- Published
- Nov 4, 2026
- Reading time
- 11 min read
DV360 Authorized Sellers: Audit Settings and Supply Hops
Audit DV360 seller settings against bought inventory and declared supply chains. Separate DIRECT authorization, hop limits and CTV exceptions.

In this guide (8 sections)
DV360's documented authorized-seller setting is an authorization policy, not a numerical supply-chain hop limit. To audit a buyer's “direct only” requirement, preserve the effective campaign settings, inspect the seller states reported for purchased inventory, and obtain transaction-path evidence for any separate hop-count requirement. A DIRECT label alone cannot establish all three.
An older public question about DSP control of multi-hop supply describes an SSP asked to meet a buyer's zero-hop requirement. Replies discuss both account authorization and chain-length controls, revealing a practical ambiguity: people may use “direct” or “hop” to mean different things. This guide checks the current DV360 contract; it does not turn those historical replies into claims about another DSP's present capabilities.
The existing ads.txt, sellers.json and SupplyChain investigation guide owns the file and identity checks. Here, the distinct task is accepting a buyer's configured DV360 policy against evidence of what was bought.
Translate “direct only” into an auditable requirement#
Ask the buyer to write down what must be excluded. The answer determines the evidence and who can supply it.
| Intended requirement | Evidence to accept | Shortcut that does not answer it |
|---|---|---|
| Buy only the platform's authorized DIRECT category | Effective authorized-seller configuration and the relevant reported seller-state breakdown | A partner saying it has a direct commercial relationship |
| Allow no resale step under a defined chain rule | Complete declared chain, agreed node interpretation and seller-identity validation | An authorization label without the transaction path |
| Avoid particular intermediaries | Defined entity exclusions and path evidence showing their application | Counting only the visible exchange names |
| Avoid duplicated access to the same opportunity | Agreed opportunity identifiers and evidence of how competing routes are identified | Calling every repeated domain a duplicate impression |
| Improve traffic quality or business outcomes | Separately scoped delivery, IVT, viewability or outcome evidence | Assuming fewer intermediaries guarantee better results |
Keep authorization, path length and duplicated access as separate clauses. An SSP may need one implementation to validate sellers and another to manage route duplication. Do not accept “zero hops” as a complete specification until the parties agree whether the originating selling system is counted, which participants belong in the path and how incomplete records are treated.
For a strict path requirement, write an acceptance rule in ordinary language: for example, a complete declared chain with one identified selling-system/account node, validated against the applicable owner relationship. This is a hypothetical policy definition, not a claim that a particular DV360 switch implements it.
Turn direct only into a checkable rule
Authorized-seller settings do not set a numerical supply-path limit. Discuss what campaign and transaction-path evidence is available.
Discuss measurement optionsA short inquiry to scope relevant options; no campaign access is needed.
Understand what the seller setting expresses#
IAB Tech Lab's ads.txt specification defines DIRECT through the publisher's control of the named account in the specified system. It is an account relationship. The related SupplyChain specification describes ordered participants in the direct flow of payment. Those are different records with different questions to answer.
The DV360 authorized-seller enum reference maps its direct-only option to a DIRECT declaration. It does not define that option as a maximum number of SupplyChain nodes. Do not convert that relationship filter into a numeric path-length guarantee.
For human review, Google's Line Item QA format exposes an authorized-seller options field with three named choices: direct sellers, direct sellers and resellers, or authorized and non-participating publishers. QA files are read-only; they are useful audit artifacts and cannot be uploaded as configuration changes. Record the version actually exported rather than assuming every account uses the version linked here.
When a file identity is disputed, hand it back to the existing supply-path guide. Resolving the account controller, a monetization manager or an inventory-sharing relationship requires evidence beyond choosing a stricter buyer label.
Capture effective settings rather than relying on defaults#
Google's Authorized Digital Sellers guidance describes settings at multiple buying levels and inheritance for newly created entities. A current line item therefore needs its own readback. A partner-level policy screenshot cannot establish the settings of every existing child.
