The search for high-quality impressions
For performance marketers operating in digital advertising, the challenge is not inventory availability, it's efficiently reaching impressions that align with measurable outcomes.
Our answer to this is the tvScientific Supply Side Optimization Agent, a highly specialized tool designed to inject efficiency and precision into the bidding process.
What is tvScientific’s Supply Side Optimization Agent?
The Supply Side Optimization Agent is not just a filter; it is an intelligent preselection engine that operates directly within our supply side partner’s infrastructure. Its primary function is to inspect 100% of available supply and prefilter it before it ever arrives at our bidding algorithms.
Jonathan Watkins, Sr. Director of Product Management, frames it more simply:
"You could think about this thing as being a private shopper for you that's doing some initial screening and throwing away things that it knows you are not interested in."
By filtering for high-intent impressions before they reach the bidder, we avoid processing millions of bid requests that are unlikely to align with advertiser objectives.
Curating quality inventory & driving major customer ROI
The strategic value of the Agent lies in its ability to dramatically improve the quality signal reaching our core bidding system. Rather than letting our bidder wade through endless opportunities, it drills the supply down to the high-quality inventory, based on the parameters we provide.
This approach aligns with logistical strategies seen in other mature industries (like big box retail and shipping) that optimize distribution by bringing curated inventory closer to distribution points.
We apply these same principles to programmatic inventory, seeking to bring high-quality, curated supply closer to our bidder so that our customers’ ads perform better and their ad spend is more efficient. By performing this prefilter and preselection process, our technology drives up the average customer performance for the inventory we purchase, ensuring every dollar of ad spend is focused on the highest-quality opportunities.
Using prescreened supply, tvScientific customers have seen a 50-70% improvement in cost efficiency compared to performance against unscreened inventory. This improvement reflects a fundamental shift in how efficiently the bidder can evaluate and select inventory.
"If you wanted a certain kind of apple and I brought you directly to the row of trees where that kind of apple is grown, you’re going to have a much higher success rate selecting that piece of fruit than if you had to search the whole orchard for it. We're prefiltering and preselecting inventory in order to drive up the average customer performance for the inventory that we are buying." - Jonathan Watkins
Engineered for precision
The engineering behind the Supply Side Optimization Agent reflects a core commitment to technical rigor, particularly because of the challenging deployment environment. The Agent runs inside our SSP partner’s infrastructure, meaning we do not control the deployment ourselves and have limited telemetry and control over the underlying hardware.
This uncertainty forces us to design software that is extremely conservative around resource usage. Michael Bilow, VP of Data Science & Machine Learning, explains:
“It’s a lot more difficult to develop. The software has to be highly adaptive because the same workload might run on a large server one moment and a much smaller one the next. That uncertainty forces us to be extremely conservative — especially around memory usage, threading, and resource assumptions.”
This technical rigor ensures the software is robust and designed to tolerate constant change in its execution environment.
Intelligence with minimal data
Beyond infrastructure, the Agent also operates under strict data constraints. It operates early in the funnel, meaning the data it receives is generally a subset of what reaches our core bidder, often lacking later enrichments like geographic or household identifiers. This requires training specialized machine learning models that can operate effectively within this limited infrastructure. Our models demonstrate high efficiency despite this reduced data input, proving that our AI is not reliant on costly, late-stage data aggregation but on highly efficient, early-stage analysis.
Governance and speed: A clean room alternative
To address privacy and governance concerns, the industry has leaned on clean rooms — often sacrificing real-time decisioning in the process. Securely hosting our models directly within the auction path changes what’s possible for bidding and optimization. This real-time hosting capability delivers many of the core benefits people look for in a clean room: no raw data moving around, decisions made inside a trusted environment, and higher-fidelity use of signals in real time. The result is the accuracy and quality benefits of deeper data collaboration without the cost, operational overhead, and lag of traditional clean rooms.
Built for cost efficiency
The Supply Side Optimization Agent’s strategic placement creates a significant economic advantage by minimizing the cost of data transfer, which is one of the biggest expenses in running a digital ad auction. Transferring the high volume of bid requests happening every second is costly, especially when transferring data between disparate cloud providers, such as Google (where many SSPs run) and AWS (where we primarily operate).
By prescreening traffic on the supply partner’s side, we eliminate unnecessary data transfer while preserving full visibility into relevant inbound traffic, allowing us to go directly to the inventory we’re most likely to buy.
This system creates efficiency on both sides of the marketplace:
"Since we have a higher propensity to bid on the traffic we receive, it’s more efficient for supply-side partners to route that traffic to us. Everybody wins: we get a much broader view of the available inventory, and our supply partner gets a higher auction fill rate on the requests they send us, which reduces wasted auction costs and overhead.” - Michael Bilow
This pre-curated supply ultimately drives up the key business KPIs our supply partners care about, such as bid rate and demand pool size.
Looking ahead
The Supply Side Optimization Agent represents a shift in how CTV inventory is valued and purchased. By pushing intelligence and optimization capabilities directly into the supply path, it proves that CTV is moving deeper into outcome-based performance, operating more like search and social performance media.
Interested in learning more about our Supply Side Optimization Agent? Download the one-pager or contact us.
