# DV360 Troubleshooting & Root Cause Guide ## Enterprise Operating System Guide **Purpose:** Enterprise framework for diagnosing delivery, pacing, bidding, audience, inventory, deal, creative, measurement, quality, reporting, and business-outcome issues in Display & Video 360. **Companion guides:** Use alongside the related operating and specialist guides in this knowledge library. **Core rule:** Diagnose before changing. A symptom can have several causes, and one campaign can have several simultaneous failures. --- # 1. Troubleshooting Mandate The DV360 Agent must identify the narrowest defensible root cause before prescribing action. Every investigation should answer: 1. What is the symptom? 2. When did it begin? 3. Which entities, markets, devices, audiences, inventory, creatives, or outcomes are affected? 4. Is the evidence trustworthy? 5. What changed before the symptom appeared? 6. Which plausible causes have been tested? 7. What is the business risk? 8. What containment, correction, and prevention are required? 9. Who owns the action and approval? 10. What confirms resolution? Troubleshooting is not complete when delivery resumes. The team must confirm the cause, assess downstream impact, document the correction, and prevent recurrence where practical. --- # 2. Diagnostic Principles - Separate symptoms from causes. - Validate measurement before interpreting performance. - Start with simple structural failures before advanced theories. - Review the entire campaign hierarchy. - Check recent changes and copied settings. - Change one major variable where possible. - Use the narrowest correction that addresses the cause. - Protect privacy, policy, rights, contractual commitments, and experiment integrity. - Preserve before-and-after evidence. - Escalate critical risk immediately through the approved process. - Do not invent platform behaviour, thresholds, or resolution status. Official DV360 spend troubleshooting checks budget, active status, matching impressions, creative approval, creative dimensions, targeting, and auction competitiveness. This supports a structured investigation rather than a single-cause assumption. --- # 3. Symptom, Cause, and Contributing Factor ## Symptom The observable issue. Examples: - No spend - Slow pacing - High CPA - Missing conversions - Creative not serving ## Root cause The condition that created the symptom and would prevent recurrence if corrected. ## Contributing factor A condition that increased the impact but did not independently create the issue. ## Example - **Symptom:** A line item stopped spending. - **Root cause:** No eligible creative remained after expiry. - **Contributing factor:** Only one creative had been assigned. - **Correction:** Assign an approved compatible creative. - **Prevention:** Require expiry monitoring and creative redundancy. Do not label the first correlation as root cause. --- # 4. Diagnostic Evidence Pack Gather before material changes: - Advertiser, campaign, insertion-order, and line-item IDs - Current status and dates - Budget and pacing settings - Bid strategy, bid, and target - Inventory sources and deals - Targeting and exclusions - Audience size and status - Frequency settings - Creative assignment and approval - Floodlight selection and event health - Reporting period, timezone, currency, and cost basis - Change history - Publisher, exchange, app, site, device, and market delivery - Verification and suitability evidence - Analytics, CRM, and business outcomes Capture only information needed for the investigation and follow the approved data-handling policy. --- # 5. Universal Diagnostic Decision Tree ## Step 1: Confirm the symptom Reproduce the issue using the correct scope, dates, timezone, filters, and source. ## Step 2: Confirm measurement health Check reporting latency, Floodlights, analytics, attribution, consent, and business-system reconciliation. ## Step 3: Check hierarchy and eligibility Check advertiser, campaign, insertion order, line item, creative, dates, status, budget, and policy. ## Step 4: Check opportunity availability Check targeting, audience, inventory, device, geography, frequency, creative compatibility, and deals. ## Step 5: Check auction participation Check bid, floor, strategy, target, competition, and eligible opportunities. ## Step 6: Check post-impression journey Check creative, click, landing page, form, checkout, app journey, offer, product availability, and conversion. ## Step 7: Segment the symptom Review market, audience, supply, creative, device, format, time, and customer stage. ## Step 8: Identify recent changes Check deployments, bulk uploads, budgets, bids, audiences, supply, creative, URLs, tracking, consent, and attribution. ## Step 9: Test the hypothesis Use platform evidence, a controlled correction, or an experiment. ## Step 10: Close the incident Document cause, correction, impact, owner, approval, validation, prevention, and learning. --- # 6. Severity and Response ## Critical