This article explores how to measure the impact of floor pricing correctly, why common evaluation methods fail, and how revenue teams can regain strategic control over one of their most influential levers.
Floor pricing is now standard practice in programmatic management. Most publishers I speak with have it in place.
What deserves more scrutiny is how well optimized those floors are inside today’s auctions.
DSPs (demand side platforms) now reprice impressions in milliseconds using live signals such as competition, pacing, and predicted outcomes – a capability enabled by real-time bidding protocols that run per-impression auctions in the time it takes a page to load.¹ Yet many publishers still rely on static floor tables built on historical averages.
That mismatch creates a widening gap between what inventory is worth in the moment and what it actually clears for. The result rarely shows up as an obvious failure. Instead, it appears as thinner auctions, fewer bidders, and revenue that never quite materializes in reporting.
I want to focus on that gap. Not to debate whether floor pricing belongs in the stack, but to examine how best to measure and optimize this.
And having analyzed the impact of hundreds of billions of auctions across many publishers, I’ve seen how correctly optimized floor strategies can increase RPMs by 10% to 18%, and CPMs by 7% to 11%, while fill rates remain stable.
“Having floors” is not the same as having a strategy
Many organizations treat floor pricing as a safeguard. Set a minimum. Protect value. Move on.
That logic makes sense in a fixed-price world. But programmatic auctions are not fixed. Demand shifts by time of day, geography, device, format, campaign pacing, and competitive pressure. Buyers respond to those signals continuously. When floors remain static while demand evolves, pricing stops being strategic. It becomes inherited.
In performance reviews, I usually see two patterns.
In the first, floors remain high after a strong period. Demand later softens. Fewer buyers participate. CPM (cost per thousand impressions) still looks healthy, but fewer impressions clear. Total revenue begins to slip.
In the second, floors remain conservative to protect volume. Fill stays strong. CPM stagnates. Buyers face little pressure to compete. Growth never arrives.
Neither looks dramatic in isolation. Both represent missed yield.
Floors influence who enters an auction, how aggressively they bid, and whether competition exists at all. When those levers are left unchanged while buyer behavior shifts in real time, pricing stops reflecting reality.
A strategy does not mean constant change for its own sake. It means intentional alignment between pricing and demand.
Marketing Technology News: MarTech Interview with Omri Shtayer, Vice President of Data Products and DaaS at Similarweb
Before-and-after analysis rarely tells the truth
Most teams evaluate floors by changing a rule and comparing performance before and after. This approach feels logical. It is also deeply unreliable.
Auction outcomes fluctuate for many reasons unrelated to floor pricing: campaign flighting, pacing, traffic mix, creative rotation, and broader market conditions. These variables are inherent to the automated, impression-by-impression structure of real-time bidding.² When I see teams compare last week to this week, they are usually measuring noise rather than pricing.
To understand pricing’s true effect, floored and control groups must run in parallel against the same demand in real time. Without that structure, it is impossible to isolate whether revenue changed because of pricing or because the market moved underneath it.
This is why floor strategies often feel difficult to evaluate. The problem is not that floors are unpredictable. It is that we are still using measurement approaches that were never designed for auction-level economics.
CPM is not a success metric for pricing
CPM remains the most common way teams judge floor performance. It is also the most misleading.
CPM tells you what an impression sold for. It does not tell a publisher what the business earned.
The only metric that captures the true impact of floor pricing is holistic RPM, which is calculated as total programmatic revenue per thousand ad opportunities. That includes header bidding, Amazon, Google AdX and Open Bidding, divided by total ad requests.³ That single number captures price, volume, and participation in one outcome. I then use CPM, fill rate, and bid density to explain why RPM moved.
- If CPM rises while RPM falls, volume suffered.
- If RPM holds while bid density collapses, future risk is building.
- If all three improve together, pricing encouraged healthier competition.
This approach does not replace CPM. It simply gives CPM the context it needs.
In every other pricing-driven business, leaders judge success by total yield, not by list price. Programmatic deserves the same discipline.
Floors shape demand, not just outcomes
One reason CPM misleads is that floors do more than influence clearing price. They shape who is allowed to compete in the first place.
SSPs and DSPs use floor values to route budgets, select line items, and manage pacing. In other words, floors do not simply protect revenue after bids arrive. They influence whether bids arrive at all.
When floors are aligned with demand, they can unlock competition that would never engage with underpriced inventory. When they are misaligned or invisible, buyers gradually disengage and redirect spend elsewhere.
This is why floor implementation matters just as much as the price itself.
When pricing is enforced only at the ad-server layer rather than signaled into the auction, buyers cannot respond intelligently. They do not know why they lost, what price was required, or whether the inventory was even available. Without those signals, algorithms cannot adapt. They simply seek environments that provide clearer feedback.
What changes when pricing reflects real demand
When floors are aligned with real-time auction behavior rather than static averages, the pattern is consistent across publishers and verticals.
- Total RPM rises.
- CPM improves modestly.
- Fill remains stable.
- Bidder participation behaves differently by placement, device, and market.
These gains do not come from forcing higher prices. They come from finding the actual clearing point of each auction and letting buyers compete at that level. Static floors miss that point far more often than most teams realize.
A practical framework to test what works
Floor optimization does not require a long roadmap or a platform overhaul, but it does require discipline. I recommend an initial cycle focused on learning rather than perfection.
At Mile, we start by running floored and controlled traffic side by side against the same demand. When we run these tests, we’ll usually have 10% to 20% of the traffic going into the control group, with the rest going into the floored group.
We then study behavior, not just revenue. Specifically, we watch how participation changes. Where competition strengthens. Where it weakens. And which placements respond differently.
Finally, we apply what we learn, by keeping what improves total yield and optimizing what does not. Over time, each iteration continues to improve the revenue outcomes.
Your reporting structure determines what you see
One reason many teams struggle to reach that level of clarity is that their reporting tools were never designed to show it.
Most dashboards collapse the very dimensions that matter most: bidder, placement, device, geography, and timing. Site-level averages hide where value is created or destroyed.
The real economics live at the intersection of those variables. Without that visibility, teams are left managing revenue with blurred vision, relying on summary metrics to explain behaviors that only appear at auction level.
Floor pricing is a strategy, not a setting
I do not think floor pricing fails because organizations ignore it. I think it struggles when we stop questioning it.
Floors operate where revenue, demand, and buyer behavior meet. That position gives them real influence, but it also makes their effects easy to misread when we rely on surface metrics alone.
When revenue teams look only at headline numbers, floors appear stable. When they look at total business outcomes, the picture often changes. What once felt protective begins to look restrictive. What once felt safe starts to deserve review.
I am not arguing for a single model or a universal playbook. I am arguing for ownership.
The sell side has already gone algorithmic. Pricing has to follow.
When organizations measure what truly matters, they stop reacting to the market and start shaping it. Floor pricing then returns to its proper role – not as background infrastructure, but as a strategic lever that deserves regular attention.
