Menu

Programmatic: WTF Is Throttling?

Programmatic advertising throttling and SSP bid request filtering

A French publisher recently told me something surprising: when comparing reports from their ad server and one SSP, they found 50% fewer bid requests in the SSP report. Where had the other half gone? One possible explanation is a mechanism that remains poorly understood in programmatic advertising: throttling.

Key takeaway: SSP throttling is the selective reduction or filtering of auction requests to manage infrastructure costs and prioritize economically valuable traffic. It can help an SSP remain efficient—but, depending on how it is configured, it may also prevent valuable impressions from reaching buyers. Publishers should investigate discrepancies and test more granular inventory segmentation rather than assuming every missing request is a lost sale.

What is throttling in programmatic advertising?

SSP throttling refers to an SSP's practice of limiting which bid requests it processes or forwards to demand partners. Rather than treating every available auction opportunity equally, the platform may prioritize traffic according to its expected economic value and operational constraints.

Imagine an ad server counting 100,000 auction opportunities while a connected SSP reports only 50,000 bid requests. Throttling could help explain the difference, but it is not the only possible cause (reporting definitions, time zones, auction eligibility, consent, integration settings and request errors can also affect the comparison).

Why would an SSP deliberately reduce bid requests?

The economic equation is simple: an SSP generally earns revenue when advertising is sold, while its infrastructure must handle a substantial flow of bid requests and bid responses—including auctions that generate no revenue.

With the growth of header bidding, several SSP seats can participate in monetizing the same publisher's inventory. More connections may mean more opportunities to compete, but they can also mean more duplicated processing costs.

Being connected ten times to the same website does not guarantee ten times more sales. It can, however, increase the number of technical interactions the SSP must handle (the actual impact depends on the configuration and how many connections are active).

This is closely related to our earlier analysis of resellers, duplicated bid requests and Supply Path Optimization.

A concrete French example: Le Monde's ads.txt

In an audit of Le Monde—a major French news publisher with a highly developed programmatic setup—I observed the following number of declared seats in its ads.txt file:

SSPSeats listed in the original observation
Xandr16
Magnite14
Equativ13
PubMatic13

(These counts are from the original audit, not a live verification of the current file. An ads.txt declaration is an authorization record: it does not prove that every listed seat is currently active or generating traffic. The original analysis also cited a 9/10 Sellers.Guide score, which should be understood as a point-in-time assessment.)

The lesson is not that this publisher has a poor setup—quite the opposite. It is that even sophisticated publishers can operate in a complex ecosystem of SSP connections and intermediary relationships.

How can SSP throttling work?

A simplified way to think about the process is to group inventory by attributes such as ad slot × website × format × device, then assess the revenue generated relative to the number of requests processed.

SignalWhy it matters
Bid-request volumeRepresents processing load and potential infrastructure costs
Revenue generatedIndicates whether that traffic produces enough economic value
Expected yieldCan inform which traffic is prioritized or restricted
Granularity of the inventory groupDetermines whether valuable and weak opportunities are evaluated together

If a group consistently generates little revenue for a large volume of requests, the SSP may reduce the share of requests processed or sent downstream.

(This is an explanatory model, not a description of every SSP's proprietary algorithm. Different platforms may use additional signals, thresholds and predictive models.)

The catch: not all bid requests in a group are equal

Suppose a broad inventory group contains both low-value traffic and opportunities that are especially attractive to buyers. A filtering model that is insufficiently granular may restrict some of the latter as well.

Several dimensions can affect a bid request's value:

The risk is that high-quality bid requests may never reach the DSP if the SSP's decision is made at a level that does not distinguish them adequately.

This also intersects with our articles on cookie, cookieless and no-consent audience fragmentation and why viewability alone does not capture advertising attention.

What can publishers do to reduce unnecessary throttling?

My recommendation at GreyFox Ads: investigate whether a more granular tagging and inventory strategy can help SSPs identify valuable traffic instead of evaluating it as part of a broad, mixed-quality pool.

  1. Reconcile reporting first. Compare equivalent time periods, auction definitions and eligible inventory between the ad server and SSP.
  2. Identify where filtering happens. Ask SSP partners which reporting dimensions and diagnostics they can provide.
  3. Segment meaningful inventory. Where operationally justified, separate high-quality placements or formats into dedicated ad slots or inventory groups.
  4. Measure the outcome. Monitor request acceptance, fill, revenue, buyer participation and page performance—not just raw bid-request volume.
  5. Coordinate with SSP partners. Avoid adding tags or seats indiscriminately; more connections can increase complexity without creating incremental demand.

With better segmentation, valuable opportunities may become easier for an SSP to recognize, potentially reducing unnecessary filtering. But there is no guarantee that a new ad slot automatically changes throttling rules or increases revenue.

And while you're auditing your supply chain, don't overlook the other side of auction efficiency: how MFA websites can flood the market with excessive bid requests.

Better programmatic economics start with better signals

Throttling is not inherently bad. It is an understandable response to the economics of real-time bidding. The problem begins when cost-saving filters are too blunt to distinguish genuinely valuable impressions from low-yield traffic.

Here in France, we believe the quality Open Web deserves better than unnecessary technical friction. At GreyFox Ads, we may be the smallest ad network in France, but not the least competent. Our Master Deals aim to give Media Traders clean supply, smart targeting, simplicity and transparent costs.

About the author

is the founder of GreyFox Ads, an independent French programmatic curation specialist focused on quality, efficiency and transparency.

Read the original LinkedIn post

Need a smarter programmatic strategy in France?

Work with a French partner focused on quality video and display inventory, smarter targeting and a transparent supply chain.

Contact GreyFox AdsExplore the Master Deal