Examples: - Wrong advertiser, budget, market, or destination - Material privacy, policy, legal, or rights failure - Broken primary measurement affecting optimization - Unapproved creative serving - Contractual commitment at immediate risk Contain affected activity through the authorised process, notify accountable owners, preserve evidence, and document recovery. ## High Material delivery, measurement, quality, or business-outcome risk requiring prompt correction and approval. ## Medium Contained operational degradation with manageable impact. ## Low Housekeeping or limited issue that does not materially threaten outcomes or compliance. Severity must follow approved enterprise policy. The Agent must not invent classification. --- # 7. No-Spend Playbook Investigate in this order: 1. Is the relevant entity active? 2. Are the insertion-order and line-item dates current? 3. Is budget available at every controlling level? 4. Is pacing configured as intended? 5. Are eligible and approved creatives assigned? 6. Do creative dimensions, duration, or format match available inventory? 7. Does targeting produce eligible opportunities? 8. Is the audience active and viable? 9. Has frequency exhausted the reachable population? 10. Are private deals accessible and assigned? 11. Is the bid or strategy competitive for eligible opportunities? 12. Are brand-safety, viewability, technology, or inclusion controls too restrictive? Potential corrections must retain the approved business, quality, and policy boundaries. Do not remove controls indiscriminately merely to create spend. --- # 8. Slow-Spend and Underdelivery Playbook ## Structural causes - Limited audience or geography - Narrow inventory - Missing creative sizes or formats - Tight frequency - Bid below competitive or deal conditions - Restrictive suitability or viewability settings - Deal underavailability - Shortened flight or delayed approval ## Diagnostic questions - Is spend consistently slow or newly slow? - Which restriction removes the most opportunity? - Is lost delivery concentrated by market, device, exchange, deal, or creative? - Is the campaign protecting quality or failing to access intended inventory? - Is the remaining budget achievable without compromising strategy? ## Actions - Correct setup failure. - Add approved compatible creative. - Expand suitable inventory or audience. - Rebalance budget. - Adjust bid or pacing with approval. - Renegotiate or replace underdelivering deals. - Reforecast and disclose residual risk. --- # 9. Overspend and Rapid-Pacing Playbook Check: - Approved budget versus platform budget - Currency - Daily versus flight budget - Pacing type - Duplicated entities - Automated allocation - Recently added budget - Unexpected inventory access - Flight dates - Status changes Containment should protect the approved commercial plan. After correction, reconcile spend and assess whether reporting, billing, or commitments were affected. Do not label strong demand as the cause until budget and structure errors have been excluded. --- # 10. Bid and Auction Troubleshooting ## Low auction participation Check: - Bid amount - Automated target - Deal floor - Inventory price - Audience and data cost - Format and device - Competition - Quality restrictions ## High win or spend with weak value Check whether the strategy accesses low-value supply, weak audiences, poor creative contexts, or misleading conversion signals. ## Automated bidding instability Check: - Event accuracy - Conversion volume and lag - Value quality - Target changes - Budget changes - Structural changes - Audience or inventory expansion Avoid repeated bidding changes that prevent clear diagnosis. --- # 11. Audience Troubleshooting ## Audience does not deliver Check: - List status - Market availability - Membership duration - Data freshness - Audience size - Inclusion and exclusion conflict - First-party matchability - Inventory compatibility - Additional targeting layers - Frequency ## Audience delivers but business quality is weak Check: - Audience role - Source and provenance - Creative relevance - Inventory and device mix - Destination - Conversion definition - New versus existing customers ## Retargeting saturates Check pool growth, recency, suppression, frequency distribution, creative rotation, view-through attribution, and marginal conversions. Do not conclude the audience is the cause until media and measurement conditions are considered. --- # 12. Geography, Language, and Device Troubleshooting ## Wrong market delivery Check inherited geography, location logic, duplicated line items, deals, bulk changes, and reporting filters. ## Low local delivery Check population, audience size, inventory, language, device, creative approval, bid, and local policy. ## Device concentration Check bid adjustments, creative compatibility, inventory availability, audience behaviour, format, and automated optimization. ## Language mismatch Check platform targeting, creative language, page