By Greg Laugero, Vice President Strategy
One thing that stood out for me at the recent Predictive Analytics Innovation Summit in Chicago is this: Companies that are leading the way in the creative use of data understand that deep-dive analysis and marketing optimization are two sides of the same coin.
Here are just a few examples of how marketing optimization showed up at an advanced analytics conference:
-Netflix presented on “Quasi-Experimentation (Beyond A/B Testing).”
-We learned how Verizon uses big data and predictive models to feed marketing campaign management.
-Legendary Studios uses big data to inform how they test for sequel feasibility, concept evaluation, and casting for movies.
-The talk from Walgreens was all about optimization – “Delivering Value & Driving Best Customer Loyalty Through Pricing & Promotions.”
I think that such prominence of Marketing Optimization within a Predictive Analytics conference was a bit odd for the audience. When asked by one of the speakers how many of the attendees are involved in A/B testing at their companies, very few of the 200 people in the room raised their hands. Our experience at Numeric is that there is typically a separation between those who analyze and those who experiment and test. Sometimes they work together, sometimes they don’t. What’s clear from listening to representatives from Netflix, Walgreens, Legendary Studios, Verizon and others is that an emerging best practice is to have a very close alignment between these two disciplines.
To me, combining the ability to answer tough questions by digging into your data with the ability to run data-driven experiments makes complete sense. Rarely does a deep dive into your data yield definitive answers. Rather, it typically leads to more questions. Some of those questions are answered by more deep dives, but some are answered by devising different types of tests – A/B, multivariate, etc. – which are the specialty of Marketing Optimization pros.
Here’s how it works at these companies:
-Someone has a question about some aspect of the business – typically involving a better understanding of customer behavior.
-A data analyst figures out the available data to answer the question, works some transformational magic to come up with a clean and curated data source, and starts to run queries to find correlations. (The analyst might use a tool like Pentaho to do the ETL; then turn to a new tool like BeyondCore to do some automated identification of correlations between variables.)
-The analyst creates some insights that are communicated to the business. But more often than not, those insights lead to additional questions that call for the kinds of experimentation that typically falls into the realm of Marketing Optimization.
The presentation by the Chicago Cubs is a good example of how this back-and-forth works. The first speaker – a data analyst and employee of the Cubs – presented some compelling data related to how the Cubs and other sports team’s operations have been transformed by big data. For example:
-The NBA uses mapping technology to understand how effective a player is from different areas of the court. (See the Lebron James image included.)
-Pitcher/catcher/umpire combinations are used to predict the effectiveness of pitchers.
-The length of the baseball game affects concession revenue – fast working pitchers yield less revenue.
But after this enlightening discussion, another person – a consultant – came on stage and talked about how they used data to essentially run marketing optimization programs at Wrigley Field. For example, data analysis told them that one particular concession stand had a relatively high rate of beer sales compared to others. Hypothesizing that they could leverage the beer sales to drive sales of a salty product like peanuts, they ran an experiment (using the other concession stands as control groups). Using a digital menu board with a periodic interruption showing a peanuts promotion yielded a 50% increase in peanut sales at that stand.
Within the same presentation we go from big data analytics to experimentation and optimization. For the presenters, this was a natural combination. For others in the audience, this may have been a bit jarring. Nonetheless, the point is that the disciplines of data analysis and experimentation are merging – or at least finding common ground in leading organizations.