There is also a reason to avoid copying a default from a reference page: the enum document's introductory default wording differs from its UNSPECIFIED row. The shared assigned-targeting detail describes the absence of that assigned option as direct sellers and resellers. These descriptions are not evidence of your account's actual setting.
Use this recommended read-only collection sequence:
- Identify the partner, advertiser, campaign, insertion order and exact line items covered by the buyer's policy.
- Save the line-item quality setting and applicable parent context with an observation timestamp. Record whether the evidence is a UI readback, exported QA field or API response.
- Preserve the original export and its schema/version. Keep identifiers as identifiers; a renamed line item should not break the audit join.
- If using API data, record the resource, targeting type, assigned value and any relevant inheritance context. An absent field is a question to resolve, not a self-authored fallback rule.
- Reconcile disagreements between surfaces with the platform operator or support before changing delivery policy.
The goal is a reproducible record: this scope had this effective policy at this time. If a change is proposed later, preserve its approval, exact affected entities and readback separately. Downloading a file is not proof that a policy was changed or enforced.
Compare the policy with bought seller states#
Google documents an Authorized Seller state reporting dimension for the participation or authorization status of inventory purchased. Use the supported breakdown to compare the policy with the bought scope. Keep ineligible or non-participating populations visible rather than merging them into a reassuring combined “direct” total.
Save a completed interval with the reporting timezone, line-item identifiers, inventory scope, exact metric and export timestamp. Check whether the period crosses a policy change. Split the periods where possible, or state that a mixed interval cannot test the new policy cleanly.
A report row is a classification under the platform's methodology. It is not a list of the chain's participants. If the buyer requires a specific maximum chain length, the report is a useful configuration check but does not finish the path audit.
Hypothetical example: authorization share is not hop compliance
Assume a hypothetical completed seven-day report in Asia/Kolkata contains 1,000,000 impressions for the exact line items under review:
| Reported category | Impressions | Share of this report |
|---|---|---|
| Authorized direct sellers | 700,000 | 70% |
| Authorized resellers | 150,000 | 15% |
| Non-participating | 100,000 | 10% |
| Ineligible | 50,000 | 5% |
These are invented teaching values, not an observed campaign or an expected DV360 distribution. The direct category accounts for 700,000 divided by 1,000,000 impressions, or 70% of this report. It does not prove that 70% of requests met a one-node chain rule.
If the buyer expected an applicable direct-only policy throughout the entire scope, the other categories create specific questions. Were all line items covered? Was the policy changed during the interval? Is the environment within the setting's enforcement scope? Is there a documented classification or reporting limitation? Preserve the original categories while those questions are answered.
Do not call the remaining 30% unauthorized traffic. Authorized resale, non-participation and ineligibility have different meanings. Nor does the table establish billed waste, duplicate exposures or an entitlement to a refund.
Verify a chain rule using chain evidence#
Request authorized, redacted transaction samples for the actual deal or inventory being reviewed. Preserve the request identifier available at that boundary, timestamp, property/app, selling system, seller account, chain version, completeness flag and ordered nodes. Ask the partners which identifiers can be correlated across their systems.
In the SupplyChain specification, an incomplete chain begins with the first known node rather than necessarily the original inventory owner. A one-node record marked incomplete is therefore not equivalent to a complete one-node record. The completeness flag is itself a declaration; identity and relationship validation remain necessary.
Suppose a separate, hypothetical sample of 100 bid requests contains:
- 60 complete, one-node chains;
- 25 complete chains with two or more nodes;
- 15 incomplete chains, each displaying one node.
Seventy-five records display one node, but only 60 satisfy the sample's declared complete-one-node condition. Twenty-five fail that particular rule, while 15 remain unresolved because an upstream path may be missing. The completeness rate is 85%, and the declared complete-one-node share is 60% of the sample.