context, browser or user signals where applicable, and landing-page language. Treat material unapproved geography as a high-risk incident. --- # 13. Inventory and Supply Troubleshooting ## Unexplained publisher concentration Check inventory eligibility, bid strategy, floor, audience availability, creative compatibility, deal assignment, and supply-path duplication. ## Cost rises after supply restrictions Check whether exclusions removed cheap or high-scale opportunities, concentrated demand, reduced reach, or increased bid pressure. ## Quality weakens after expansion Check newly accessed exchanges, sites, apps, sellers, devices, formats, and audiences. ## Duplicate paths Review publisher and seller relationships, direct and reseller routes, authorised-seller signals, SupplyChain information where available, and outcome differences. Do not classify a publisher or seller as fraudulent from one missing signal or weak KPI alone. --- # 14. PMP and Deal Troubleshooting ## Deal accepted but not delivering Check: - Deal ID - Advertiser and seat access - Dates - Line-item assignment - Inventory type - Floor and bid - Seller availability - Targeting conflicts - Creative requirements - Frequency - Status ## Deal underdelivers Compare actual eligible supply with the documented forecast or commitment. Check delayed launch, creative approval, targeting, frequency, and seller-side restrictions. ## Deal costs more without clear benefit Assess unique access, priority, certainty, audience, context, quality, and business outcome. ## Deal delivery differs from publisher report Align period, timezone, currency, impression definition, fees, invalid-traffic treatment, and reporting source. Do not assume a private deal guarantees delivery unless the agreement explicitly does so. --- # 15. Creative Approval and Serving Troubleshooting ## Creative rejected Check the platform reason, format, content, claims, destination, category, technical requirements, and required review. ## Approved creative not serving Check assignment, line-item type, dimensions, duration, market, device, inventory, rotation, dates, rights, and landing page. ## One creative dominates Check eligibility, available sizes, format coverage, platform optimization, rotation, audience, and inventory mix. ## Clicks but no qualified engagement Check misleading message, accidental interaction, rendering, destination speed, product match, offer, form, and tracking. Never claim approval from an upload status alone. State the actual approval stage. --- # 16. Landing-Page and Journey Troubleshooting Check: - Final URL - Redirects - Tracking parameters - Market and language - Product and offer - Page availability - Mobile experience - Page speed - Consent - Cross-domain tracking - Form or checkout - Confirmation event - App-store destination ## Common symptoms - High clicks, low sessions - High sessions, low journey starts - High journey starts, low completion - Conversion tag fires without completed business outcome Map the customer journey stage by stage. Do not change media to compensate for a broken destination. --- # 17. Floodlight and Conversion Troubleshooting Floodlight setup foundations include Counter and Sales activities, counting methods, revenue capture, custom variables, and tag implementation. ## No conversions Check: - Correct activity selected - Attribution enabled where intended - Tag firing - Consent - Event trigger - Date and report filters - Actual business volume - Reporting latency - Cross-domain or app journey ## Too many conversions Check duplicate firing, refresh behaviour, repeated submissions, counting method, test activity, broad micro-events, and attribution windows. ## Incorrect revenue Check currency, transaction value, order ID, gross or net basis, product data, refunds, cancellations, and default values. ## Platform and CRM disagree Align event definitions, attribution, windows, latency, qualification, cancellations, and identity matching. --- # 18. Click and Session Discrepancy Playbook Potential causes include: - Click and session definitions - Consent - Ad blockers or browser restrictions - Redirect failure - Landing-page speed - Cross-domain configuration - Repeated clicks - Accidental clicks - Invalid traffic - Reporting timezone - URL-parameter handling Compare trends and segmented differences rather than expecting exact equality. State which source governs engagement decisions. --- # 19. Reporting and Data Troubleshooting ## Totals change unexpectedly Check source updates, filters, dimensions, metric compatibility, attribution, latency, timezone, and field-definition changes. ## Download differs from interface Check report type, saved configuration, sampled or modelled data where applicable, date logic, row limits, currency, and update timing. ## Duplicate rows after transformation Check joins, entity keys, many-to-many mappings, creative associations, and duplicated extracts. ## Missing markets or entities Check filters, status, permissions, naming, data latency, and extraction logic. Preserve the raw extract and reconciliation evidence. --- # 20. CPM Root-Cause Tree When CPM rises: 1. Confirm cost basis and currency. 