What Counts as a Competitor? Eligibility, Ranking, and the Supply Path
To explain a CPM, we need to inspect the boundary of competition: who was allowed into the auction, who made each decision, and which records link the opportunity to the result. This chapter treats an ad auction as a system of interfaces rather than a single ranking formula, so readers can tell the difference between a weak market, a filtered market, and a result whose supply path is simply not visible.
Competition begins before ranking
A bid that appears in a buyer's log is not necessarily a bid that competed. The Media Rating Council's definition of a qualified bid makes this boundary explicit. Before an auctioneer compares demand, a bid may have to be technically valid, come from an eligible demand partner, satisfy a deal or publisher threshold, meet a price floor, pass brand-safety and competitive-separation requirements, and survive checks on the buyer, seat, advertiser, and seller's allow or deny lists. In OpenRTB, a malformed bid is not processed, and the auctioneer may stop if no demand partners are eligible.
That sequence changes how to interpret a thin auction. Suppose a report shows only two meaningful bidders and a low clearing price. One explanation is that few advertisers value the impression. Another is that several bids were filtered because their supply source, creative, deal, brand-safety condition, or price did not qualify. The observed bidder count is therefore an outcome of admission rules as well as demand. Without the qualification results, “low competition” is a conclusion about the market that may actually be a conclusion about the system's gatekeeping.
The ownership of that gate matters. The buyer controls the submitted bid and campaign conditions. The auctioneer controls technical processing and often the eligibility logic. The seller or publisher may set inventory, deal, brand-safety, or competitive-separation requirements. A measurement provider may later record the result without controlling either admission or ranking. A CPM report that combines these layers can make the final number look like a direct expression of buyer willingness to pay even when the set of available choices was narrowed earlier.
Qualification is not the same as winning. It is the condition for entering the contest. Once bids are admitted, an auctioneer may use more than price to determine the winner. The MRC standards identify demand-source priority, seller-defined business rules, ad quality, context, and relevance as possible factors. They require the variables and their relative weighting to be made known and reported where applied, along with the method used to construct variables such as relevance and quality and the actual values applied.
This is why the phrase “the highest bid won” can be too strong. It is an accurate shorthand for some simple auctions, but it is not a safe description of every closed-loop system. A bid with a lower nominal CPM could receive a higher effective ranking value after a quality or relevance adjustment. A higher visible bid could lose because a business rule gives priority to another demand source. A direct deal or managed-service buy could supersede the auction altogether. The fact that one ad was served does not, by itself, reveal which of these paths occurred.
These factors have different meanings and should not be collapsed into a single idea of “quality.” Relevance may describe how well the creative matches the opportunity. Quality may be a separate score. Priority may be a seller rule. A direct deal may be a contractual arrangement that bypasses ordinary competition. The MRC asks auctioneers to disclose the methods and actual applied variables precisely because the labels alone do not tell a participant how much each factor mattered.
The system-level consequence is that the competitor set and the ranking order are both part of cost. If a platform removes a class of supply before bidding, changes the eligibility threshold, or changes the weight applied to relevance, the clearing-price distribution can move even if advertisers leave their nominal CPMs untouched. That movement does not prove that the auctioneer made an error; it does prove that an explanation must include the decision layer before the payment layer.
Records have to follow the opportunity
The next boundary is the path connecting an impression opportunity to a bid and an outcome. In a closed-loop auction, the auctioneer may host most or all of the functions inside one platform. Participants can still need direct disclosure of the rules, demand-source treatment, pricing, and winner determination because an internal system is not automatically an observable system.
Open programmatic auctions use a different interface. OpenRTB is an IAB Tech Lab protocol for automated trading between buying platforms, often called demand-side platforms or DSPs, and selling platforms, often called supply-side platforms or SSPs. It carries bid requests, bids, and auction outcomes. That shared protocol makes certain facts easier to connect, but it does not guarantee that every optional field is present or that the full price logic is exposed. Private auctions can use OpenRTB while relying on additional, pre-negotiated rules.
The MRC standards emphasize a small set of links because a price claim needs a joinable trail. A transaction ID, or tid, can link bids from different bidders to a specific impression. A global placement ID, or gpid, can provide a consistent identity for the ad placement. A complete SupplyChain object and SupplyChainNode object, including the RequestId, can describe the path through which the opportunity traveled. Loss codes can tell the buying side why a bid did not win. Reporting against the transaction ID where applicable and the RequestID helps a participant connect the auction record to buyer and seller systems.
