First-Price Auctions and Bid Shading: What Changed for Publishers

When Google moved AdX to a first-price auction in 2019 — following the exchanges — it permanently changed publisher yield strategy. Understanding the mechanics explains why dynamic floors went from nice-to-have to essential.

Second price vs first price

In a second-price auction, the winner pays just above the runner-up's bid; bidding your true value is optimal, and floors mostly matter when bid density is thin. In a first-price auction, the winner pays exactly what they bid — so every rational buyer bids below their true value. That gap between value and bid is bid shading, and every major DSP runs shading algorithms that learn, per publisher and per segment, the minimum bid that still wins.

The arms race

Shading algorithms probe your inventory constantly: bid lower, still win, lower again. Left unopposed, they ratchet your clearing prices down to whatever your floors allow. The publisher's counterweapon is the floor: a well-placed floor forces shaded bids back up toward true value. This is why floor optimization directly fights an adaptive adversary — and why static floors lose. The DSP updates its shading model daily; your quarterly floor review never stood a chance.

Practical consequences

• Floors are now offense, not defense: they actively recover shaded value rather than just blocking junk demand.
• Granularity wins: per-segment floors (geo × device × unit × demand channel) beat blanket rules because shading itself is per-segment.
• Transparency in reporting matters: monitor win-rate and clearing-price distributions; a healthy floor sits where the bid distribution thickens, not at a round number someone liked.
• Holdout measurement is essential: in an adversarial system, only a control group proves your floors are netting more revenue rather than just higher eCPM on lower fill.

First-price economics rewards publishers who treat pricing as a continuous algorithmic discipline. That's the entire thesis behind AI floors.