Bid shading is a popular programmatic advertising strategy, but created as an unhappy compromise between first and second price auctions.
Advertisers are leaving at least $6.6 billion in annual savings on the table by relying on bid shaders, according to Aaron Andalman, co-founder and chief science officer at Cognitiv.
“It was intended as a stop-gap measure and has become the standards,” Andalman said. “Advertisers are not generally aware they’re using it or how much it costs. There’s a lot of risk in bid shading because the advertiser gives up a lot of control of the bid. Instead of the advertiser pricing it themselves, they allow a tool to choose how much to bid for them.
Andalman believes there is a better way. He calls it inventory forecasting. As a Stanford graduate with a PhD in neural networks from Massachusetts Institute of Technology (MIT), he is knowledgeable about the process supported by artificial intelligence (AI) and deep learning.
Cognitiv’s technology -- inventory forecasting -- looks out into the future to similar auctions to adjust the price. The bidding process is changed based on that forecast. Machine learning and AI technology are used in the process.
There are lots of ways to run an auction. Second-place auctions were shown to be a more efficient way to run an auction from the buyer side. The buyer can bid its worth and end up with a reasonable deal with the seller.
“If an advertiser gives up the choice of a bid to a tool, bid shading, they better understand what the tool does and what they give up when making the decision,” he said. “Even if the tool is being used to align with the advertiser’s goals, bid shading is only one tool to think about the guess that others will bid.”
Digital ad buyers cannot agree on what bid shading does. In a recent survey conducted by Cognitiv, 33% say it is a tool to adjust their bids for a first-price auction, 32% say it is an algorithm that optimizes win-rates and CPM, 22% say it is a tool that manipulates bids so they pay less, and 12% say it just adds another fee to their bid.
Advertisers seem to be playing a guessing game. Bid shading, in theory, lowers the bid based on what you think others will bid, but not lower than what others do bid. First price auctions are less efficient in the sense that the advertiser can’t just say how much the piece of inventory is worth and place the bid. The bidder must determine it’s worth and then bid at least one cent higher than what the advertiser thinks others will bid. Otherwise, the advertiser spends more than it needs to spend.
The demand side platforms (DSPs) and SSPs created the tool called bid shading, but Andalman's research shows not only do advertisers not understand how it work, they use it as a crutch.
Cognitiv developed a patent-pending process to forecast inventory and then base the bid on that. “Let’s say you’re bidding on a necklace, you ask yourself, how much is it worth to me,” he said. “If there were a thousand other similar necklaces you could bid on sometime in the future, it wouldn’t be as worth as much to you.”
Bid shading algorithms focus on one auction at a time to determine how much to shade and then consider other factors. It’s now been more than four years since first-price auctions and bid shading became standard practice for the majority of programmatic media buys.
The same challenges related to transparency continue to plague the industry. Advertisers, based on Cognitiv's research, do not trust the process and question the validity of their buys or they remain unclear as to how the auction system works. With this status quo, it is unclear whether programmatic advertising can continue to be efficient or if it can provide transparency to advertisers.
Intelligent bidding prices inventory before it even gets to the point that a bid shader would be applied Andalman says intelligent bidding is based on performance and not just reactionary bidding.