Price floors are one of the few yield levers a publisher fully controls, and one of the easiest to get wrong. Set them too low and buyers pay less than they would have. Set them too high and impressions go unsold. This post explains the difference between static and adaptive floors, the tradeoff they manage, how to test them properly and the mistakes that quietly cost revenue.
What a price floor actually does
A price floor (also called a reserve price) is the minimum bid a publisher will accept for an impression. Any bid below the floor is rejected, and if no bid clears it, the impression goes unfilled or passes to a fallback source.
Most programmatic exchanges now run first-price auctions: the winner pays what it bid. That changed the role of floors. In the older second-price model, a floor mainly lifted the clearing price when there was only one strong bidder. In a first-price world, buyers use bid shading (algorithms that lower bids toward the lowest price likely to win), so a well-placed floor can stop shaded bids from dropping below what the impression is really worth to that buyer.
Floors don’t create demand. They only decide which existing bids a publisher is willing to take.
Static floors and why they drift out of date
A static floor is a fixed value set by the ad ops team, often per ad unit, country or device type. It’s simple to understand and easy to audit. It’s also a snapshot of the market on the day someone set it.
Demand changes constantly. Seasonal budgets arrive and disappear, individual buyers enter and leave, and the value of an impression varies by page, time of day, user context and format. A static floor can’t follow any of that. Common symptoms include:
- Floors that were right months ago now block bids that would have been profitable.
- A single floor applied across very different inventory, such as a top-of-page video slot and a below-the-fold display unit.
- Teams that rarely revisit floors because every change feels risky.
Static floors aren’t useless. They’re a reasonable starting point and a sensible safety net. They just can’t respond to a market that moves by the hour.
How adaptive floors work
Adaptive (or dynamic) floors adjust automatically based on observed bidding behavior. Instead of one number per unit, the system estimates what each impression is likely to be worth and sets the floor accordingly.
The inputs typically include historical bid levels for similar impressions, which buyers tend to bid on which inventory, and context such as geography, device, placement, format and time. Predictive modeling turns those signals into a floor for each auction or for fine-grained segments of inventory.
The logic follows a simple idea. When the data suggests strong competition for an impression, the floor can rise to capture more of that value. When demand looks thin, the floor drops so the impression still sells rather than going unfilled.
The fill versus price tradeoff
Every floor decision trades fill rate (the share of impressions that sell) against price (what each sold impression earns). Raise floors and average price usually goes up while fill goes down. Lower them and the opposite happens.
Neither metric is the goal on its own. What matters is total revenue across all available impressions, usually tracked as revenue per thousand pageviews (RPM) or per thousand ad requests. A floor strategy that increases average CPM (the price per thousand sold impressions) but leaves many more impressions unsold can easily reduce total revenue.
There are also second-order effects to keep in mind:
- Buyer learning. Aggressive floors can train some buyers to bid less often on your inventory, which shows up slowly.
- Fallback value. If unfilled impressions pass to another demand source, the real cost of a missed sale depends on what that fallback pays.
- User experience. Unfilled slots that collapse or show house ads affect the page, which matters for layouts and page stability.
How to test floors properly
Floor changes are easy to misread because demand moves on its own. A before-and-after comparison can make a bad change look good during a strong week, or a good change look bad during a slow one. A controlled test avoids that.
- Split traffic, not time. Run the new floor logic on a random portion of traffic while a control group keeps the current setup, over the same period.
- Pick one primary metric. Use revenue per ad request or per thousand pageviews, not CPM alone.
- Watch the secondary metrics. Track fill rate, bid rate, number of active buyers and timeouts to understand why revenue moved.
- Segment the results. A change can help desktop while hurting mobile, or lift one country while dropping another. Averages hide this.
- Run long enough. Include full weekly cycles so weekday and weekend behavior are both represented.
- Roll out gradually. Expand the winning setup in steps and keep monitoring, since buyer behavior can shift after launch.
Pitfalls to avoid
- Chasing CPM. Higher average prices look good in a report while total revenue slips.
- Floors on thin data. Small segments with little bid history produce noisy estimates, so group them until there’s enough data to trust.
- Conflicting floors. Floors set in the ad server, in each demand partner’s settings and in a floor tool can stack or contradict each other. Keep a single source of truth where possible.
- Ignoring deals. Private marketplace and direct deal prices should be set deliberately, not accidentally overridden by open auction floors.
- Set and forget. Even adaptive systems need people checking outcomes, reviewing edge cases and adjusting guardrails such as minimum and maximum floor values.
This is why Vortex pairs adaptive floors with smart auction routing, predictive modeling and hands-on optimization from experts who review publisher setups, rather than leaving the configuration to a dashboard alone.
Key takeaways
- A price floor is the minimum bid you’ll accept, and in first-price auctions it helps protect value against bid shading.
- Static floors are a useful baseline but drift out of date as demand changes.
- Adaptive floors use bidding history and context to raise floors when competition is strong and lower them when it’s thin.
- Judge floors by revenue per request or pageview, not by CPM or fill rate alone.
- Test with a traffic split, segment the results and keep human oversight after rollout.
If you want a second look at how floors are set on your inventory, see how Vortex works with publishers.
