• Kanhaiya Maheshwary

How Target Inc. predicted a girl's pregnancy before her father did - Retail Data Analytics

In 2010, Andrew Pole, a data analytics expert managing Target Retail's big data program gave a rather solid keynote presentation at Predictive Analytics World's annual conference. He spoke about a wide range of deployments he was planning for Target Retail, which for those who don't know is America's 5th largest retail chain. At the end of his speech, he briefly mentioned something known as a "pregnancy predictor".


Now let us fast forward 2 years from that day, and go inside a Target Retail store in Minneapolis sometime in 2012. An angry dad walked inside the store with a booklet in his hand that Target had sent, addressed to his teenage daughter. It contained coupons for baby products such as baby clothes, cribs etc. The dad alleged that his daughter was still a teenager and there was no reason why Target should send such coupons to her. Taken by surprise and not knowing what to do, the store manager simply apologized. But that evening, the store manager received an apology call; from the dad. It turned out that his daughter was indeed pregnant, and hadn't told him yet. But Target Retail knew.




Retail businesses and customer loyalty

Pareto's "80-20" principle, which states that 80% of the consequences arises from 20% of the causes, finds utility in various situations and settings. It is largely true in the business world, where you will find that 80% of your revenue comes from 20% of your customers, basically alluding to your repeat buyers. That's why retail companies focus so heavily on the loyalty loop and offer their customers various incentives so they can keep buying.


In a bid to keep customers coming back and increase their purchase frequency, retail companies have started harnessing big data; and the practice is only getting stronger with the passage of time as newer technologies emerge. Through the use of predictive analytics, they are able to make smart recommendations to fuel more buying. For example, Amazon's "You may also like" is based on this predictive algorithm. However, Target's algorithm was too invasive, and it sparked a huge debate. But before that, let's see how Target did it.



Target's "Pregnancy Predictor"

If you are a Target member, you are automatically assigned a unique ID which maps your purchases. If you are not a member, there are various ways to assign you a dummy ID and then cross reference your purchases. This can be done if you are using the same credit card for payments, which is usually the case in US. People with a slightly higher purchasing power have specific credit cards for specific purposes such as fuel, restaurants, and grocery / supermarket purchases. Even if not, they will have one common credit card for all their purchases. In both these use cases, it is easy to cross reference purchases and assign them to an ID.



Besides assigning these IDs to both members and non-members, and then tracking their purchases, Target Inc. started creating various customer cohorts and back-tracking their purchases to determine purchasing patterns of those customer segments. One of these cohorts was that of pregnant women. Target was able to identify at least 25 products that pregnant women buy leading up to their due dates, and even beyond. And these 25 products were mapped according to the trimesters. For example, pregnant women bought unscented lotion at the beginning of their 2nd trimester. Another observation was the purchase of calcium, magnesium, and zinc supplements roughly 20 weeks into pregnancy.


One can argue that a lot of people might be buying vitamin supplements and unscented soaps. But there were several other purchases that were done alongside these, such as big cotton balls, sanitizers, wash cloths etc. Based on the purchases and the mix of goods in the shopping basket, Target would assign a probability score to each ID. This would compute the likelihood of a customer to be pregnant, and even calculate their expected due date! It so turns out at this angry dad's daughter probably ranked very high, and thus the baby product coupons.



The After Effect

Target realized that such heavily invasive advertising can creep out the customers. And instead of buying more, they may actually shift to another retailer who doesn't seem to "stalk" them as much. That's why Target's data analytics and advertising team started mixing up the promotional coupons to keep it generic, but also insert a few that are specific to the customer.


This Target case happened back in 2012. The world has changed massively since then. In fact, every single year the degree of technological upgrades we experience are much more than several past years put together. Today, every single app on your phone is collecting tremendous amounts of data about you. Whatever you type is being read and whatever you say is being recorded, even while you're not using the apps. The apps, however, continue to work in the background because remember (or maybe you don't) you gave them certain "permissions" during installation? In fact, if you try to disable any permissions, either the app will threaten you that it "can't perform" optimally or the process to disable those permissions will be so complicated that you will eventually give up.



That's one of the reasons why so many people are deleting their social media accounts, too. Because what is the price you are paying in exchange for your data and a few store coupons? It's pretty much the privacy of your existence. What you once guarded so deeply and let out only to a handful of close people is now absorbed by your phone apps, social media platforms, and e-commerce and retail stores in exchange for a $20 coupon. That's what it has come down to.


#100DaysOfBlogging #Day50

66 views0 comments