Theory of Gravidity
At both the DHL Innovation Day and at the Production and Operations Management conference I have heard speakers use the example of Target, the US retailer, being able to predict whether customers are pregnant as an example of how the analysis of data is becoming intrusive.
The original source is an article in the New York Times from February 2012 by Charles Duhigg “How Companies Learn Your Secrets”. Among other things, the article describes how Andrew Pole of Target developed a model to analyse consumer purchasing data and demographic data to predict the probability that customers were pregnant, so that they could be targetted with vouchers for related products.
Duhigg’s description of the prediction’s use is engagingly graphic: “About a year after Pole created his pregnancy-prediction model, a man walked into a Target outside Minneapolis and demanded to see the manager. He was clutching coupons that had been sent to his daughter, and he was angry, according to an employee who participated in the conversation.
“My daughter got this in the mail!” he said. ‘She’s still in high school, and you’re sending her coupons for baby clothes and cribs? Are you trying to encourage her to get pregnant?’
The manager didn’t have any idea what the man was talking about. He looked at the mailer. Sure enough, it was addressed to the man’s daughter and contained advertisements for maternity clothing, nursery furniture and pictures of smiling infants. The manager apologized and then called a few days later to apologize again.
On the phone, though, the father was somewhat abashed. ‘I had a talk with my daughter,” he said. “It turns out there’s been some activities in my house I haven’t been completely aware of. She’s due in August. I owe you an apology.'”
Duhigg reported that Target denied that they specifically aim marketing at customers predicted as being pregnant and Andrew Pole refused to speak to him again. This colourful tale of Minnesotan life has most of the signatures of being an urban myth: the random narrative detail, the lack of enough facts to be corroborated and, most significantly, its playing on people’s concerns about corporations knowing more about them than they do combined with parental fears about teenage sexuality. Marketing analytics works to identify products that consumers are not buying but it is probable that they might buy and then target them with coupons, for no better reason than to avoid sending out coupons that will never be redeemed. How “improbable” it needs to be before you would not send out coupons is an interesting question. It is natural for people to seek narrative causal explanations about why a customer is now likely to be a purchaser of X, but the system just needs to calculate the probability. It is possible to calulate the probability that customers have had major life events, including estrangement, pregnancy and illness, from purchase data, but the reputational risk from errors prevents companies acting on it.