• by Steve Smith , Staff Writer @popeyesm, February 22, 2011

Optimizing video ad campaigns effectively lifts key performance metrics, but even when experienced humans do it manually some inefficiencies are inevitable. Ad marketplace SpotXchange boasts that its new automated optimization engine, cleverly named OTTO, is already showing that a well-tuned robot works wonders in a new age of real-time bidding where nano-second decisions are required to serve highly individualized impressions. "With 30 to 40 million calls into our system every day you can't expect humans to evaluate each and every call in a marketplace to determine whether it is the right placement," says Michael Shehan, founder and CEO of SpotXchange. The RTB system in place at SpotXchange can target individual impressions based on audience segments so an ad call from the same publisher on the same video content can be different moment to moment. Even days after a campaign has been well optimized by human managers, he says, because audience exposure and other parameters may have changed that quickly.

OTTO watches campaigns and learns from past performances for a specific placement, a certain publisher or even a segment. It learns for other campaigns and then from real time feedback from the currently running campaign. "It steers campaigns towards higher performing audience segments and publishers and away from less engaging and lower performing traffic for a campaign that generates less waste," Shehan says. The system was tested recently on a travel advertiser who was using CTR to define the overall campaign goal. "Before OTTO was in place, our team was doing a good job to achieve a .78% CTR, which is a little below network averages. On day one with OTTO it achieved about the same of about .75%. On days 2 and three OTTO's learning started taking over to determine which placements were working better and it got a .87% CTR. On days 4 and 5 it got a 1.48% CTR."  

Of course with any ad exchange that claims heightened efficiency, the question arises, "efficiency for whom?" Is the job to deliver better rates in the end for the advertiser or for the publisher? Shehan insists the algorithms in OTTO reward effective inventory with strong pricing and incentivizes publishers to direct better inventory to the exchange. "They can see the effective CPMs go up," he argues. "It is hard to achieve optimization both for the client and the publisher," he admits, but the inherent inefficiencies in the manual optimization system was producing waste that skewed bidding. "We found the highest quality publishers were being punished by the lowest quality publishers. If a publisher or an audience segment is generating higher CTRs then OTTO will change the bid rate to reflect what they should be getting." Theoretically an engine like OTTO could be built to optimize SpotXchange's own margins. But Shehan says that is exactly what they didn't want to do. We wanted to built it to pay publishers the most they can be paid. We want them to send as much of their inventory to the exchange to compete with the networks at their higher net CPMs and deliver more healthy partnerships than they have with the ad networks."

For the buy side he says the system has been able to increase effectiveness. "We see much higher click volumes with OTTO in place," he says. But ultimately, Shehan says he would like to see some of the video ad buying, which now tends to mimic the methods and assumptions of TV spot buying, embrace the deeper analytics online targeting and optimization can provide. "We would like to challenge the assumptions and randomize the delivery of a campaign with all of its creative and evaluate the placements against the marketplace." He thinks that by applying performance metrics against some of the broad audience assumptions TV spot buyers make, online targeting could uncover unexpected and promising segments to target.