A Method to March Madness
You don't have to wait until Sunday to know which 65 teams will fill the coveted spots in the men's NCAA College Basketball Tournament. Jay Coleman, a professor of operations management at the University of North Florida, and Allen Lynch, an associate professor of economics at Mercer University, have used predictive analytics to determine the 34 teams not granted an automatic bid to the Tournament.
Using analytics software from SAS--which at the end of the month is launching its biggest analytics upgrade in 15 years--Coleman and Lynch have created the Dance Card, based on the same 29 basic-data points considered by the Tournament's selection committee. The Dance Card's formula includes such factors as ratings percentage index rank and a team's number of wins against the top 25 teams. Each of the 29 data bits is assigned a weight to create the Dance Card equation.
According to AMR Research, investments in analytical CRM applications will grow at nearly double the rate of operational CRM systems. The analytics market is anticipated to expand to $4.4 billion, or 19 percent of the total CRM market, by 2005. SAS is readying a new version of its analytics software for release later this month that will make it accessible to a wider variety of user types.
Coleman and Lynch have come out with the Dance Card for the past six years, but the pair used historical data to go back 10 years and do the predictions. So far, they have a track record of being 94 percent accurate when determining which teams will participate in March Madness.
The question is, if they can predict who will be in the tournament, why not predict who will win? After all, March Madness trails only the Super Bowl in Las Vegas as the most-bet-on event, and think of all the NCAA "pools and friendly wagers" that go on at college campuses, offices, and among friends.
"There is a pattern as to who gets in," Coleman says. "There is a randomness as to who is going to win. The relationships we have set would quickly break down. That's why we haven't done it."
The same predictive analysis that Lynch and Coleman apply to Tournament team standings is relevant to business. "If you can predict whether an event will take place, you can predict whether or not a customer will come back, a product will be successful, and if a certain customer will purchase a specific product," Lynch says.
It's not just about the data, though. Lynch notes that statistics and human judgment compliment each other, and that one is not a substitute for the other.
"You have to have creativity and business insight," Lynch says. "Firms need to more cleverly mine data. Too many are just gathering piles and piles of data. That does not replace insight, thinking, and perspective. The firms that are doing this have developed an advantage."
Businesses need to approach analytics with a good questions, and need to use the analytics to determine whether the question or hypothesis is correct or not, Coleman says. "You can't just use conventional wisdom. Sometimes it's just conventional and not wisdom."
Coleman and Lynch agree that several factors, such as the significantly lower price point of analytics packages and easier-to-use software, are allowing for a wider range of users to take advantage of analytics.
"Before it was just too costly to answer certain questions, or you had to have a Ph.D in statistics to get answers," Lynch says. "Now, many people like marketing executives can ask the questions, and get the answer, all the while knowing that the data is there to back it up."
Coleman says that too many businesses rely on historical data only, which leaves out too many factors: "Things are constantly changing. Historical data provides a lot of information on what will happen, but it can't do any anticipation of wide scale changes. So, firms might be only looking backwards, and there needs to be a balance between looking back and looking forward."