A price floor is the lowest bid a publisher will accept for an impression. Setting one looks simple and is one of the more consequential decisions in programmatic selling.
Floors exist because bids are not offers of value
In an open auction a buyer bids what an impression is worth to that buyer, which may be far below what it costs the publisher to produce and serve.
Without a minimum, inventory would clear at whatever the weakest demand was willing to pay, and the publisher would have no way to hold a price.
The floor is the publisher's only lever for refusing a sale, since it cannot negotiate individually across millions of automated transactions.
The trade-off is direct and measurable
Raising a floor removes every bid beneath it, which increases the average price of the impressions that do sell and reduces how many sell at all.
Whether that is profitable depends on the shape of the bid distribution, specifically how much demand sits just below the proposed level.
Where bids cluster tightly, a small increase can eliminate a large share of demand. Where they are dispersed, the same increase costs almost nothing.
Floors influence bidding behavior, not just outcomes
Buyers observe the prices at which they win and lose, and their bidding systems adapt, so a floor teaches demand what it must pay to access the inventory.
A floor set slightly above the usual clearing price often pulls bids upward rather than removing them, because the impression remains worth having.
A floor set far above value simply moves the buyer elsewhere, and the publisher loses both the impression and its place in that buyer's future planning.
Dynamic floors replaced fixed ones
Fixed floors are blunt, because the value of an impression varies with geography, device, page, time and how often that user has already been reached.
Dynamic systems set the minimum per request using recent bidding history for similar inventory, capturing more value from strong impressions without blocking weak ones.
The cost is opacity: buyers see a price that moves without explanation, and disputes about apparent inconsistency become part of the relationship.
The floor interacts with everything else in the stack
A floor applied at one point in the chain but not another produces confusing results, since some demand paths are filtered while others are not.
Floors also interact with what fills the leftover space, because an impression rejected on price becomes an unsold impression that must still be filled with something.
Publishers who evaluate floors on average price alone therefore miss half the equation, which is the value of what runs when nobody meets the price.