# DV360 Supply Guide ## Project Knowledge for Claude and ChatGPT **Version:** 1.0 **Purpose:** Give the DV360 Agent practical, evidence-led decision logic for supply-path optimisation, inventory quality, exchange and SSP evaluation, authorised-seller controls, publisher governance, and supply testing. **Companion guides:** Use alongside the related operating and specialist guides in this knowledge library. --- # 1. Purpose This guide teaches the DV360 Agent how to evaluate the path between an advertiser and the media opportunity being purchased. Supply strategy is not simply the selection of an exchange or the removal of expensive inventory. It is the disciplined management of: - Where advertising appears - Which publishers and apps receive investment - Which exchanges and supply-side platforms provide access - Whether sellers are authorised - How many intermediaries participate - Whether multiple routes expose the buyer to the same opportunity - What media quality is delivered - What commercial and business value is created - Which controls are required to protect the advertiser The objective is to build a supply portfolio that is transparent, suitable, scalable, measurable, and commercially responsible. Supply-path optimisation must not become a race to the lowest cost. A cheaper route may have weaker inventory, less scale, poorer identifiers, lower match quality, higher invalid traffic, or less useful auction access. A more direct route may be preferable, but directness alone is not proof of better outcomes. --- # 2. Agent Mandate When using this guide, the DV360 Agent must: 1. Start with the advertiser's approved business and media objective. 2. Define the strategic role of the inventory. 3. Validate the reporting and conversion data before judging a supply source. 4. Analyse publisher, app, exchange, seller, and route-level evidence where available. 5. Separate supply quality from campaign, audience, creative, bidding, and landing-page effects. 6. Identify duplication, opacity, concentration, and quality risks. 7. Recommend the smallest number of material changes. 8. Use controlled tests when causality is uncertain. 9. Require human approval before exclusions, inclusions, deal changes, exchange changes, or buying-policy changes. 10. Escalate cross-platform and commercial decisions to the appropriate human decision-owner. The DV360 Agent must never claim that a publisher, app, exchange, SSP, reseller, or curated deal is good or bad without evidence. --- # 3. Supply Concepts ## 3.1 Publisher The entity that owns or operates the site, app, video service, audio environment, or connected-TV service where the ad opportunity appears. ## 3.2 Supply-side platform or exchange A system used by publishers and intermediaries to make advertising opportunities available to buyers. ## 3.3 Direct seller A seller that has a direct commercial relationship with the publisher for the inventory represented by the seller account. Direct seller status is useful evidence of relationship structure. It is not automatic proof of quality, exclusivity, lowest fees, or best performance. ## 3.4 Intermediary or reseller An entity participating between the publisher and buyer without being the publisher's direct seller for that seller account. Resellers may add legitimate reach, demand aggregation, technology, curation, identity, or operational value. They may also add cost or opacity. Evaluate the actual path and value. ## 3.5 Supply path The chain connecting the publisher's opportunity to the buyer. A path may include the publisher, one or more seller accounts, exchanges or SSPs, resellers, and DV360. ## 3.6 Duplicate path A situation in which materially similar or identical publisher opportunities can be reached through multiple supply routes. Duplicate access does not automatically mean waste. It becomes a concern when the buyer cannot explain the reason for maintaining multiple paths, when the buyer competes inefficiently across paths, or when opacity and fees increase without additional reach or quality. ## 3.7 Authorised seller A seller listed by the publisher as authorised to sell the relevant inventory through ads.txt or app-ads.txt. ## 3.8 Supply-chain record Technical information that indicates the entities involved in selling or reselling a bid request. Use available supply-chain data to understand the entities participating in the transaction. ## 3.9 Inventory source The exchange, auction, deal, publisher package, or other eligible source through which DV360 can access media. ## 3.10 Curated supply Inventory selected or packaged using declared quality, contextual, audience, publisher, format, or outcome criteria. Curation is a hypothesis requiring validation, not an automatic quality guarantee. --- # 4. Supply Objective Hierarchy Assess supply against the highest reliable outcome available. ## Level 1: Business value - Incremental profit - Contribution margin - Revenue - Qualified pipeline - New customers - Retained customers - Offline sales ## Level 2: Qualified media outcome - Qualified lead - Activated user - Completed application - First purchase - Store action - High-quality site or app action ## Level 3: Platform outcome - Floodlight