Hearst Magazines announced the winners for its second annual Hearst Challenge last week. This year’s challenge, which was designed as a competition to build a model that predicts an individual’s likelihood of opening and/or click through rate of a given promotion, gave away more the $25,000 in prize money for the winners.
“Ultimately for us, the challenge at Hearst is really how do we leverage and bring the best thinking of the world to address the basic problems we face in the business—the Hearst Challenge continues to be a great source of innovative thinking,” says Charlie Swift, vice president of database strategy and marketing for Hearst Magazines. “It’s a public thank you to the industry and a recognition that a lot of people out there can help you solve your problems and they’re a great resource to have.”
The grand prize of $25,000 was awarded to U.S. based company Global Decision, the company’s team was led by Jaysen Gillespie. The second winning team—EURO RSCG Discovery, also from the U.S. and led by Ward Thomas—got a prize of $5,000 for presenting the most innovative solution.
“Global Decision was picked because they presented the lowest amount of error in predicting open and click through rates,” says Swift. “The winner chosen is the most formulaic, we take out judgment and say ultimately, ‘Did their model work the best when compared to all of the other folks that submitted their entries?’ The model from the folks at Global Decision had the least amount of errors which is why they were the winners.”
Swift says the judging criteria for EURO RSCG Discovery was more subjective.
“They got the innovation award,” he says. “Yes, their model had to be predictive—so we looked at the top 10-20 range submissions that were in the realm of winning—but then we went through and said ‘Is someone doing something innovative in their approach to their thinking?’ What we mean by that is have they presented a new technique that we could learn and leverage in future modeling and analytics? It was the creativity we were looking for.”
More than 500 entries from 51 countries submitted materials for consideration. The results were announced at the Direct Marketing Associatio’s 2011 Conference in Boston, MA.
“The panel felt that the modeling technique and the element of innovation they felt was most pronounced was the clever approach of what they call forcing in demographic variables,” says Swift. “In historic modeling, everyone knows that the past is the best prediction of the future, but in many cases you’ve got a lot of people who have limited past so you must break the model apart and force in additional information to deal with the new segment. Their ability to force in demographic variables was one of the basses of the award.”
Swift adds that the group was able to simplify the equation and reduce its complexity to amplify success. The algorithms provided from the challenge will, going forward, allow Hearst to test theories of actions that the company takes.
“Now, if we take action X, Y or Z we can try and determine what the impact will be on the behavior of our customers—if we increase or reduce frequency, what will the impact be on the likely response? If we change the timing of the email, what will that do to the response?” says Swift. “Now we can come up with simple guidelines and rules of how we execute our strategy—we want to apply these models to help teach us the ramifications of some of our choices so we can make better choices. The way you change the response rate is by all these strategies.”