Suppose you’re pondering a major consumer-goods purchase, such as a complete, top-end home theater setup. You want the works, but you know it’s going to cost a pretty penny, and of course you want the best deal you can get. Being a savvy buyer, you’ve spent some time online looking at reviews and studying product websites.
After narrowing your choices to a couple of favorites – one has everything you want and more, but the price is more than you meant to spend, while the other is affordable but not as fully loaded – you decide to put in a little more time comparing.
In a free hour on a Saturday, you return to the website of the Cadillac of component systems, taking one last, longing look, wishing you could squeeze it into your budget, knowing that wouldn’t be the prudent, responsible decision. You’re just about to leave for the cheaper competitor’s site when… a miracle! An e-coupon pops up on your screen offering the very system you want at a 10 percent discount.
Incredible luck? Well, no.
In fact, your every move since you started looking at the seller’s product online has been tracked and analyzed, including how often you’ve visited, how long you stayed, what day of the week and time of day you stopped in, and how you arrived each time. Everything that can be learned about your buying habits and even your lifestyle has been filtered in as well, and your customer profile has been building on an unseen server running a suite of algorithms, waiting for just the right moment to present the bait – a reasonable discount – and set the hook – getting you to buy.
Metadata systems that track online traffic and basic customer information have been around quite a while. It’s no surprise at all to get an email or chat from a seller whose website you have visited only once, and briefly at that.
What’s new here is that human behavioral psychology, condensed into a statistical model, has been integrated to a much greater degree, to the point that this system has a very good idea of what you really want, what you’re likely to do and when you will do it, and whether you are a good prospect to actually do it if enticed in a timely fashion.
Whether you are the seller closing the deal or the consumer feeling satisfied that you’ve gotten the deal you wanted, you may thank two UWM faculty members for laying the groundwork.
Amit Bhatnagar is a Lubar School of Business associate professor of marketing who specializes in online retailing analysis, both from the consumer side and the advertising and sales side. Bhatnagar teamed with Atish Sinha, professor of information technology management at the Lubar School, who specializes in business intelligence and analytics, and Arun Sen of Texas A&M University.
Melding their expertise, Bhatnagar, Sen, and Sinha set out to decrypt the brain of the online buyer and put the keys to that code in the hands of any online seller. They have created a marketable e-business strategy, expanding on the previous work of a giant in online consumer information analysis – Google.
“Essentially, what Google has done is that they have come up with a mechanism that is the equivalent to ranking all calls coming in to a call center,” Bhatnagar said. “The standard approach is to handle these calls on a first-come, first-served basis. Google’s new approach is that the more valuable calls should be handled first, and they have developed a technology to do that.”
But, Bhatnagar said, Google’s system is, in the first place, proprietary to Google, and second, limited to whether a person acted in response to a Google search engine ad or not.
Bhatnagar, Sen, and Sinha created their statistics-based approach, as compared to Google’s purely technology-based one, to work with any search engine and for any retailer, and to take into account much more data about the potential customer’s online behavior alone, without those web visitors ever picking up a phone to dial a call center.
“Many e-commerce websites have online salespeople who approach website visitors via online chat windows on a first-come, first-served basis,” Bhatnagar said. “We argue that site visitors should be ranked in terms of their probability of making a purchase, and they should be approached according to this ranking.”
The researchers worked with a major consumer-goods retailer – unnamed because of its proprietary interests – to access a deep trove of online visitor data and determine the behavioral profile of the potential customer who was most likely to buy. They were able to detect distinct patterns among those people who were most on the verge of a decision.
“Our model looks at many factors,” Bhatnagar said, “such as day of week, time of day, visit history, when the visitor came to a website.”
The result was what Bhatnagar, Sen, and Sinha call the prime “window of opportunity” when a retailer should engage the visitor for the greatest chance of “converting” that visitor from a shopper to a buyer. And in the hypothetical situation described above, it need not even (although it could) involve a human salesperson. A computer program alone could, using artificial intelligence and data gleaned from online visitor behavior, learn to dangle that discount in front of a likely buyer at just the right time to get him or her to pull the trigger.
In fact, one thing the researchers learned is that the best window of opportunity occurs when a visitor returns during weekend hours, as in our example of the Saturday shopper. But that’s only after a first visit of just so many minutes made during weekday hours. The first visit, to be ideal, should have come through a direct web search for the product rather than through a banner ad. And the sales contact, whether from a person or a computer, should come about three minutes into that subsequent weekend visit.
Someday, perhaps soon, Bhatnagar and Sinha’s UWM research will fuel a quiet revolution in marketing, in which artificial intelligence trumps classic sales techniques, and in which, perhaps, ultimately the customer will become a willing partner, playing a tactical game of trading time and information for deals and steals – but playing against a purely digital counterpart.