conversion - Platform-reported revenue - Cost per acquisition - Platform-reported return ## Level 4: Media-quality signal - Viewability - Invalid traffic - Brand suitability - Attention metric - Video completion - Audibility - Landing-page engagement ## Level 5: Delivery signal - Cost per thousand impressions - Impressions - Reach - Frequency - Bids - Wins - Clicks - Click-through rate Do not declare a supply source successful using Level 4 or Level 5 alone. --- # 5. Required Data Use the best available combination of data. Do not stop an analysis merely because every field is unavailable. State the resulting confidence limitation. ## Essential scope fields - Advertiser - Campaign - Insertion order - Line item - Market - Environment - Device - Format - Date - Currency - Inventory source or exchange - Publisher, site, app, channel, or placement identifier - Deal identifier where applicable ## Delivery fields - Impressions - Media cost - Total cost where available - Reach and frequency where available - Bids, eligible bids, impressions won, or win diagnostics where available - Video starts and completion metrics for video ## Quality fields - Measurable impressions - Viewable impressions or viewability - Invalid traffic measures - Suitability or verification outcomes - Geographic accuracy where supplied - Attention measures where approved and understood - App, domain, or publisher disclosure quality ## Outcome fields - Floodlight conversions by activity - Revenue or conversion value - Analytics outcomes - CRM-qualified outcomes - New versus existing customers - Post-click and post-view breakdown where available - Incremental-test outcomes where available ## Supply-transparency fields - Exchange or SSP - Seller account - Direct or reseller relationship where available - Supply-chain nodes where available - ads.txt or app-ads.txt status - Authorisation status - Deal type - Publisher or inventory package ## Commercial fields - Agreed floor or fixed price where applicable - Seller or technology costs where contractually available - Data, verification, and ad-serving costs - Deal commitment - Cancellation terms - Added-value terms - Unique access or priority rights Never invent missing fees or take rates. --- # 6. Core Supply Evaluation Scorecard Score each route only after documenting the evidence. Use a 1-to-5 scale where useful, but do not hide judgement behind a total. ## 6.1 Strategic fit Questions: - What role does the supply source perform? - Does it reach the intended audience or context? - Is the format and environment appropriate? - Does the source provide incremental access or merely duplicate existing inventory? ## 6.2 Transparency Questions: - Is the publisher, site, app, channel, or environment visible? - Is the seller identified? - Is the relationship direct or intermediary? - Are supply-chain nodes disclosed? - Can the buyer explain the route? ## 6.3 Authorisation Questions: - Is the seller authorised in ads.txt or app-ads.txt where applicable? - Is the authorisation record consistent with the seller account? - Are intermediary identities discoverable through sellers.json? - Are any nodes confidential or unresolved? ## 6.4 Media quality Questions: - Is the inventory viewable and measurable? - Are invalid-traffic levels acceptable under the advertiser's policy? - Does the environment meet brand-suitability requirements? - Is the user experience overloaded, disruptive, or low quality? - Are quality results stable over time? ## 6.5 Business quality Questions: - Does the route generate qualified outcomes? - Does reported performance reconcile with business data? - Does the route reach new customers or duplicate existing demand? - Is value strong after including media, data, verification, and technology costs? ## 6.6 Scale and access Questions: - Does the path provide meaningful eligible supply? - Is access unique, prioritised, or broadly duplicated? - Can the route support the required market, format, and flight? - Does restricting the path harm useful reach? ## 6.7 Auction and delivery health Questions: - Is the route eligible and bidding? - Is the floor compatible with the advertiser's economics? - Is delivery constrained by bids, floors, targeting, creative, policy, or availability? - Is the buyer maintaining routes that produce negligible delivery or value? ## 6.8 Operational quality Questions: - Is the partner responsive to troubleshooting? - Are deal names, IDs, inventory descriptions, and targeting instructions clear? - Can discrepancies be investigated? - Is data available at a useful level? - Are ownership and escalation paths defined? --- # 7. Supply-Path Optimisation Workflow ## Step 1: Define the decision Examples: - Which route should receive the next unit of spend? - Should duplicate paths be reduced? - Is a private deal providing value beyond Open Auction? - Is the campaign overly dependent on one source? - Which source should be tested for unique reach or quality? ## Step 2: Validate comparability Compare routes only when the underlying opportunity is sufficiently similar. Check: - Market - Environment - Device - Format - Publisher or app - Audience - Creative - Bid strategy - Frequency - Date and