What's the Big Deal with Big Data?

written by Paul Feinberg

 

They know you better than you know yourself.

The collection of data by the private and public sectors has become so ubiquitous that it seems no one even questions it anymore — whether it be the federal government gathering information (allegedly) for national security purposes or Facebook mining for marketing and sales reasons. We barely notice that after we spend half an hour researching headphones or the nutritional information for our favorite snacks, the ads we’re served while browsing are suddenly for Beats by Dre or M&M’s. This type of data collection is not new. Large companies have been doing it for years. What is new is how technology has made it possible for organizations of all sizes to not only collect data, but also to quickly — even instantly — analyze it to make fast, often real-time, decisions.

Talk to UCLA Anderson professors such as Dominique Hanssens and Sanjog Misra, two of a number of Anderson professors doing research in the big data space, and they’ll tell you that, as academics, they’ve been leveraging big data for decades. The differences, they’ll remind you, are that 1) academic research is not-for-profit; 2) they have not been pressed for time when compared to private sector executives; and 3) their analytic findings weren’t previously needed for profit or loss decisions. But faculty are now leveraging the latest technology in their research and in the classroom, where students who once demonstrated disinterest in data now insist that their

Industries spending the most on Big Data
courses include discussion of marketing analyses. Those students are acting out of self-interest. They know that marketing managers are now expected to lead the quest for new and better data acquisition. “In the past, managers were thought of as consumers of analytics; that’s not true anymore,” Misra says. “Today, as a [marketing] manager, you’re often required to produce the analytics that you consume.”

Faculty doing research in the areas of finance and economics have long studied mass amounts of data, as have alumni working in those fields. However, processing that data wasn’t always easy. For example, the original UCLA Anderson Forecast reports were done at night when economists could get time on a mainframe to parse information. Marketing managers and other professionals didn’t usually have access to the computing power needed to make grand, micro-focused analyses. But, as is becoming more obvious every day, that has changed.

The New World Order

Tony Fagan is the director of quantitative research at Google. Fagan has guest lectured recently for marketing courses at UCLA Anderson, including those taught by Misra and Professor Peter Rossi. His participation is, in part, rooted in the respect he has for the work being done by UCLA Anderson faculty. “These guys are bringing the concept of big data into the MBA program, and they’re really ahead of the curve,” he says. Fagan cites three critical changes over the past decade that served as catalysts for the current big data tsunami.

The first came as a result of Google’s attempt to index the Web. The existing technology wouldn’t scale large enough, so the company developed MapReduce, a programming model and associated implementation for processing and generating large data sets. In layman’s terms, Fagan describes it thusly: “Rather than bring the data into a large computer for processing, MapReduce brings the processing to the data.” So instead of a huge mainframe or supercomputer, MapReduce utilizes lower-cost, commodity servers. The more data one uses, the more machines one simply adds.

The second innovation came when an open-source version of MapReduce called Hadoop was released by The Apache Software Foundation. This created an open standard and allowed companies other than Google to collect and analyze very large data sets in a less expensive way. As Fagan puts it, Hadoop democratized big data and created a competitive advantage for those diving into the data deep end.


“It seems they now know you better than you know yourself”

The third catalyst was the dawn of cloud computing, which created to even greater equality among companies. Instead of investing in hardware and having to manage the hardware/software environment themselves, companies are now able to pay only for the server time they use. “It’s almost a return to the mainframe model,” Fagan says. “Instead, now you have access to the cloud.

Now companies can just focus on applications.”

These trends contribute to the headlines as companies not only gather copious amounts of data, but they are also able to process it and put it to use in a cost-effective manner. Their decisions now go beyond marketing products they think you might want. Companies can improve — some might say “manipulate” — the overall customer experience through those targeted ads and purchasing recommendations, not to mention the results each individual receives to online queries.

The Case for Big Data
People Power

“We’re not at the point where decisions are made by computers, though there are algorithms that buy advertising and the market is moving toward that type of automation,” Misra says. “[Computers] can estimate which [advertising] channels are working, which are not working, and this information is fed into real-time algorithms that make decisions on your behalf. You can eliminate human beings for these types of decisions. But humans still have to make the big decisions.

Humans are still needed to look at the creative and strategic issues. This is why we teach the subject. If computers could do all the work, humans wouldn’t need to be conversant in big data. But, thankfully, humans are still needed to interpret and implement plans, even in the world of big data. That is where Anderson programs and classes on analytics and strategy have an important role to play.”

