Dotomi Rolls Out Multivariate Testing for Display Ads

Web ads are getting smarter and smarter. Dotomi, a Chicago interactive agency, announced a new multivariate testing methodology yesterday for banner ads. It follows Web users around to determine which ads ultimately work best — not to make users click, but to persuade them to buy.

Dotomi COO Ken Treske says clients saw anywhere from 5 to 15 percent improvement in customer conversions, on average, by incorporating the methodology into their campaigns, and some saw bumps as high as 47 percent.

The company works with clients to design banners with several variations of several variables, such as colors, fonts, button shape, specific wording, offers, and call to action. Testing many variations at once allows the tester to pinpoint not only which combinations work the best, but the effect of each specific element.

Dotomi combines the testing methodology with a proprietary “personalized” approach that uses cookies to track certain customers from the client’s site across their Web travels. Attaching geographic and demographic data to those users, the company can track which ads each test user sees and whether a sale ultimately results. “We can observe how we’re able to change the behavior of different individuals, and the relation of a near-term click to the person’s long-term purpose,” Treske said.

“It’s very easy to create survival of the fittest around clicks because they happen in the moment and it’s easy to see, but the client’s end goal is how messaging ultimately influences consumer behavior to make a purchase,” he added. “The conversion cycle can take days or weeks or months, and you don’t want to narrow down to an ad that leads to short-term clicks but not to a sale. We think our big innovation is applying multivariate testing principles to display advertising at an individual impression level, and making that a consistent part of a campaign on an ongoing basis.”

Privately-held Dotomi, which doesn’t publish its client list but claims more than 100, concentrated among large consumer brands, is in the first wave of what will probably be standard practice throughout the online advertising industry within a couple of years, says Michael Andrew, director of search and analytics at the San Francisco agency Mediasmith. He points to the Yahoo! SmartAds program , which uses the mass of data collected on millions of Yahoo users, as well as third-party data sources, to serve ads dynamically. The first two third-party partners in the program, Tumri and Teracent, also promise tightly targeted ads using their own proprietary technologies.

Dapper, another contender, pulls live information from the advertiser’s Web site. The analytics company Fetchback allows advertisers to retarget recent visitors with new offers to lure them back to the advertiser’s site.

“All of these offering are relatively new in the market, and we’re talking to a lot of companies who can do customizations based on all kinds of data,” said Andrews. “They’ll optimize on whatever action you think is important, whether it’s seeing a movie trailer or getting to a shopping cart. It’s still new ground and that’s why there are all these players emerging. We don’t have a concrete judgment on what’s better. You don’t know what these algorithms are like until you’ve worked with them.”

Related reading