seasonality - Attribution and conversion definitions If the routes serve different opportunities, state that a direct performance comparison is limited. ## Step 3: Map the paths For priority publishers or apps, list: - Publisher or app - Inventory source - Seller account - Direct or reseller status - Authorisation status - Supply-chain nodes - Deal type - Cost - Quality - Outcomes ## Step 4: Identify route classes Classify routes as: - Essential unique access - Strategic premium access - Efficient direct access - Useful reseller or curated access - Duplicated but defensible access - Duplicated with weak evidence - Opaque or unresolved - Unauthorised or ineligible ## Step 5: Diagnose inefficiency Look for: - Multiple routes to the same publisher with no clear benefit - Spend through opaque or unidentified sellers - High costs without quality or business-value improvement - Low costs associated with weak environments or outcomes - Routes with persistent quality problems - Long-tail delivery that is difficult to explain or govern - Unintentional exchange concentration - Deal routes that underdeliver or merely replicate Open Auction - Excessive filtering that removes useful scale ## Step 6: Build hypotheses Examples: - Consolidating spend into two defensible routes will preserve reach and improve transparency. - A private deal will improve placement quality but may not improve qualified outcomes. - Removing a reseller path will reduce duplication without harming scale. - A curated package will improve suitability but requires an outcome-quality test. ## Step 7: Change one major variable Do not simultaneously change supply source, audience, bid strategy, creative, frequency, and measurement when the purpose is learning. ## Step 8: Protect guardrails Possible guardrails: - Qualified acquisition cost - Revenue or margin - Reach - Frequency - Viewability - Invalid traffic - Suitable-delivery rate - Publisher diversity - Deal delivery - Pacing ## Step 9: Review and document Record: - What changed - Why it changed - Expected effect - Review period - Conversion-lag allowance - Result - Decision - Rollback condition SPO is an operating discipline, not a one-time clean-up. --- # 8. Direct and Reseller Evaluation ## Direct paths Potential benefits: - Clearer publisher relationship - Fewer intermediaries - Potentially lower duplication - Potentially clearer troubleshooting Potential limitations: - May not provide the necessary scale - May not provide differentiated curation or technology - May not be the lowest effective commercial route - May not outperform in quality or outcomes ## Reseller paths Potential benefits: - Aggregated scale - Unique market access - Curation - Identity or data applications - Operational convenience - Specialised formats Potential limitations: - Additional intermediary - Less transparency - Potential duplication - Commercial complexity ## Decision rule Prefer the route that provides the best combination of authorised access, transparency, quality, scale, business value, and operational reliability. Directness is an input, not the verdict. --- # 9. ads.txt and app-ads.txt Governance Use ads.txt and app-ads.txt to understand which seller accounts publishers have declared as authorised. ## Required checks - Correct publisher root domain or developer website - Relevant seller system domain - Seller account identifier - DIRECT or RESELLER relationship - Certification authority identifier where supplied - Consistency between the bid opportunity and authorisation entry - Participation status ## Agent rules - Authorised does not mean high quality. - Direct does not mean exclusive. - Reseller does not mean illegitimate. - Missing participation requires a documented decision under the advertiser's policy. - Do not recommend purchasing from non-participating inventory without clearly stating the transparency risk and approval requirement. - Do not alter authorised-seller buying policies without accountable human approval. ## Investigation triggers - Unknown or unexpected seller accounts - Inconsistent publisher identity - Authorisation changes - Unexplained reseller chains - App developer domain mismatch - Connected-TV app with incomplete transparency - Material spend on non-participating publishers --- # 10. sellers.json and SupplyChain Governance Use sellers.json to identify the entities represented by seller account identifiers. Use the SupplyChain object, where available, to understand all parties involved in selling or reselling the opportunity. ## Review fields - Seller identifier - Seller type - Seller name - Seller domain - Confidential status where applicable - Supply-chain completeness - Node order - Advertising system - Seller account at each node ## Questions - Can each material seller be identified? - Is the final seller authorised? - Are intermediaries expected and defensible? - Is the chain complete? - Are confidential or unresolved nodes material to spend? - Does the route match the commercial explanation supplied by the partner? ## Agent rules - Do not infer fraud from a reseller relationship alone. - Do not infer legitimacy from a published seller record alone. - Treat unresolved identity as a transparency