MACHINES ARE PREDICTING WHAT YOU’LL BUY

Written by | IoT & Big Data

Your favorite retailer already knows what you’re going to buy next. Customer spending data are goldmines of intelligent insight. Typically, a company will look at how you spend to determine how you are likely to spend in the future, and thus, how to target their marketing to your likely purchases. All of this is made possible through predictive analytics.

First of all, it’s important to note that most predictions are based on the assumption that consumers spend based on their interests. Financial behaviors that companies consider are the types of products customers are purchasing, when they purchase, and where they are spending their money. Having access to this kind of information allows companies to improve their value proposition by personalizing product offerings and delivering them to you when you’re most likely to need or want them.

The story of Target predicting a teenage pregnancy before her father knew has become a classic case study among
retailers. Target was sending the teenager coupons for baby clothes and cribs a few months before she was due, while her father was completely unaware of the pregnancy. Based on the daughter’s updated spending and browsing behaviors, Target knew she was algorithmically more likely to be a pregnant shopper.

However, there are only so many things retailers can do until a consumer moves on to his or her next major life event- getting married, buying a new house, or getting pregnant. That’s when the consumer’s usual routine breaks, and retailers need to adjust. To account for this, retailers have statisticians like Andrew Pole, who created what he called a ‘pregnancy-prediction model’ for Target. This model was built from an existing database associated with their customer ID numbers which included their historical transactions through credit cards, coupons used, opt-in survey responses, emails, demographics, and various other types of data. Using these data, the system can predict which females are most likely to be pregnant, and Target can start placing ads and offerings for an array of maternity products to those customers, while not making it intrusive.

The act of following a consumer’s spending pattern does not occur solely on the retail floor.

Recently, The University of Texas at Dallas published a study that shows how companies can target consumers based on the amount of money they have already spent elsewhere. This model requires a new kind of methodology–instead of looking at how a consumer buys at their own stores, companies can now examine share-of-wallet, or the amount of a consumer’s spending in a defined product category that a business captures. For instance, if a customer spends a total of $100 on toiletries every month, with $60 of that amount at Walmart and the other $40 at Target, the share-of-wallet for each company would then be 60% and 40% respectively.

This method makes sense as it also looks at a consumer’s total size of wallet, which means how much they can afford for a specific product category alone. If a consumer owns a wallet size of $100 for toiletries per month compared to those who can only afford $50 for the same products, then Target and Walmart would want to put more efforts into the former customer. The same goes for banks. A customer can either decide to put 100% of his or her savings into a bank, or split them at 50/50, or whatever proportion that they’d like. If a bank can entice the customer with a certain size of wallet into putting more of their savings into that bank’s system, then it subsequently gains a higher level of profit.

However, getting competitive data can be challenging as it isn’t always available. Without accessible data, companies can turn to predictive models to gain an estimation of their customers’ spending. Most of the data feeding into these models originates from past surveys or information aggregators, as well as historical transactions and approximated customer average income within a demographic or geographic segment.

In addition, interrelationships between product spending behaviors can also be used to facilitate cross-promotional efforts. If customers tend to spend more on soaps after buying shampoos, but not quite the reverse, then companies might want to promote shampoos more heavily.

Using consumer data for retail is one thing, but aggregating life patterns is quite another. Banks in this modern era are in a tremendous position of power to understand, predict, and exploit customer trends.

There is a lot of value that a bank can derive from your monthly bank statements. For example, Kasisto developed KAI, an Artificial Intelligence based conversational platform with an ‘expertise’ in finance, to help consumers with managing money and tracking expenses through simple messaging. The bot can also show you bargains that you might be interested in based on your spending history and geography (for example, deals from the exact store you’re standing in).

Furthermore, even the banks that you are connecting with can tap into your data and offer additional banking services when it makes sense. By combing through your statements, credit scores, reviews, and many other streams of resources, banks can start offering deals and saving options that not only match your interest, but also know when you need them most.

For such a long time, banks, retailers, and companies have retained a fluid amount of data streams that have gone untapped. With Artificial Intelligence, machine learning, and predictive analytics advancements, companies can now put these data into use–not only to increase profits, but also to improve processes, and in turn, benefit their customers.

 

Last modified: October 26, 2017