Asking The Right Questions In An Age Of Predictive Analytics


In 2012, New York Times reporter Charles Duhigg wrote a story about how Target used transaction data to create a pregnancy-prediction model. When the model predicted one customer was pregnant, Target promptly sent coupons for baby clothes and cribs. The only problem was that the customer was in high school. According to Duhigg, the dad stormed to the local Target to complain about the coupons but when he returned home, the daughter confessed. The model was correct.  

We’re living in an era in which brands use our data to anticipate what we need next with alarming precision. And yet, there’s one problem that our data can’t fully solve. Like a politician, a brand has a reputation to maintain. It has to design images and messages that charm and persuade. While it might seem like brands are using our data surreptitiously, one of their primary tactics—advertising—is plain to see. 

Last fall, a major retailer approached me with an interesting question. They were redefining a customer segment and needed help with some research. They knew how old, wealthy, and educated the segment was. Most consumer-facing brands have access to this kind of information. It’s not precise and it doesn’t need to be. The idea is to divide a customer base into a few big segments. 

However, if you’re a consumer-facing brand, you need something more. You need to observe something that your customers feel but have yet to consciously acknowledge. Then you need to broadcast that insight back to them. A great ad is not meant to move you off the couch and into the store. The goal is an involuntary head nod—the subtle “I get it” gesture.

The question my client asked me was: “What are our segment’s top ‘stress drivers’?”

When I sat down to draft a few survey questions, the first thing that came to mind was money. I knew from previous research that most people stress about not having enough—even people who have plenty. The client knew that. Like most retailers, their business was built around this central theme of consumer life.  

The next thing that came to mind was work, and then relationship problems. Writing survey questions is difficult. Some behavioral scientists compare the self-reporting mind to a press secretary and the brain to the oval office. A survey doesn’t have access to decisions that happen in the oval office. You can get a clue if you ask the press secretary the right questions. But you have to be clever.

I started with something I knew and argued outward. I imagined walking up and down the aisles of a store. For me, shopping comes with a dose of low-level stress and I wondered how many people were like me, so I jotted down a question: “When you’re shopping in the store, do you normally feel like you’re in a hurry?” Everyone had to pick: “Yes or No.”

Next, I noticed that my low-level stress doesn’t have an object. When I shop, I don’t really have a reason to be in a hurry. I jotted down another question. “When you’re shopping in the store, are you normally late for something? Yes or No?”

I shared these questions with the client who was initially hesitant. Most market research surveys are long and filled with questions that measure stated preference such as “What’s important to you when choosing a retailer?” and “How many hours a week do you spend shopping?” My survey was about 90 percent shorter than the surveys they normally conducted. It wasn’t obvious what it was measuring.  

Luckily, the client liked the idea and made a great suggestion. They pointed out that some people would say “yes” to the first question about feeling in a hurry and “no” to the second question about being late for something. For this group, they recommended the logical follow-up: “If you’re not typically late for something, why do you feel like you’re in a hurry?” It’s a great question because it catches the press secretary off guard.

When I analyzed the data, I found that about 30 percent of the respondents admitted that when they shop they feel like they’re in a hurry—even though they usually have nowhere else to be. When prompted to explain why they felt so rushed, this group responded with a variety of answers, each of which we categorized and then used to create a list of “stress drivers.” We found that some people shop during the day and can’t be late getting back to work, some rely on drivers who idle in the parking lot, and some have family members waiting at home. Some just don’t like shopping and some get anxious in crowds. Just about everybody dreads waiting in line. And then there’s the kids, which were universally described as ticking time bombs. For parents with small kids, a shopping experience is one tantrum away from an early exit.  

Customers in this group liked the retailer—they even preferred it over competitors, according to our research. The problem was that despite being filled with friendly employees, the store was getting in the way of something else going on in their lives. They didn’t expect the retailer to solve their problems. I imagine that they just wanted their problems acknowledged.

Duhigg’s juicy anecdote about Target’s pregnancy-prediction model is probably false. Many people on Target’s mailing list receive such offers—you just never hear about all the women who received the coupons who weren’t pregnant. In the digital era, Duhigg’s fabricated story offers a shift in perspective. Don’t ask, “How are brands using my data?” Ask, “Do brands know the right questions to ask about me?”

The media theorist Marshall McLuhan observed that agencies are trying to make “effective contact with a genuine feeling.” He urged us to see ads as more than a means of manipulation. Like a poem, book, or film, ads are simply participating in the “perennial flow of perception and impulse from the few to the many and from the many to the few.”   

The research we conducted will eventually make its way into this perennial flow in the form of more images and messages. I hope a few customers will see them and react with an involuntary head nod. 

Photo Credit:

Sam McNerney