Are We Overcomplicating Floor Pricing Optimization?

Discover how behavioral economics offers a simpler, more effective approach to floor pricing optimization. Kean Wang, VP of Product and Strategy at Intowow, reveals best practices for balancing Header Bidding and Google Ad Manager to maximize publisher revenue.

Floor pricing optimization is making waves again in the publishing world, but have we been overthinking it?

Kean Wang, VP of Product and Strategy at Intowow, dives into the evolution of floor pricing strategies and unveils a refreshing shift from complex mathematical models to the practical realm of behavioral economics. By understanding bidder behaviors and leveraging the strengths of Header Bidding and Google Ad Manager, publishers can streamline their approach and boost revenue without getting lost in the computational weeds.

Floor pricing optimization has regained popularity among publishers. Over the past five years, we have perfected our dynamic floor pricing algorithms. However, it wasn’t until a year ago, as we gained access to more data from major publishers around the world, that I realized we might have been approaching this problem unnecessarily.

For quite some time, we adopted a purely mathematical approach to floor pricing optimization, focusing on determining each price P that maximizes the RequestRPM according to the function 

RequestRPM(P) = SellThroughRate(P) × eCPM(P)

for each infinitesimally meaningful inventory segment. If demand is static, the calculations are straightforward and manageable. However, randomness introduces uncertainty in a more realistic scenario where hundreds of thousands of campaigns run concurrently and complete at different times. This uncertainty significantly complicates the computational process and requires intensive predictive modeling to find an optimal solution.

It turns out that an easy way out is to approach the problem from a completely different discipline – behavioral economics.

A Simple and Elegant Approach

In the real world, campaigns are managed by DSPs who bid for impression opportunities in auctions. These DSPs vary widely in their technical capabilities and operational strategies. So, if we could target the behaviors of different types of bidders and provide the right incentives and signals to facilitate communication and competition among them, we could reach an alternative solution that is more elegant. This approach allows the market to optimize by itself to maximize publishers’ benefits without too much interference and the need for excessive calculations.

Generally speaking, bidders buy through two open auction channels: Header Bidding and Google Ad Manager (GAM), both of which are extensively integrated by most publishers. By analyzing bidding behaviors across these channels, we have consolidated the following best practices:

  1. For Header Bidding, set up floor prices low enough on SSPs to encourage bid tendencies but high enough to filter out low-quality ads.
  2. On Google Ad Manager, use the winning Header Bidding bid prices and dynamically trigger Unified Pricing Rules (UPRs) to provide competitive price signals through Google Ad Exchange and Open Bidding.
  3. Ensure that Header Bidding line items are correctly priced on GAM with your net earnings to facilitate an efficient unified auction.

These best practices take advantage of a key behavioral distinction between these two channels of bidders: 

Google Ad Manager bidders, predominantly Google Ads and DV360, primarily adjust their bids based on floor price signals, whereas Header Bidding bidders make extensive adjustments in response to changes in win rates.

Excessive floor prices do not stimulate Header Bidding bidders; instead, they block their bids and reduce competition. By allowing more Header Bidders to participate in auctions, we maximize the competitive bid signals sent to GAM. On GAM, triggering these signals with UPRs can restore the “last look” advantage for Google bidders, encouraging higher bids that benefit publishers. (Google’s decision to cancel this feature was due to pressure from other SSPs, but this move also negatively impacted publisher revenue.)

With more competitive bids from Google bidders taking over some winning opportunities from Header Bidders and driving down their win rates, Header Bidders are incentivized to adjust their bid prices, which in turn encourages more competitive bids from Google. This fosters a perpetual cycle of healthy competition across these two channels.

For publishers with extensive Header Bidding coverage, these best practices are generally sufficient. Beyond this, additional efforts would likely yield only marginal benefits unless you are determined to invest intensive R&D to further optimize price ranges specifically for Google bidders across each traffic segment, where the benefits could add up to be significant.

