The industry has made striking advanced in tracking the value and attributes of biddable online ad inventory, but the road ahead is long. That much was clear at the Interactive Advertising Bureau's Ad Operations summit in New York.
In a discussion of inventory packaging and pricing, Neal Richter, chief scientist at Rubicon Project, and Art Schram, VP of Glam Media's Adapt Platform outlined the challenges facing publishers. "There isn't a Yellow Pages for inventory and it's a problem of packaging and forecasting for real-time bidding," Richter told attendees at the New York conference on Monday, suggesting that the kind of categorization and clarity available in an old media environment like a large directory is missing for online.
Richter pointed to some hopeful signs, offering an example of "old RTB" and what is starting to be represented by the "new RTB," though he noted that the advancements are shining a brighter light on what needs to be fixed.
"The assumption of 'old RTB' is that impressions would be sent out with fairly poor descriptors, but at the end of the day, we will still get bids back and the highest bidder always wins -- that's it," Richter said "The impressions are something that the industry sells for cents per thousand, but we only auction off individually. 'New RTB' is about having the extension to identify not only the buyer clearly, not just the advertiser, but the order ID as well. The objective of the buyer and the seller is to fulfill the order, not just select the winning bid. That idea can be lost if you’re always selling to the highest bidder, which is really beside the point."
Forecasting is another issue. If you think about traditional forecasting models in ad servers, they’re based on defining ads in an ad platform, knowing how they’re targeted and accounting for their future delivery against available inventory, Schram said. Even with all the greater efficiencies provided by ad tech, a programmatic call is generally only a single object in an ad platform, representing a lot of different possible campaigns, with diverse targeting, so the traditional model becomes more difficult to implement, and a more approximate approach has to be taken. These same “expectations” are largely what drive prioritization decisions, as opposed to full, complete data telling the publisher exactly what to expect in terms of inventory fill.
The notion of "Nirvana" for a forecasting system is when a publisher has a dashboard and advertisers and platform operators can just start selecting targeting criteria for expected avails at clearer price points, Richter added.
"The problem from our perspective is that we already have that," Schram responded. "We just don’t know how to build in the notion of ‘booked’ into that context."
At this point, Richter described the current playing field for programmatic pricing and packaging as looking like a game called "Blind Auction Wars," where there is a ton of random products available for various prices -- but no one knows exactly what's being offered for bids.
"Imagine you’re going into Wal-Mart and all of the products are in brown cardboard boxes with the price posted," Richter said. "That’s it. And the only way you have to differentiate whether you’re getting shampoo or cereal or a loaf of bread is the approximate size of the thing. Deal ID and the Yellow Pages problem are about having a well-lit store where the inventory has more description than a low-placed moniker."
Publishers do have to address this, but overall, this is a problem for the industry, Schram said. The connective tissue between supply and demand are specifics about the nature and value of the inventory being sold. "As publishers start to address this, that will force the demand side platforms on the other side to respond in kind," he said.