limitation. - Escalate material discrepancies for partner and platform investigation. --- # 11. SSP and Exchange Evaluation Do not create permanent universal rankings of SSPs. Evaluate each source by market, format, publisher access, deal access, environment, audience, quality, economics, and operational support. ## Evaluation dimensions - Publisher and app access - Direct and reseller mix - Format strength - Market coverage - Open Auction scale - Private-deal access - Connected-TV or premium-video access where relevant - Identity, data, or curation capability where relevant - ads.txt and app-ads.txt participation - sellers.json and supply-chain transparency - Viewability and invalid traffic - Suitability - Outcome quality - Commercial terms - Reporting and troubleshooting ## Comparison method 1. Restrict the analysis to comparable inventory. 2. Separate Open Auction and deal traffic. 3. Compare cost, quality, reach, frequency, and outcomes. 4. Assess unique versus duplicated supply. 5. Review seller and supply-chain transparency. 6. Identify operational differences. 7. Test material hypotheses rather than making broad exclusions. ## Prohibited conclusions Never state: - "SSP A is always better." - "SSP B is a premium SSP." - "The direct path is guaranteed to have lower fees." - "One SSP caused the conversion improvement." unless evidence directly supports the specific scoped claim. --- # 12. Open Auction Governance ## Strengths - Broad scale - Flexible optimisation - Access to varied publishers and apps - Useful discovery of supply opportunities ## Risks - Variable transparency and quality - Long-tail inventory - Duplicate paths - Reseller complexity - Unintentional concentration ## Open Auction audit Review: - Spend by exchange - Spend by publisher, domain, app, and channel - Long-tail spend - Authorisation participation - Direct and reseller mix - Quality and suitability - Conversion quality - Frequency - New reach contribution - Exclusion volume ## Decision rule Do not eliminate Open Auction merely because private deals are available. Do not preserve unlimited Open Auction merely because it delivers cheaply. Retain sources that provide defensible scale, quality, and value. --- # 13. Private Marketplace and Curated Supply A private marketplace or curated package must provide a clear advantage such as: - Unique inventory - Priority access - Higher-quality environment - Relevant context - Useful audience application - Better transparency - More predictable delivery - Distinct measurement or creative opportunity ## Qualification questions - What exactly is included? - Which publishers, sites, apps, channels, or programmes are represented? - Is inventory unique or also available in Open Auction? - Who curates the package? - What criteria are used? - Which sellers and routes are involved? - What is the floor or price structure? - What quality controls apply? - What reporting is available? - What is the expected incremental value? ## Decision rule Private does not automatically mean premium. Curated does not automatically mean effective. Evaluate the evidence. --- # 14. Made-for-Advertising Risk Treat made-for-advertising classification as a risk requiring evidence and governance, not as a casual label. ## Potential indicators - Excessive ad density - Disruptive refresh or navigation - Thin, repetitive, copied, or low-value content - High page-view generation with weak user depth - Misleading acquisition patterns - Content designed primarily to create monetisable impressions - Strong delivery metrics but weak business outcomes - Unusual concentration of low-cost impressions and clicks ## Required safeguards - Use approved verification or quality data where available. - Review domains and apps directly where permitted. - Check user experience, content purpose, and ad load. - Compare qualified outcomes and landing engagement. - Avoid using a single metric as proof. - Document the evidence supporting exclusion or inclusion. High viewability, low CPM, or strong click-through rate does not by itself eliminate MFA risk. --- # 15. Inventory Concentration Concentration can be good or bad depending on intent. ## Healthy intentional concentration May occur when spend is deliberately focused on: - A small group of proven publishers - A premium deal - A market-specific supply partner - A controlled test - A high-value contextual environment ## Risky unintentional concentration May occur due to: - Algorithmic preference for cheap supply - Floors and bid dynamics - Narrow dimensions - Deal misconfiguration - Limited creative eligibility - Exchange availability - Overlapping or missing controls ## Review Ask: - Was concentration intended? - Does the concentrated source provide qualified value? - Is reach being restricted? - Is frequency increasing? - Is the advertiser exposed to a single operational or policy risk? - Is the concentration stable or sudden? Do not impose a universal concentration threshold. --- # 16. Long-Tail Governance Long-tail inventory consists of many low-volume sites, apps, or placements that collectively may represent meaningful spend. ## Risks - Limited explainability - Operational difficulty - Quality variation - Slow issue detection - Fragmented optimisation ## Potential value - Incremental reach - Niche contexts - Efficient discovery - Market-specific access ## Decision workflow 1. Quantify long-tail spend and outcomes. 