Floor prices have long been an important part of publisher yield strategies. But finding the right floor is becoming increasingly difficult as auction conditions, demand and user behaviour constantly change. This makes floor price optimisation a natural machine learning use case. Instead of relying on static rules or manually adjusting floors based on historical averages, machine learning can analyse far more signals and continuously adapt to changing auction conditions.
Why Granularity Matters
Effective floor price optimisation requires more than looking at CPM alone. Demand can behave differently depending on placement, device, format, market, and other auction characteristics. Optimising floors therefore requires enough granular data to understand these differences and how they affect the wider auction. Relevant Yield’s new Automated Floor Price Optimisation uses publisher-specific machine learning models to analyse thousands of auction signals across more than 60 metrics and dimensions. The models update several times per day and can also incorporate publishers’ first-party data.
This lets optimisation reflect the characteristics of each publisher’s inventory, rather than relying on the same assumptions across publishers and environments. The highest CPM does not necessarily mean the best overall outcome. Floor prices influence which bidders participate in an auction and which bids remain eligible to compete. Setting floors too aggressively can therefore affect other parts of the auction chain. Automated Floor Price Optimisation considers the wider auction environment when setting floors. The aim is to find floor prices that support revenue while maintaining healthy competition rather than optimising an individual metric in isolation.
The solution covers Prebid bidders, ad servers, Amazon TAM, and OpenAds, and works across client-side and server-side setups on desktop, mobile web, and native app environments.
Automation Without Giving Up Control
Automating optimisation does not mean that publishers should lose control over their floor price strategy. Within Relevant Yield, publishers can define minimum floor prices, choose which placements use automated optimisation and manage ad server pricing rules. Real-time reporting provides visibility into how the optimisation performs, allowing publishers to track impact rather than rely on a black-box system.
This combination of automation and transparency matters as machine learning becomes a larger part of day-to-day ad revenue optimisation. The technology can process more data and react faster to changing conditions, while the publisher remains in control of how and where it is applied.
What Does It Mean for Publishers?
The impact of floor price optimisation varies by publisher, depending on factors such as inventory, demand pressure and the existing floor price strategy. Early results have shown a few per cent revenue uplift for most sellers, with some seeing considerably higher gains. As the optimisation engine learns from and adapts to each publisher’s data over time, it may take a few months for the full impact to be seen at scale.
The goal is not simply to push floor prices higher. It is to use data to determine when a higher or lower floor is likely to produce a better overall result and automate that decision-making at a level of granularity that would be difficult to manage manually.
Automated Floor Price Optimisation is included with Relevant Yield’s HB Analytics and HB Manager at no additional cost. It gives publishers a more dynamic way to manage floor prices, continuously adapting to changing auction conditions while maintaining visibility and control over how the optimisation is applied. Read more about HB Manager here or contact us to learn more.