competing brands.
• Abandon your online shopping cart for a better deal. Select an item you intend to buy, start to check out, and enter your contact information but not payment. Then leave the site. In a few days you may get an e-mail reminder, often with a discount or free shipping.