How Publishers Set CPM Floor Prices (and Why They Matter)

What a Floor Price Actually Is

A floor price is the minimum CPM a publisher will accept in a programmatic auction. It's not what most impressions sell for — in a healthy auction with real competition, winning bids land comfortably above the floor. The floor exists specifically for the impressions that would otherwise sell for less than they're worth, setting a line below which the publisher would rather not fill the impression at all.

You can sanity-check any floor against actual results using the standard CPM formula — divide realized revenue by impressions sold, multiply by 1,000, and compare that to the floor you set, using the CPM calculator:

Why Set a Floor at All

Without a floor, an auction with thin competition can clear at a price well below what the inventory is actually worth — a single low bidder in an otherwise quiet auction can win an impression for next to nothing. A floor protects against that scenario, guaranteeing that inventory either sells at an acceptable rate or doesn't sell at all, rather than sliding to whatever the weakest available bid happens to be.

Floors also give publishers leverage in a market with many competing demand sources. In header bidding setups, where multiple ad exchanges bid on the same impression simultaneously, a consistent floor across exchanges prevents one exchange from consistently underbidding the others and skewing where the impression ultimately sells.

The same logic applies to what gets billed at all — an impression that's technically sold but never actually seen is a distinct problem covered in CPM vs vCPM.

The Risk of Setting It Too High

A floor that's too aggressive doesn't just filter out low-quality bids — it filters out legitimate demand that would have cleared slightly below the floor but well above what the impression is actually costing the publisher in lost opportunity. The impression goes unsold, fill rate drops, and that revenue is gone entirely rather than captured at a lower rate.

This is the central tension in floor pricing: raising the floor increases the value of each sold impression, but past a certain point it reduces how many impressions sell fast enough that total revenue falls even as the per-impression rate rises.

Static vs Dynamic Floors

A static floor applies one fixed minimum CPM across a block of inventory — simple to set and reason about, but blind to the fact that demand for the exact same placement can vary enormously by time of day, device, geography, and audience. A dynamic (or algorithmic) floor adjusts per impression based on those signals, aiming to set a higher floor when historical bid density suggests strong demand, and a lower one when it doesn't.

Dynamic floors generally outperform static ones on total revenue, since they capture more value from high-demand impressions without pricing out the lower-demand ones that a single static number would have to compromise between. The tradeoff is complexity — dynamic floors require enough historical bid data to set intelligently, which smaller publishers with thinner traffic may not have.

A Worked Comparison

Consider 100,000 available impressions on a placement with genuinely variable demand:

• A $3 static floor fills 90,000 of those impressions at an average $4 CPM, for $360 in revenue.
• A $5 static floor fills only 60,000 impressions — the higher-demand share — at an average $6 CPM, for $360 in revenue: the same total, despite selling far less inventory.
• A dynamic floor might fill 85,000 impressions by setting a higher bar only where demand supports it, landing at an average $4.70 CPM for roughly $400 in revenue — capturing more value from strong-demand impressions while still selling the weaker ones.

These are illustrative numbers, not universal outcomes, but the pattern they show is real: the floor that maximizes revenue is rarely the highest one, and it's almost never a single fixed number applied uniformly across genuinely uneven demand.

Practical Tips for Setting Floors

• Start conservative and adjust — a floor set too low is a smaller loss than one set too high; it's easier to raise a floor gradually than to recover fill rate lost to an overly aggressive one.
• Segment by real demand differences — even without full dynamic pricing, splitting floors by device, geography, or placement type captures much of the same benefit as a fully automated system.
• Watch fill rate alongside CPM — a rising average CPM paired with a falling fill rate is a signal the floor may be costing more in unsold inventory than it's gaining in price.
• Keep floors consistent across header bidding partners — inconsistent floors across exchanges create a bias in the auction that has nothing to do with which exchange actually has the best demand.
• Revisit floors periodically — demand for a given placement shifts with seasonality and market conditions, and a floor set once and never revisited gradually drifts away from what the inventory is actually worth.

Frequently Asked Questions