Where Industry Meets Innovation

  • Contact Us
  • sign in Sign In
  • Sign in with certificate
mit campus


Search News

  • View All
  • ILP News
  • MIT Research News
  • MIT Sloan Management Review
  • Technology Review
  • Startup Exchange

ILP Institute Insider

May 1, 2017

From punch cards to deep learning: 50 years of marketing analytics

The power and influence of the customer has come a long way since Glen Urban arrived at MIT in the mid-60s, and much of that power can be tied directly to information technology.

Eric Brown

The power and influence of the customer has come a long way since Glen Urban arrived at MIT in the mid-60s, and much of that power can be tied directly to information technology. “When I arrived in the 60’s, computers were just beginning to enter marketing,” says Urban, former MIT Sloan School dean, and currently Retired Emeritus, Dean Emeritus. “We translated sheets of paper into punch cards and carried around big boxes of cards to feed into the latest computers with 128KB of memory.”

Urban recently gave the keynote address at the 2016 MIT Consumer Dynamics Conference, in which he revisited the many changes that have transformed marketing analytics since the ‘60s. “In the last 50 years we’ve moved from simple crosstabs and regression to deep learning and individual parameter estimation and multi-level networks,” says Urban. All these technologies have enabled a shift toward a more consumer-centric approach to marketing, he adds. “Today, companies can’t just push the products out like they did in the 60s and 70s. They look at customers as intelligent agents.”

The transformation of marketing analytics has also been accelerated by Urban, who is known for his Theory A concept of trust-based marketing. Urban is currently working half time at MIT, researching new marketing trends in story-telling, social networking, and deep learning integration.

Glen Urban
Professor in Management
and Marketing Emeritus
MIT Sloan School
of Management

Just as clickstream analysis, social media, and sophisticated targeting algorithms are transforming marketing analytics today, technology has pushed the field forward since the 60s. “The big driver was the computer,” says Urban. “We moved pretty quickly from big batch to online computers. In the 1970s, UPC code dramatically expanded the amount of data we had, and logit models were used to predict brand share and consumer choice. As we got better at the technology, consumers became more complex, so we needed even more analytics, data, and computing to make better decisions.”

The next big shift came in the ‘80s with the arrival of personal computers and online services. “When people started to get online with invoice prices, and not just retail prices, it launched a fundamental behavioral change in the way consumers interacted with companies,” says Urban. “I remember sitting at a meeting at General Motors when a manager in the back of the room said ‘People can get invoice prices on our cars? We’ve got to stop this.’ That was how naïve people were. But we were starting to realize that consumers had more power.”

In the ‘90s, the internet further expanded consumer buying power. “Customers could now access extensive information about autos, make their own choices and decisions, and walk into a dealership with a printout showing the cost of competitive cars,” says Urban.

In the first decade of the new Millennium, the internet began to change the way used cars and other goods were sold. “eBay started selling used cars,” says Urban. “Because of its rating system and guarantees on claims, used car dealers had no choice but to be honest. If you were not a 5-star used car dealer, you just didn’t sell cars. Today eBay sells $20 billion of used cars, and many people order cars without even seeing them. The internet was breeding trust, which was a big shift.”

Theory A and the changing customer relationship

Technology has continued to empower consumers in recent years via Google searches, online retailers like Amazon, and social media, Along the way, there have been misfires, as well. Companies often tried to apply new tools to old marketing concepts, leading to overhyped customer relationship management software and useless corporate mobile apps.

“With the rise of the internet, companies saw CRM as a better way to relate to the customer experience, but for many companies it was only a better way to do the same old push marketing,” says Urban. In the early 2000s, Urban addressed these issues with his Theory A, which stands for advocacy. “The idea of Theory A is to advocate for customers and offer transparency to provide confidence in the brand. In that context, CRM can be a form of trust-based marketing that builds long term relationships.”

Urban is frustrated with the slow progress of Theory A in the corporate world. “I had hoped that the CRM revolution from push to benevolence would have happened sooner,” says Urban. He adds, however, that companies that have used such methods, such as Amazon and L.L. Bean, “have really moved ahead.”

Recent examples in which “trust has been blown completely” could also help change opinions, says Urban. “I’m not sure VW will recover effectively from its emissions scandal. The Samsung Galaxy Note 7 is another example. Once you lose trust, it’s very hard to get it back.”

Companies must first understand that what customers want from a relationship is not usually what the company wants. “Companies thought the customers would love them if they put up a website or a YouTube channel, or tweeted about everything,” says Urban. “Yet, the brand relationship is really not what customers want. They want brand to customer to customer back to brand. They want the dialog.”

Indeed, the key to effective use of social networking is to enable customers to discuss the brand among themselves, says Urban. This concept is terrifying to many marketers, however.

“Companies are afraid of negative input so they want to control it, but that defeats the purpose,” says Urban. “There will be no secrets. If you don’t offer the venue, the information sharing will come out on other channels rather than the brand. However, by using social networking while having a mechanism to cut out the outliers of slanderous and wrong statements, you can have a very successful dialog.”

Brand-hosted customer to customer communications need not be limited to forums. One effective form of marketing that combines customer input with more traditional advertising is what Urban calls story-telling: encouraging customers to produce their own videos or other content to tell a story about using a product.

“One of our recent storytelling projects was with BMW, which encouraged customers to produce stories about their cars,” he says. “We found that story-telling can be a good way to build interaction and trust.” In one story, a man’s parents told him he had a secret twin. The twin turned out to be a BMW car they bought when he was born, and which had been driven for 30 years. “The story was about the life of his twin, and it was very engaging and positive,” says Urban.

Among people who saw the stories, there was a 30 percent increase of those who would consider BMW for their next purchase. After the study, BMW developed a new site called #bmw.stories, and posted videos on Twitter, YouTube, and their website.

From targeting to morphing

One of the biggest marketing transformations in recent years has been the rise of targeted marketing -- using purchase history and clickstream analysis to deliver targeted advertising. Some companies miss out on the subtleties of consumer behavior, however. For example, if you just purchased a chainsaw, it’s unlikely you’re going to want to look at ads asking you to buy another. “In that case, it’s much better to tell you about chainsaw repair services, sharpening, oil, or safety tips,” says Urban. “Dumb targeting gets in the way of trust. This is why we’re looking at big data to do more modeling on bigger click stream databases, so we have a better sense of the purchase and the process.”

One emerging technique promoted by Urban is website morphing: tailoring not only the content of marketing, but also the style. “Targeting tells you who to talk to, while morphing tells you how to talk to them,” says Urban. “Morphing measures cognitive styles based on website clicks, and then experiments to find the best targeted message. Every customer has a different cognitive style. A car buyer who is very analytic wants technical specs. Someone who is more visual and holistic wants images. Other types include impulsive vs, deliberate and visual vs. verbal.”

Urban’s research team recently worked with GM to develop 10 different styles of messaging. They then did real-time machine learning adaptive testing to see which ad was most effective for which cognitive style.

Urban is also investigating another machine learning technology called deep learning that makes use of multi-level neural networks. “We’re looking at whether deep learning can help us identify consumer needs for new products,” says Urban. “We’re now looking at acceptance of credit cards from sites like CreditKarma or NerdWallet, analyzing consumer clickstreams and card selections. It’s very promising, and if it really does work, we could use it to create new products and concepts.”

Looking forward, Urban is also interested in the potential for integrating deep learning with input from bio-bracelets, face and voice analysis, and other emerging technologies. Technology will continue to push marketing analytics the same way it did when the first mainframes were applied to the problem back in in the 60s. “It’s an exciting time,” says Urban.