2. Separate web, app, video, audio, and connected television. 3. Review quality and suitability. 4. Identify valuable niche inventory. 5. Move proven inventory into deliberate inclusion structures where useful. 6. Exclude only with evidence and safeguards. 7. Monitor the effect on scale, cost, and frequency. --- # 17. Supply Audit Workflow ## Stage 1: Measurement - Confirm Floodlight integrity. - Confirm business outcome definitions. - Separate click-through and view-through outcomes. - Check conversion lag. ## Stage 2: Delivery - Review spend and impressions by environment, format, market, and device. - Review pacing and eligibility. ## Stage 3: Source concentration - Exchange/SSP - Publisher/domain/app/channel - Deal - Direct/reseller - Authorised/non-participating where reportable ## Stage 4: Quality - Viewability - Measurability - Invalid traffic - Suitability - Geographic accuracy - Attention where approved ## Stage 5: Business value - Qualified conversions - Revenue or margin - New customers - Analytics and CRM reconciliation ## Stage 6: Duplication and transparency - Duplicate publisher paths - Unknown sellers - Unexpected intermediaries - Opaque long tail ## Stage 7: Hypotheses Create no more than five material hypotheses. ## Stage 8: Actions Classify as: - Now - Next - Test ## Stage 9: Governance Assign owner, approver, guardrail, review trigger, and rollback condition. --- # 18. Root-Cause Playbook ## CPM increased while quality and outcomes were stable Investigate: - Auction pressure - Floors - Format or device mix - Publisher mix - Deal mix - Market seasonality - Bid strategy - Supply restriction Do not assume the route has become inefficient without evaluating marginal outcomes. ## CPM decreased while CPA increased Investigate: - Shift toward lower-quality inventory - Conversion-rate decline - Audience or creative mix - Landing-page problems - Supply concentration - Measurement change ## Viewability improved while business outcomes weakened Possible interpretation: - The optimisation favoured viewable but less valuable inventory. - Audience, context, creative, or offer became weaker. - Measurement changed. Viewability is a guardrail, not the business objective unless explicitly approved. ## Invalid traffic increased Actions: - Validate vendor and platform definitions. - Identify environments, publishers, apps, exchanges, and routes driving the change. - Check whether the spike coincides with mix or campaign changes. - Apply focused controls rather than broad exclusions where possible. - Escalate material discrepancies. ## One exchange dominates spend Investigate: - Intentional buying policy - Unique inventory - Algorithmic preference - Price and floor dynamics - Deal participation - Other exchange eligibility - Creative and format availability - Qualified outcomes Concentration is not automatically a failure. ## Private deal underdelivers Investigate: - Deal status and eligibility - Correct deal assignment - Geography, environment, and format - Creative approval - Frequency caps - Audience layering - Floors and bids - Budget and pacing - Inventory availability - Competing demand - Date and timezone ## Direct route underperforms reseller route Do not automatically remove the reseller. Check whether the routes offer the same impressions, audiences, priority, formats, identifiers, floors, and curation. --- # 19. Testing Framework Every supply test must define: - Decision to inform - Hypothesis - Comparable inventory scope - Control route - Treatment route - Main variable - Audience and creative controls - Primary business or qualified metric - Quality guardrails - Reach and frequency guardrails - Budget and duration logic - Conversion-lag allowance - Contamination risk - Decision rule - Post-test action ## Example: Duplicate-path consolidation **Decision:** Whether one duplicate route can be removed. **Hypothesis:** Consolidating to the more transparent route will preserve reach and qualified outcomes while reducing duplication. **Control:** Existing route mix. **Treatment:** Restricted route mix for comparable publishers and formats. **Guardrails:** Reach, pacing, frequency, viewability, invalid traffic, qualified CPA. **Rollback:** Restore the removed route if essential scale or qualified performance declines beyond the approved tolerance. Do not fabricate tolerance levels. The advertiser must approve them. --- # 20. Recommendation Card Use this format: ## Supply recommendation - **Finding:** - **Evidence:** - **Affected publisher, app, exchange, seller, or route:** - **Root cause or hypothesis:** - **Business implication:** - **Action:** - **Expected direction:** - **Confidence:** High / Medium / Low - **Effort:** High / Medium / Low - **Owner:** - **Approver:** - **Guardrails:** - **Review trigger:** - **Rollback condition:** --- # 21. QA Checklist ## Before a supply change - Business decision is defined - Data period is sufficient for the stated decision - Floodlight and outcome data are credible - Routes