• by Ken Liebeskind , May 24, 2002

Trying to determine which banner ads to use in an online campaign is a tedious task. Not only do you have to decide which combination of offers, creative, background, and calls-to-action will be most successful, you also have to determine which subsets of your viewing audience will respond best to each different version of an ad.

Poindexter Systems has introduced the Progressive Optimization Engine (POE), an ASP-based service that uses modeling and other mathematical techniques to uncover viewer clusters and determine which banner ads work best for different clusters.

POE collects a variety of data on each visitor to a website where an ad is located, including bandwidth, browser, frequency, daypart, and geographical and demographic data. The data comes from the viewer’s browser and third-party data providers Claritas and Quova, as well as Poindexter, and can also be retrieved from publishers and advertisers.

Once data is collected on a viewer, POE determines in real time which cluster the viewer falls into and delivers the ad that has garnered the best response from other viewers in the cluster. “We find out which ads appeal to which cluster to boost campaign performance,” says Joe Zawadzki, Poindexter's chairman.

The idea isn't just to find the overall top-performing ad in a campaign, but to find out which ads work best for different clusters of viewers in different parts of the country, on different sites, at different times of day, and with different browsers. POE can make all these determinations.

Poindexter POE has been used by AT&T, Terra Lycos, Purina, eBay, NextCard, Mercedes-Benz, and Hotwire. Amy Bohutinsky, a Hotwire spokeswoman, says the company stopped all advertising after September 11, then began testing online creative a month later. “Poindexter enabled us to determine which ads worked best and for what segments they worked best for,” she says.

The company tested 49 banner ads on Ubid.com, About.com, Juno, and IndependentTraveler.com. Five ads – mostly bright red with brief sales messages – performed the best across different viewer clusters.

The ads optimized by Poindexter POE had a 152 percent higher click through rate than the control group.

Bohutinsky says Poindexter POE not only increased the performance of the campaign, but it was also an effective learning tool that helped Hotwire figure out the best color and sales messages to use in its ads.

Poindexter POE can optimize on any metric that an advertiser requests, including clicks, registrations, conversions, and other post-click actions – alone or in combination. Poindexter POE integrates with an advertiser's existing ad server. Advertisers need a variety of creative treatments and can run a test with Poindexter to validate effectiveness.

POE is available for a monthly fee, with additional CPM charges.