More About the Behavioral Distinction

Upon further research and some reverse engineering, we were able to gain a clearer understanding of the factors contributing to such a behavioral difference.

For DSPs, floor prices are one of the pre-auction signals useful for optimizing bid decisions to maximize campaign ROIs. However, for floor prices to serve as reliable indicators, the environment must meet three criteria:

  1. Floor prices, along with the associated traffic metadata, should be consistently supported across all stakeholders.
  2. Floor prices must maintain their integrity during transmission and should not be lost, overridden, or manipulated down the supply path.
  3. To benefit from these signals, bidders must possess powerful predictive capabilities with the technical bandwidth to perform cost-efficient real-time calculations.

Only Google, with its unified and streamlined programmatic ad supply path, meets these criteria across all stages. Information from publisher web pages is collected by the standardized Google Publisher Tag (GPT) library, consistently formatted as ad requests, and transmitted to the centralized Google Ad Manager. From there, bid requests are uniformly assembled and sent to DSPs via a robust server-to-server connection using the Authorized Buyer Real-Time Bidding Protocol.

On the buy-side, Google Ads and DV360, operating with powerful predictive capabilities leveraging Google’s highly integrated cloud infrastructure, can accurately estimate ad performance using real-time client-side signals and determine reasonable bid prices for each impression opportunity before an auction occurs. 

In contrast, for Header Bidding, every bid request is processed by at least two entirely separate parties (e.g., publisher-hosted Prebid.js, vendor wrappers, or SSPs) before reaching DSPs. Even for large DSPs with strong predictive capabilities, such a fragmented supply path makes it difficult to ensure the integrity of information, forcing them to downplay the option of adjusting bid decisions based on real-time sell-side signals. A rather reliable source of information is win rate data, which each bidder processes post-auction but is often delayed and lacks granularity.

These disadvantages of Header Bidders are particularly evident when we compared bid CPM trends from early to subsequent ad refresh instances or across different ad position series. For example, for the same inventory, when comparing the CPM of the first ad to the fifth ad refresh, average bid prices from Google bidders can drop by over 50%.

However, for Header Bidders, the average win bid CPM only decreased by 3%, which falls within the margin of statistical error. Such inability to perform per-impression bid adjustments from Header Bidders can also be highlighted by the fact that, on average, they purchase the same inventory with the same level of CTR and viewability performance at approximately a 35% premium compared to Google bidders.

To further pinpoint the issue, we took the same Header Bidding bidder and compared its bidding behaviors across two different channels: Header Bidding and Open Bidding (a unified server-to-server bidding solution provided by Google).

Under the same floor pricing strategy for the same bidder, through Open Bidding RequestRPM could be improved by 8 to 10%, compared to only a marginal 2% improvement through Header Bidding. This suggests that a fragmented supply path is the primary factor preventing per-impression bid adjustments for Header Bidders, forcing bidders through this channel to forgo floor prices and focus on win rate signals instead.

Future Outlook for Floor Pricing Strategies

Floor prices, like other real-time client-side signals, provide valuable information that encourages bidders to recognize the fine nuances in publisher inventory. However, these signals can only be effective when information integrity can also be guaranteed across the entire supply path.

The above behavioral distinctions between Google bidders and Header Bidders underscore the importance of Supply Path Optimization (SPO) and demonstrate how a more streamlined supply path can encourage bidders to utilize more sell-side signals, ultimately improving efficiency across the industry.

But for now, with these simple yet elegant best practices, publisher ad inventory is effectively categorized into two groups: one with extensive competition from both Header Bidders and Google bidders, and the other with only Google bidders. Publishers with more inventory in the first category can significantly benefit from the competitive cycle facilitated by these simple steps.

However, for publishers whose inventory primarily falls into the latter category, an active floor pricing strategy using a traditional mathematical approach is still necessary to realize the huge growth potential from Google demand.