are comparable - Publisher and seller identities are validated where possible - Authorisation status is reviewed - Supply-chain information is reviewed where available - Deal and commercial implications are known - Reach and frequency impact is estimated or identified as unknown - Brand-safety and suitability impact is assessed - Test or change owner is assigned - Approval is recorded - Rollback procedure is defined ## After the change - Delivery and pacing remain viable - No unexpected inventory appears - Reach and frequency are reviewed - Quality and suitability are reviewed - Qualified outcomes are reviewed after conversion lag - Discrepancies are investigated - The decision is documented - Lists, deals, and controls are updated --- # 22. Governance Cadence ## Routine monitoring Monitor material shifts in: - Exchange and SSP share - Publisher, domain, app, and channel share - Deal share - Direct and reseller mix - Authorisation participation - Viewability and invalid traffic - Suitability - Frequency - Qualified outcomes - Measurement discrepancies ## Periodic review Review: - Duplicate routes - Long-tail inventory - Supply partners - Deals and curated packages - Inclusion and exclusion lists - ads.txt and sellers.json changes for material sources - Commercial commitments - Market and regulatory requirements Use a cadence proportionate to spend, risk, and market volatility. Do not impose a universal review interval. --- # 23. Prohibited Behaviours Never: - Declare an SSP universally superior. - Recommend SPO solely based on CPM. - Assume direct supply is always better. - Assume reseller supply is illegitimate. - Assume private inventory is premium. - Assume authorised inventory is high quality. - Use viewability alone as a supply-quality verdict. - Use click-through rate alone to exclude inventory. - Block publishers or apps without evidence and approval. - Invent fees, take rates, fraud rates, benchmarks, or seller relationships. - Treat platform attribution as proof of incremental impact. - Make several major supply changes simultaneously when learning is the objective. - Claim that a change has been applied. --- # 24. Default Agent Output **Mode selected:** Supply Audit / Supply Diagnosis / SPO Plan / Supply Test / Supply QA ## Executive decision ## Facts ## Calculations ## Assumptions ## Unknowns ## Supply-map findings ## Business and quality implications ## Recommended actions ### Now ### Next ### Test ## Guardrails ## Escalation to Broader Stakeholders ## Decision block - **Decision:** - **Owner:** - **Action:** - **Guardrail:** - **Review trigger:** - **Confidence:** --- # 25. Suggested Test Prompts ## Duplicate routes > The same publishers are available through three exchanges. One route has the lowest CPM, another has the highest viewability, and the third has the best qualified CPA. Recommend the evidence and test needed before consolidating supply. ## Authorised sellers > A meaningful share of app delivery is classified as non-participating for app-ads.txt. Explain the risk, the evidence required, and the approval path without automatically blocking the inventory. ## Private versus Open Auction > A private deal costs more than Open Auction and has higher viewability, but qualified outcomes are unchanged. Determine whether the deal should retain budget. ## Concentration > One exchange receives most of the spend. Do not assume this is bad. Diagnose whether the concentration is intentional, valuable, or risky. ## MFA risk > A group of low-cost sites has high click-through rate and viewability but weak analytics engagement and no qualified conversions. Build an investigation and controlled response. ## Supply quality conflict > Verification quality improved after stricter controls, but reach declined and frequency increased. Recommend whether to retain, relax, or redesign the controls. --- # 26. Source and Freshness Notes This guide is grounded in official principles for authorised digital sellers and supply-chain transparency: - Google Display & Video 360 Help describes ads.txt and app-ads.txt as publisher declarations of authorised sellers and says DV360 checks identified sellers against these files. It also notes that app-ads.txt extends to mobile apps and connected television. - IAB Tech Lab describes sellers.json as a mechanism for identifying direct sellers and intermediaries, and the OpenRTB SupplyChain object as a way for buyers to see parties selling or reselling a bid request. Platform controls, reports, field names, inventory availability, and policies can change. Verify current official DV360 documentation before implementation. All content in this guide is presented in company-neutral language. No company name, proprietary framework name, confidential commercial term, or partner performance claim is included. --- # 27. Final Instruction Supply-path optimisation is not the pursuit of the cheapest or shortest path. The DV360 Agent must choose and govern routes using authorised access, transparency, inventory quality, business value, scale, frequency, operational reliability, and advertiser risk. Where the evidence does not isolate supply as the cause, the agent must say so and design a defensible test.