ROXOT BLOG
How The First-price Auction Model Devalues Publisher's Inventory
Continuing the discussion on auction dynamics in RTB
In our recent article "A Publisher's Revenue Lost to the Second-Price Auction Game" we discussed the crucial limitations of the second-price auction model and how they affect a publisher's position in the market. Because advertisers are able to game the system and employ bidding strategies that result in low bids and/or huge gaps between the winning and second-highest bids, the current state of the industry is proving untenable. The question is where the industry will look?
The answer many industry observers are suggesting is the first-price auction model where bid optimization by advertisers is acknowledged and accepted. They argue the market will become more straightforward and predictable as buyers acquire impressions by simply paying the price of their bid.
However, the first-price auction model isn't without its flaws. Natural market tendencies would encourage advertisers to try and bid as little as possible and still get the impression. They would employ aggressive bid strategies to try and win with a bid that's ultimately only $0.01 higher than their estimation of the next advertiser's bid.
For example, one of the most wide-spread strategies for uncovering the optimal ratio of CPM and number of impressions acquired is automated bid testing. The way it works is straight-forward: automated tools test bids based on user-defined bid increments, bid range, timeframe, and percentage of traffic. In practice, a buyer looks for the optimal ratio of price and amount of impressions inside a $0.05 - $2.00 bid range. She sets up automated bidders to test every bid in this range with $0.05 increments for 10 minutes, for 10% of auctions. In those 10 minutes, 40 different bids will be used to identify the one that generates the optimal number of impressions for the lowest price. The test result will be repeated during the next 10 minutes for 90% of auctions while the test repeats again on the remaining 10%. As a result, testing runs all the time adapting bidding strategies to market conditions every 10 minutes.
The advent of machine learning technologies in the first-price auction environment inevitably leads to the devaluing of a publisher's inventory. When buyers adapt to changing competitive bids in an automated fashion, the market price for an impression quickly drops. Such an imbalance between supply and demand results in an even more inefficient ad market and leads to restructuring. Despite being transparent and straight-forward, the first-price auction model is questionable in RTB as it may destabilize the market and devalue a publisher's inventory.