Twelvefold does all of its display buying through AppNexus, but before it transitioned to video in early 2016 the company was spending several million dollars a year to maintain a custom-built solution that attempted to model data signals out of AdX inventory.
Budzik and his team have since been able to shut that project down, save the cash Twelvefold was expending on maintenance and assign the engineers who had been working on it to other projects. They’ve also cut video campaign execution time by half, Budzik said.
“There really is a benefit to having one platform doing all your display buying and also having video as a capability in the same ecosystem with viewability, which is top-of-mind for advertisers,” Budzik said.
In March, Twelvefold, a member of Integral Ad Science’s certified viewability partner program, started offering its clients a 100% in-view guarantee on display. Video is far more nascent than the display space, so guarantees aren’t on the table, but Budzik said he’s able to use AppNexus’ video viewability solution to assure clients when their video viewability and completion rate goals are achievable at particular prices.
Of course, that doesn’t obviate the need for third-party verification partners. AppNexus still works with Moat, DoubleVerify and Integral Ad Science.
“Many agencies require them as their system of record and we respect that,” Hoffert said.
Although the “fox can’t guard the henhouse,” said Budzik, Twelvefold is able to use AppNexus’ viewability solution to help plan and optimize tricky campaigns. In some cases, the data Twelvefold uses for targeting is highly proprietary.
Say a food-related client wants to target users based on an interest in gourmet food or an interest in meat-and-potatoes-style cooking, and it turns out that people who like casserole have a far better video completion rate than people who go for haute cuisine.
“We can’t see that with third-party reporting, but when all of the video and viewability metrics are baked into one platform where I also have all of my data, I can see differentiated performance based on content signals,” Budzik said. “That lets us do optimization using our data, whereas before it wasn’t really possible. ”