Gaylord described how the New York Times uses predictive insight, based on web site traffic the prior day and other factors, to determine how many papers it should print. He shared an example of how the intuition of the editors following a recent big news event suggested printing between 35,000 and 40,000 newsstand copies. The predictive algorithm the newspaper developed suggested 43,000. The actual demand was 45,000.

Ayres said online publications can improve customer satisfaction – and save money -- through randomized testing of content. But he warned against organizations spending too much time determining the content to test because it can sap the savings gained by randomized tests. He also suggested testing as many as 20 different creative treatments for ads or content. Sometimes marginal ideas have traction with customers or readers, he said.

Ayres would like to see analytics dashboards that not only show trends, but provide predictions about the certainty or accuracy of the trends. Gaylord said he’s most interested in dashboards providing the nuggets of insights he needs to improve his marketing. He wants to avoid dashboards with “data about data.”

To leave comments, please login or sign up to the Engage Network at the top of this page.