Amazon Bull Case Tested by $20 Billion Ad Overcharge Allegation

The history of advertising is a history of unmeasured waste. We are currently witnessing an dispute at Amazon (AMZN) that highlights the fundamental tension of modern digital platforms: the discrepancy between the auction mechanics promised to the advertiser and the reality of the price paid ⁸. The Federal Trade Commission alleges that Amazon’s Sponsored Products, long framed as a generalized second-price (GSP) auction, instead utilized an undisclosed “soft reserve price” ^10,46. The question is not whether it works, but how you know it works.

The Attribution Disconnect

For years, Amazon signaled to its millions of advertisers that it operated a standard GSP auction ^{1,2,12,29,30,39}. In this model, the winning bidder pays only one cent more than the runner-up ^{23,24,26,37,38,40}. This design theoretically encourages participants to bid their true maximum value, confident that they are protected by the second-price floor ^10,34,38,41.

The FTC’s evidence suggests a different reality: a “proxy 2nd price” system where the winner often pays their own bid ^{15,23,24,31,32}. Internal documentation refers to an “invented auction participant” utilized to artificially elevate prices ^19,23,35, a mechanism functionally equivalent to a shill bid ⁴⁶. The consequence of this undisclosed floor is clear: when the auction price is no longer set by a competing bidder, the nominal auction result and the actual cost-per-acquisition diverge ^5,6.

Escalation and Economic Impact

This is not a matter of a single technical pivot, but a shift measured in years ¹⁰. While only 4% of advertisers were charged their full budget cap in 2020 ¹³, the FTC alleges that this figure rose to 30–40% in 2021 ^10,25,46, and reached roughly 80% by 2024 ^{14,15,20,23,25,27,41,44,46}. The cumulative overcharges attributed to this practice exceed $20 billion since 2019 ¹¹.

Amazon contests this narrative, pointing to metrics such as a 50% decline in average winning bids from 2019 to 2024 ^{30,32,33,36,42} and flat inflation-adjusted cost-per-click ^{2,3,10,21,27,28,33,37,39,43,45}. Yet, the industry currently lacks the transparent audit protocols required to validate these competing claims ^21,42. We must ask: if the auction was functioning as described, why would internal documents associate disclosure with a “downward spiral” of bids ^37,44,46?

The Operational Reality for Sellers

For the merchant, the arithmetic is unforgiving. High-intent paid placements frequently displace the algorithmic “best match,” which often resides at position 17 ^4,22, while top-row placements carry a 25–29% price premium ^4,22. Every incremental increase in CPC flows directly into the seller’s customer acquisition cost ¹⁰. With margins often razor-thin—FBA-based selling nets roughly 7% at scale ⁷—the integrity of the auction is not just a regulatory concern; it is a direct constraint on business viability ¹⁶.

To manage this, the prudent operator must prioritize TACoS (total ad spend over total sales) and “true landed cost” ^9,17,18. Until transparent, third-party verifiable auction mechanisms are implemented, the only defense is rigorous, independent measurement ⁴². What is the true ROI when the auction itself is a black box?