Big Data: Wow, what an over-used expression this past year. 2012 was the year of headlines such as “Barack Obama’s Big Data won the US election,” “How Target Figured Out A Teen Girl Was Pregnant Before Her Father Did” and “Big Data knows when you’re walking past a pharmacy and texts you a coupon.” Blah, blah, blah.
The application of analytics and mining data is indeed part of a very powerful revolution, but it’s only part of the story. Analytics can tell you a lot of about the who, the what and the when – but what happened to the why? We need to combine analytics with primary research to understand more. This combination of big data analytics and primary research is the solution behind marketing analytics 2.0.
I was recently invited to a review of the insightful, multi-million dollar development of customer segments at a major US weight loss company, but I left the room puzzled. With dozens of richly described personas, the company developed an excellent research study to uncover the primary segments and their personal struggles to lose weight. Some struggled with urban living and eating on-the-go, some used food to take comfort, some simply didn’t care and were joining for peer pressure, along with a myriad of other personal challenges with weight loss.
The key question from several timid marketers around room though, was, “How do we take action? How do we help each consumer, each different in their needs, to lose weight? How do we even recognize who is who?”
Performing primary research and understanding consumer attitudes, preferences, backgrounds, price perceptions and propensity to purchase lead to rich insights that bring a customer base to life. We can understand within which segments a brand excels or fails, and we can help a brand re-align itself towards a set of target segments or white space in the competitive landscape. But in this age of data collection and personalized marketing, the sole use of primary research seems incomplete.
On the other end of the spectrum is the world of big data predictive modeling, made famous by Netflix movie recommendations, the infamous Target case, and credit card fraud detection. These examples involve analyzing hundreds of gigabytes of purchasing records, understanding individual customer purchase patterns over time and uncovering the signals that predict future behavior.
Most results are not so clear-cut and are certainly less news-worthy. For example, you may find that if a person signs up to a program on a certain day of the week, they are less likely to retain. Or if they buy a certain category of product, they are less likely to come back. The results leave you with basic questions of causation vs. correlation or simply not knowing the reason why someone has behaved the way they have. What is it about that day of the week for one individual, or that product category for another individual that is making them a low value customer? The power of this approach is that a company can dynamically recognize who belongs to what segment, and what that segments trigger signals are. With this information we can differentiate marketing and take powerful personalized action on individual customers.
Unfortunately, the underlying historical data is often skin-deep and the “why” is often left unexplained. Recognizing someone and their trigger signal is only half the battle. What do you actually do with them? How do you truly change behavior? Here’s where we come back to the traditional approach, which helps uncover the “why” – their needs and motivations, from which we can design truly effective marketing outreach. It is the “why” that changes behavior.
An integrated approach of the traditional methods and the big data approach yields the best of both worlds –the “who and what” from predictive analytics and the “why” from the traditional approach that make true behavior change possible. Modeling on customer data allows you to uncover “what” signals drive a desired or undesired behavior and the “who” to take action on. Traditional research approaches then step in and provide a voice-of-the-customer lens that attacks the “why,” or the motivation behind the action, and the underlying unfulfilled need.
The unsung beauty of the Target modeling was not necessarily finding the key purchases that predicted pregnancy, but rather knowing that pregnancy was what they were looking for in the first place! The combination of the human (voice of the customer) and the machine (historic data) allows you to monitor and alert on key signals that predict churn, a purchase or some other key behavior (e.g. fraud false positives) and act on it using the insight into their needs and preferences that lead to behavior change.
Big Data makes big promises – but it doesn’t wrap problems up with a big bright bow. What truly is the gift-wrapped package is the practical combination of multiple tools. We need to combine the multiple lenses of traditional research tools and modern day number crunching. And that’s what will ultimately get us to the game-changing state of Marketing Analytics 2.0.