This request sample is a different population from the million reported impressions above. Unless sampling and request-to-delivery linkage support it, do not extrapolate the 60% to bought impressions or multiply it by the report total. Do not call the 15 incomplete records fraud. Document the missing evidence and its owner.
The distinction is useful even when both counts look good: a matching authorization file and a short declared chain still do not establish valid human delivery, correct geography, viewability or incremental value.
Handle CTV and no-spend questions as separate branches#
Google's authorization guidance explicitly describes a CTV exception: devices without app-ads.txt support are treated as non-participating and may still be bid on regardless of the advertiser's targeting criterion. It also identifies inventory outside the authorization mechanism's scope. Check the actual environment before treating a category mismatch as a failed setting.
For a deal that loses delivery after a policy change, preserve both the control change and the applicable eligibility evidence. The DV360 deal troubleshooter includes Authorized Seller state among its targeting diagnostics; that targeting breakdown is not available for the Current day date selection. These diagnostics concern opportunities not bought, rather than a ledger of delivered impressions.
Do not subtract a targeting-filter count from a delivery total to construct an invented waterfall. Record what population each screen describes, which reasons can overlap and whether the platform permits the comparison. If the investigation reaches a win with no matching impression, use the ad-server impression reconciliation to trace the subsequent delivery and counting boundary.
A delivery decline does not automatically justify broadening to non-participating inventory. Decide whether that change fits the buyer's quality policy, obtain the required approval and verify its actual scope. A forecast is planning evidence; a controlled policy readback and subsequent comparable records are enforcement evidence.
Close the audit with a decision record#
Keep four answers on one page:
- Requirement: the exact authorization rule and any separately defined path-length or duplication condition.
- Configured scope: current line-item evidence, parent context, versions and effective period.
- Observed result: bought seller-state counts, supported limitations and separately scoped path samples.
- Disposition: accepted, accepted with a named limitation, or unresolved pending specific evidence.
Assign each gap to the team that can close it. The buyer owns the policy and account settings; the relevant platform or seller must explain its provided records. Do not ask the publisher to repair a buyer override by changing a truthful DIRECT or RESELLER declaration.
The anonymized dairy video case study illustrates a related boundary: impression-quality filtering has its own format and population. It does not establish any seller-path result. Keep that analysis alongside the authorization audit rather than treating one as proof of the other.
For a supported measurement implementation, bring the scoped review to AdProof. Export the relevant AdProof events or aggregates using explicit campaign, date and timezone selections, and compare them with DV360 reports and authorized path evidence obtained separately. Confirm the available observation scope before promising a combined analysis; an event export does not automatically contain a DSP's full supply chain.
Frequently asked questions#
Does Authorized Direct Sellers only enforce a zero-hop rule?
The documented DV360 option filters the authorized account relationship associated with DIRECT. A numerical chain rule requires its own definition and transaction-path evidence. Do not describe the setting as a generic zero-hop guarantee.
Why does an unset API value not settle the effective policy?
An enum placeholder, an absent assigned option and a current line-item configuration are different evidence. Reference descriptions also contain conflicting default wording. Preserve the actual resource and readback, then resolve ambiguity with the platform rather than inventing a default.
Can a one-node SupplyChain record still be incomplete?
Yes. The completeness flag identifies whether the declared chain reaches back through all required participants. If it is incomplete, one visible node may only be the first known seller. Keep that record unresolved under a strict complete-chain requirement.
Does a clean seller-state report prove good-quality impressions?
No. Authorization, declared transaction path and measured delivery quality answer different questions. A clean classification does not independently prove human viewers, viewability, correct targeting, lower costs or better sales outcomes.
Turn direct only into a checkable rule
Authorized-seller settings do not set a numerical supply-path limit. Discuss what campaign and transaction-path evidence is available.
Discuss measurement optionsA short inquiry to scope relevant options; no campaign access is needed.
