Banner advertising: Measuring effectiveness and optimizing placement

The article describes how one company improved banner advertising response rates by taking advantage of the medium's rich data to optimize placement. The study identified Web surfers who were frequent visitors to the banner advertiser's site. It then identified 100 Websites that these surfers tended also to visit. These sites were cluster analyzed to yield site genre definitions (affinities). In this manner a model was built to identify a group of affinities whose visitors were disproportionately likely to respond to banner advertising. These predictions were tested by placing banners on sites forecast to perform well. The average cost per response was nine times lower for sites predicted to belong to high affinity groups than low groups.