In programmatic advertising, performance conversations tend to begin with CPMs, fill rates, win rates, and revenue. Yet one of the most important decisions happens before any auction takes place: which bid requests should be sent in the first place?
That is the role of QPS management.
Queries per second, or QPS, represents the volume of requests flowing between programmatic platforms. At scale, exchanges and supply platforms process enormous volumes of opportunities across CTV, mobile app, and web environments. Sending more traffic, however, does not automatically create more value.
Effective programmatic operations increasingly depend on identifying the traffic most likely to generate a meaningful outcome, and eliminating traffic that consumes infrastructure without creating sufficient commercial value.
More QPS Does Not Mean More Revenue
Historically, scale has been one of programmatic advertising’s defining advantages. More inventory created more opportunities to reach audiences, attract demand, and generate revenue.
But unrestricted scale introduces inefficiency.
When every available impression is routed to every eligible demand connection, platforms can generate significant volumes of requests with limited probability of receiving a bid. Infrastructure costs increase, demand partners receive unnecessary traffic, and potentially valuable supply becomes diluted within a much larger stream of low-performing requests.
The objective should therefore not be to maximise QPS. It should be to maximise the value generated by each unit of QPS.
This requires understanding performance at a granular level, including geography, device type, operating system, inventory source, format, bid floor, demand connection, historical bid behaviour, and profitability.
QPS Management Is Ultimately Traffic Intelligence
Sophisticated QPS management goes beyond applying static limits.
Programmatic operations teams need to continuously evaluate which supply and demand combinations create value. A connection performing strongly for US CTV inventory, for example, may have completely different economics when receiving mobile app traffic from another geography.
The operational question becomes increasingly specific: Which impression should reach which buyer, at what moment, and under what conditions?
At MarkApp, this philosophy is reflected in the optimisation infrastructure supporting our programmatic ecosystem. Automated connection filtering can identify low-performing or non-profitable supply-demand combinations, while geo-prioritisation and floor-price optimisation help concentrate resources where stronger commercial outcomes are achievable.
The result is a more disciplined supply path rather than simply a larger one.
Less Traffic, Better Economics
The impact of QPS optimisation can sometimes appear counterintuitive. Reducing traffic can actually improve performance.
As Giannis Syropoulos, Chief Operations Officer at MarkApp, explains:
“We reduced our outbound QPS by 35% and our win rate went up. Less is more when the traffic is right.”
That principle captures an important shift in programmatic operations. Removing inefficient requests gives demand partners a cleaner, more relevant stream of opportunities. It can improve bid efficiency, reduce unnecessary infrastructure consumption, and create healthier commercial relationships between supply and demand platforms.
For an exchange operating at scale, these efficiencies compound quickly.
The Metric Behind the Metrics
CPMs, fill rates, win rates, and revenue remain essential measures of programmatic performance. But they are downstream outcomes.
QPS management influences what enters the auction environment before those metrics are generated.
As programmatic infrastructure becomes more sophisticated, particularly across high-volume CTV, mobile app, and web environments, operational excellence will increasingly depend on intelligent traffic selection rather than unlimited traffic distribution.
The next generation of programmatic optimisation will not simply ask how many opportunities a platform can process.
It will ask how many of those opportunities were worth processing in the first place.