These identifiers do not make the auction transparent by magic. They make questions answerable. If a buyer wants to know why the same nominal CPM produced different outcomes, a transaction-level identifier can help them compare the actual opportunity, placement, competing bids, loss reason, and reported payment. If a supply-chain identifier shows a different path, the buyer can ask whether inventory source, reseller relationships, or publisher-level rules differed. If no identifier joins the records, the buyer may be left comparing aggregates that refer to different populations.
What a supply path can and cannot prove
The supply-chain record is useful because inventory can pass through more than one organization or technical node before an impression is auctioned. The MRC calls for validation between seller IDs declared in ads.txt files and the corresponding IDs in SSP sellers.json files. It also calls for the chain to end at the owner of the inventory, for buyers to receive reporting on the average number of nodes or “hops,” and for misdeclared or undeclared chains to be visible.
Those practices answer an important identity question: who was authorized to sell this inventory, and how many declared intermediaries stood between the buyer and the owner? They do not, on their own, prove how each intermediary changed the final CPM. Identity, path length, auction rules, and fees are separate dimensions. A shorter path may be easier to reconcile, but a supply-chain object without pricing and outcome records cannot establish a lower cost. Conversely, a longer path may explain why a report contains multiple records, but it does not establish that every hop added a particular fee.
A unified seller platform makes this boundary especially important. One organization may operate an auction or serve ads while also aggregating supply and requesting bids on behalf of publishers. The MRC says such a platform should report general demand-source rules, pricing, winner determination, and the factors affecting price to buyer participants. Yet publisher configurations, deal terms, auction logic, restricted DSP data, and third-party SSP involvement can limit full-path transparency. The platform may not be able to map every transaction one-to-one through all OpenRTB parameters or derive every price component for every buyer.
The appropriate response to that limit is not to invent a missing explanation. The standard calls for processes to seek publisher approval for the information that can be shared and, at minimum, disclosure of general factors affecting the auction by publisher. A buyer can then distinguish three statements: “the path is identified,” “the path has a reported limitation,” and “the path proves the cause of this price.” Only the third is a causal claim, and it requires more evidence than the first two.
Why a loss code can be more valuable than a headline CPM
A winning price tells the buyer what happened in a selected auction. A loss code can indicate why a particular bid did not become that outcome. The MRC requires appropriate loss codes to pass from seller to buyer for all transactions and to be reported by the buy-side platform. That turns a missing impression into a potentially diagnosable event. The buyer can ask whether the loss arose from price, eligibility, a business rule, a creative condition, or another disclosed category rather than treating every loss as evidence that the bid was simply too low.
The distinction is useful even when the buyer cares only about average cost. Imagine two campaigns with the same nominal CPM. Campaign A loses most of its opportunities because its creative fails a qualification condition, then wins a small set of relatively easy impressions. Campaign B is eligible broadly but loses close contests to higher effective values. Their delivered CPMs might look similar, but their strategic problems differ. Campaign A needs to examine eligibility and creative policy; Campaign B may need to understand ranking, competition, or bid strategy. An average cannot choose between those explanations.
This is also why a machine-readable record is not merely a convenience for analysts. It preserves the relationships among opportunity, placement, bid, supply path, qualification, outcome, and report. If those relationships are retained, a participant can aggregate later by placement, publisher, buyer, loss reason, or time window. If only a campaign-level CPM survives, the aggregation has already discarded the distinctions needed to diagnose it.
The MRC does not claim that every field will be available in every transaction. It recognizes optional OpenRTB fields that sellers may not provide, contractual limits on what can be shared, and cases where third-party supply relationships prevent full mapping. A missing field is therefore an evidence boundary. It may justify the statement “the cause cannot be determined from the available record,” but not “the platform must have charged an undisclosed fee” or “the supply path caused the increase.” Absence of a link is not proof of a particular hidden mechanism.
The result is a more precise interpretation of competition. An auction has a potential competitor set—the buyers or demand sources that could in principle participate—and an observed qualified set—the bids that passed the system's rules for this opportunity. The price is generated from the latter, under the auctioneer's ranking and payment logic. Supply-chain and identifier records tell us which opportunity and organizations the event belongs to; qualification and loss records tell us what entered or failed; winner and payment records tell us what cleared. None of these layers can substitute for the others.
That leaves the central operational question in a better form. When a CPM changes, do not ask only whether buyers bid more or sellers charged more. Ask whether the opportunity was the same, whether the qualified competitor set was the same, whether the ranking inputs and priorities were the same, and whether the record still connects the event to the reported cost. Once those boundaries are visible, the next task is to use the evidence to distinguish a changed market from a changed rule, model, or measurement process.