Hanssens’ take on the current state of data analytics begins with what he calls “the three Vs” — volume, velocity and variety. “Of these three things, volume is a straightforward extension of what we’ve always had. The volume of data is not new; we had that in the past, we just have higher quantities now,” he says. What’s new is the velocity and variety of the data. “Velocity reflects the ability to compile and analyze data more quickly.” What previously took a month to research now takes an afternoon, or less. As an example, Hanssens notes how a movie studio would previously have to wait until opening weekend to gauge how it would ultimately perform at the box office. Now, studio executives can obtain very good readings of public interest in the movie several weeks before launch by carefully monitoring search and blog activity levels, leaving time for marketing intervention before opening weekend. The variety of data is also improving, says Hanssens.

Data is no longer simply quantitative; i.e. neatly organized rows and columns of numbers. It is now qualitative as well; through, say, audio and video data feeds. New data analysis methods are being developed to capture the richness of such data sources, for example in computational linguistics, where it’s “the most exciting and the most difficult” to put into use.

Even though air travel gets a bad rap these days, Hanssens, a frequent international traveler, believes that the data airlines collect actually has improved the customer experience. By understanding what customers want and what they’re willing to pay for a service, “Airlines have made the flying experience much better,” he says. “Whether it’s reservations, cancellations or seat choice, they are managing to deliver personalized improvements in service by understanding what individual customers want.”
John P. Kelly (’98) leads Berkeley Research Group’s Predictive Analytics practice out of the company’s Century City office. BRG’s core business is providing expert testimony and data analytics to law firms and major corporations around the world. Kelly’s group brings a new business to BRG’s core competencies, utilizing BRG’s econometrics — the application of mathematics, statistics and computer science to economic data — and data science capabilities to help clients make better marketing, sales and operational decisions. “Our approach is to leverage three areas of our expertise,” Kelly says. “First is industry and functional experience and an understanding of the data available and relevant to that industry. You can’t come into an application greenfield – you have to have lived the problem or opportunity yourself to some degree.”

“The second expertise is the econometrics foundation,” Kelly says. “Automation is used to plum out relationships in the data, but automation often can’t do it alone.” Kelly says deep marketing-science talent can parse out meaning in the data with much better accuracy than what computers alone can tell us. Ironically, Kelly believes that a shortage of aptitude in the field — and here he cites Hanssens as one of the true talents — is actually slowing the potential of data analytics.

“The third part of the equation, is in fact the technology,” Kelly says. If you’ve honed in on
the right correlations, it is amazing what software innovation, raw computing power and massive data availability have created in terms of the opportunities for discovery.” In his work, this means using data science to make predictions about the future, whether it’s for increasing sales or finding ways to improve blood and plasma donation rates.

Ivan Markman (’02) is COO of MarketShare, a B2B company that provides analytic software-as-a-service solutions to major companies worldwide, with an eye on providing greater returns on marketing investments. MarketShare was co-founded by Hanssens and utilizes some of the mathematical models Hanssens developed. A recent Harvard Business Review article authored by MarketShare’s co-founder and CEO Wes Nichols cited three stages for companies that want to move to what he calls “analytics 2.0.” They are: 1) attribution, which refers to quantifying the contribution of each element of advertising; 2) optimization, which uses predictive-analytic tools to run scenarios for business planning; and 3) allocation, which redistributes resources across marketing in real-time.

Markman says that human capability converts recommendations from analytics and makes them actionable. “It’s both an art and a science,” he adds. “Some things can be automated. But this is the “Moneyball” of marketing,” he says, referring to the way baseball teams use statistics to build their teams along with traditional (i.e. human) methods of scouting players. Markman believes marketing professionals possess a skill set and intuition that take them only so far, with data now scaling their intuition by providing them with levels of information never before available. “Great levels of precision and sophistication are required in making marketing decisions today,” Markman says. “Going forward, it will be hard to compete on just pure gut and intuition.” X

Alumni Weekend
Create, Captivate & Engage with Big Data

Would you like to stay updated on Anderson’s big data workshops and events? Subscribe to the Center for Management of Enterprise in Media, Entertainment and Sports (memes@anderson.ucla.edu) mailing list.