BUSINESS ANALYTICS

A Lesson on Predictive Analytics

Salvatore Salamone | Date: 10-31-11 | Comments

The architects of "Moneyball" talk about data quality and offer lessons to businesses about predictive analytics.

When I tell my friends what I write about, I can usually see their eyes glaze over in about 10 seconds as I drone on about analytics, virtualization and my other favorite Smarter Tech topics. But recently, thanks to a Brad Pitt movie and two midmarket teams making it to the World Series (while teams with much higher payrolls did not), analytics is now in vogue.

Even though the book came out eight years ago and describes the efforts of the Oakland Athletics team of 2002, Moneyball is a hot topic today.

In the corporate world, analytics--and especially predictive analytics--are getting a lot of attention as businesses seek to gain competitive advantages by leveraging new data or old data in new ways. And for that reason, "Moneyball" architects Michael Lewis, author of "Moneyball: The Art of Winning an Unfair Game," and Billy Beane, vice president and general manager of the Oakland Athletics, were featured speakers at the IBM Information OnDemand 2011 conference held last week in Las Vegas.

Before their talk, the two gave an interview to IBM�s Predictive Perspectives Blog, where they discussed their work and how it relates to business.

They noted that organizations, from sports teams to businesses to government agencies, need to be looking for new ways to mine their data and new ways to think about their data. And they explained that baseball provides a great �best practice� for any business thinking about deploying an analytic solution.

For example, while the Oakland A�s set about trying to create new data and generate new information that wasn�t on the baseball field, Lewis and Beane found there were lots of inefficiencies in the way teams used existing data. �The data was there, but people were just not thinking about it properly,� Lewis said. So, for example, teams could easily calculate a player�s on-base percentage, but baseball was not appreciating the value of the statistic.

Lewis� point is that using data correctly is just as important as gathering it. �The minute you start to measure something and have a statistic, it has a tendency to become fetishized, like a player�s batting average,� he said.

In the case of baseball stats, some widely touted ones turned out not to be true indicators of a player�s offensive potential, and it led teams to make inferior decisions.

Relating this to businesses, he noted the similarity to a marketing department doing simple segmentation to identify customers for a direct mail offer. This analysis provides a somewhat superficial view of the customer and delivers one-to-some direct marketing (and lot of junk mail). Such approaches perpetuate accepted wisdom that all customers (and baseball players for that matter) are created equal, according to Lewis.

For baseball teams, businesses and government agencies, the hard part is innovating continually to find new statistics that have hidden meaning. That can be challenging. Businesses will find that the low-hanging fruit has already been plucked.

But Beane and Lewis believe smart businesses will put in the effort, because in today�s environment, acquiring a new customer is much more expensive than keeping one. Additionally, as business decisions become extremely important, �It�s worth investing in complicated ways of evaluating them, because if you find a slight edge, it means saving millions of dollars,� said Lewis.

That is why those in the C-suite (and in many instances, IT organizations) need to become more accepting of analytical techniques and not be afraid of what the data often reveals, or how it might change business processes.

To get started, Lewis� advice to businesses is:

1) Measure the right things. This will give your organization a huge advantage over the competition. Measure the wrong things, and you might dilute your efforts.

2) Invest in new ways to measure. A slight upgrade in measurement tools can have a huge effect.

3) People, especially those evaluating other people, are vulnerable to biases and mistakes of their intuition. Data-driven decisions can wake you up to your biases and make you a clearer, better and more efficient decision-maker.

4) Just because you�re unpopular doesn�t mean you�re wrong; it is sometimes ridiculously painful to innovate.