2. Check market and seasonality. 3. Check format and device mix. 4. Check audience restriction and data cost. 5. Check publisher, exchange, deal, and floor mix. 6. Check viewability and suitability controls. 7. Check bidding changes and competition. 8. Check whether quality or unique reach improved. 9. Check whether the higher cost reduced business value. A higher CPM is not automatically a failure. It may reflect better supply or scarcity. The decision depends on marginal value. --- # 21. CTR and CPC Root-Cause Tree When CTR falls or CPC rises: 1. Confirm click and impression definitions. 2. Check objective relevance. 3. Check creative concept and fatigue. 4. Check audience and customer stage. 5. Check placement, app, site, device, and format. 6. Check rendering and accidental-click patterns. 7. Check frequency and reach. 8. Check destination quality. 9. Check qualified outcomes. Do not optimise awareness or CTV activity solely toward clicks. --- # 22. Conversion-Rate and CPA Root-Cause Tree When conversion rate falls or CPA rises: 1. Confirm the conversion and denominator. 2. Check event health and attribution. 3. Apply conversion lag. 4. Check cost and media price. 5. Check click or session quality. 6. Check audience and customer mix. 7. Check supply, device, market, and format. 8. Check creative and landing continuity. 9. Check offer, availability, form, checkout, or sales process. 10. Check CRM-qualified outcome. Select the correction according to the actual driver. Do not lower quality merely to improve platform CPA. --- # 23. ROAS and Revenue Root-Cause Tree When ROAS weakens: - Confirm revenue source. - Confirm cost basis. - Confirm currency. - Review average order value. - Review product mix. - Review gross versus net revenue. - Review cancellations and refunds. - Review margin. - Review new versus existing customers. - Review attribution and incrementality. An improving ROAS can still conceal falling profit. A weakening ROAS can occur while total incremental profit grows. State the business decision clearly. --- # 24. Reach and Frequency Root-Cause Tree ## Low reach Check audience, geography, inventory, device, creative compatibility, bids, floors, frequency, deal concentration, and duplicated exclusions. ## Rising frequency and flat reach Check saturation, retargeting size, overlapping line items, separate caps, limited supply, and budget growth. ## Reach differs across systems Check unit, identity, deduplication, included inventory, modelled values, device or household treatment, and reporting period. Do not sum incompatible reach figures. --- # 25. Viewability and Quality Troubleshooting ## Low measurability Check environment, app integration, format, device, provider coverage, and inventory source. ## Low viewability Check publisher, app, placement, position, format, device, creative, supply route, and targeting. ## High invalid traffic Check source concentration, apps, domains, sellers, device, geography, and measurement coverage. Escalate according to the approved invalid-traffic policy. ## Quality improves but outcomes worsen Check cost inflation, reach loss, audience relevance, frequency, creative, and destination. Do not compare providers as if methodologies are identical. --- # 26. Format-Specific Diagnostics ## Display Check dimensions, web or app environment, viewability, placement quality, clicks, landing pages, audience, frequency, and conversions. ## Video Check subtype, duration, view definition, skippability, quartiles, completion, device, inventory, audience, sound, and brand or business outcome. ## CTV Check app identity, publisher, deal, device or household reach, frequency, completion, creative, rights, cross-device attribution, and incremental-reach methodology. ## Audio Check inventory type, publisher, device, companion, creative mix, completion definition, reach, frequency, rights, and brand or search measurement. ## Native Check component completeness, rendering, cropping, truncation, publisher context, click quality, viewability coverage, destination, and tracking compatibility. --- # 27. Privacy, Consent, and Observability Incidents Potential symptoms: - Sudden conversion decline - Session discrepancy - Missing audience growth - Reduced matchability - Modelled reporting changes Check: - Consent implementation - Tag behaviour - Market requirements - Browser and device effects - URL parameters - Data transfer - Identity permissions - Release dates Separate actual business-demand movement from reduced observable data. Involve the approved privacy and engineering owners. Do not bypass consent to restore reporting. --- # 28. Bulk-Change and Duplication Incidents Bulk operations can reproduce one error across many entities. Check: - Source file version - Entity IDs - Dates - Budgets - Status - Floodlight selection - Targeting - Deals - Creative assignment - URLs - Platform validation messages Use a before-state export, identify affected entities, correct the narrowest scope, sample-check results, and reconcile totals. Duplicated entities require full QA because inherited settings may be wrong. --- # 29. Managed-Service Troubleshooting For managed service: - Request evidence for setup, delivery, optimization, and change history. - Reconcile partner reporting with ad-server and business sources. - Confirm whether the issue sits with the brief, partner implementation, publisher, technology, measurement, or business journey. - Use the agreed escalation and correction process. - Keep one accountable internal owner. Do not accept a partner summary as sufficient evidence when material delivery or measurement risk exists. --- # 30. Root-Cause Scenario Library ## Spend but no impressions in selected report Check entity scope, report filters, data latency, metric compatibility, and whether spend belongs to another entity or period. ## Impressions but no clicks Check objective, format, clickability, creative, landing setup, tracking, audience, and inventory. Some formats are not click-led. ## Clicks but no conversions Check sessions, landing page, product, offer, form, checkout, conversion event, audience intent, and lag. ## Strong platform conversions, flat business outcomes Check attribution, view-through share, duplicate events, conversion quality, CRM latency, customer status, and incrementality. ## Weak platform conversions, strong sales Check tagging gaps, offline or cross-device journeys, attribution, analytics, CRM timing, and other media. ## A new audience consumes most spend Check eligibility, bid strategy, expansion, pool size, inventory availability, and whether the change was intended. ## Performance collapses after budget increase Check marginal headroom, frequency, inventory quality, audience expansion, auction pressure, and lag. ## Performance collapses after allowlist Check scale loss, supply concentration, price inflation, reach, frequency, and whether the quality benefit offsets the loss. ## Creative metrics improve but sales fall Check whether media allocation, audience, inventory, destination, offer, or customer quality changed simultaneously. ## App inventory drives clicks but weak sessions Check accidental interaction, app quality, software-development-kit behaviour, redirects, load time, and invalid traffic. ## CTV reports many view-through conversions Check household or cross-device attribution, windows, frequency, existing demand, app quality, and causal evidence. ## Native shows unusually high CTR Check rendering, disclosure, accidental clicks, publisher concentration, device, landing engagement, and qualified outcomes. ## Audio completion is high but search is flat Check creative recall, branding, audience, inventory, frequency, study method, and whether completion is forced or automatic. ## Deal launches late Check negotiation, acceptance, creative approval, advertiser access, dates, forecast freshness, and line-item assignment. ## One market underperforms the region Check currency, local benchmarks, inventory, audience, language, creative localization, offer, tracking, and business environment. ## Reporting changes after attribution update Restate the model and windows, quantify the effect, preserve the prior view, and notify stakeholders. --- # 31. Hypothesis Testing Framework For each plausible cause, record: - Hypothesis - Supporting evidence - Conflicting evidence - Test or check - Result - Confidence Prioritize hypotheses using: - Likelihood - Potential impact - Ease of validation - Reversibility - Risk Do not eliminate a hypothesis because one supporting metric is absent when measurement coverage is incomplete. --- # 32. Corrective Action Framework ## Contain Limit immediate risk without destroying evidence. ## Correct Fix the confirmed cause at the narrowest useful level. ## Validate Confirm delivery, measurement, quality, and business journey after correction. ## Prevent Update the checklist, naming, architecture, alert, approval, training, or automation control. ## Learn Record the incident in reusable knowledge and link it to the relevant guide. --- # 33. Troubleshooting QA Before closing an issue, confirm: - The symptom is no longer present or accepted. - The root cause is supported by evidence. - Contributing factors are documented. - The correction affected the intended entities only. - Measurement remains healthy. - Privacy, policy, rights, and commitments remain protected. - Downstream reporting and billing impacts are assessed. - The owner and approver accepted closure. - Prevention and monitoring are documented. - Rollback remains available where required. --- # 34. Incident Record Template ```markdown # DV360 Incident Record - Incident ID: - Date identified: - Advertiser and campaign: - Affected entities and markets: - Symptom: - Severity: - Business impact: - Delivery impact: - Measurement impact: - Privacy, policy, rights, or commercial impact: - First observed time: - Recent changes: - Evidence: - Hypotheses considered: - Confirmed root cause: - Contributing factors: - Containment: - Corrective action: - Responsible owner: - Accountable approver: - Validation evidence: - Prevention: - Review trigger: - Rollback condition: - Closure date: ``` --- # 35. Troubleshooting RACI Template ```markdown # DV360 Troubleshooting RACI | Activity | Responsible | Accountable | Consulted | Informed | |---|---|---|---|---| | Identify and log incident | | | | | | Classify severity | | | | | | Preserve evidence | | | | | | Validate measurement | | | | | | Diagnose campaign setup | | | | | | Diagnose audiences | | | | | | Diagnose supply and deals | | | | | | Diagnose creative and destination | | | | | | Contain critical risk | | | | | | Approve corrective action | | | | | | Implement correction | | | | | | Validate recovery | | | | | | Reconcile reporting and billing | | | | | | Notify stakeholders | | | | | | Approve closure | | | | | | Implement prevention | | | | | ``` Assign one accountable owner for each incident. --- # 36. Troubleshooting Change Control Template ```markdown # Troubleshooting Change Request - Change ID: - Incident ID: - Current state: - Proposed correction: - Root-cause evidence: - Affected entities: - Expected result: - Delivery risk: - Measurement risk: - Commercial risk: - Privacy, policy, or rights risk: - Contractual or experiment dependency: - Responsible owner: - Accountable approver: - Implementation window: - QA reviewer: - Monitoring metrics: - Review trigger: - Rollback condition: - Rollback steps: - Resolution evidence: - Final status: ``` --- # 37. Escalation Template ```markdown # DV360 Issue Escalation - Incident ID: - Severity: - Decision required: - Business impact: - Current containment: - Confirmed facts: - Unknowns: - Root cause or leading hypothesis: - Options: - Recommended action: - Trade-offs: - Required owner: - Required approver: - Decision deadline: - Monitoring and rollback: ``` Use the decision deadline only when explicitly defined by the business or incident process. --- # 38. Post-Incident Review Review: - What happened? - What was the verified cause? - Why did existing controls not prevent or detect it? - How quickly was the issue contained? - What was the delivery, measurement, commercial, or customer impact? - Was communication effective? - Which control should change? - Is training, documentation, automation, or monitoring required? - Does another campaign share the same risk? Focus on process improvement rather than individual blame. --- # 39. Troubleshooting Maturity Model ## Level 1: Reactive Issues are solved through individual memory and undocumented edits. ## Level 2: Repeatable Basic checklists and escalation paths exist. ## Level 3: Structured Decision trees, evidence packs, RACI, change control, and incident records are used consistently. ## Level 4: Preventive Monitoring, recurring-issue analysis, bulk-risk controls, and cross-campaign audits reduce incidents. ## Level 5: Learning system Incident evidence improves setup, optimization, measurement, reporting, training, and automation. Recommend the next practical maturity step. --- # 40. Prohibited Behaviours Never: - Change settings before confirming the symptom and scope. - Assume one cause explains every issue. - Remove privacy, suitability, policy, fraud, or rights controls to restore spend. - Increase bids before checking budget, status, creative, targeting, and floors. - Blame an audience, publisher, partner, or creative without fair evidence. - Present attribution discrepancy as tracking failure without checking definitions. - Treat platform recommendations as automatic approval. - Change several major variables and claim one solved the problem. - Hide an incident that affected reporting, billing, or commitments. - Claim resolution without validation evidence. - Invent platform limits, thresholds, timelines, or approvals. - Claim a live correction was implemented without evidence. --- # 41. Default Agent Output **Mode selected:** Delivery Incident / Performance Diagnosis / Measurement Incident / Deal Issue / Creative Issue / Reporting Issue / Root-Cause Audit ## Executive decision ## Symptom and scope ## Severity ## Facts ## Calculations ## Assumptions ## Unknowns ## Recent changes ## Hypotheses considered ## Root cause ## Contributing factors ## Containment ## Corrective action ## Prevention ## QA status ## RACI ## Review trigger ## Rollback condition --- # 42. Source and Freshness Notes Official DV360/Google guidance checks budget availability, active status, targeting matches, creative approval, creative dimensions, private inventory, and auction competitiveness. Platform interfaces, diagnostics, field names, report latency, product support, and deal workflows can change. Verify current official documentation and approved enterprise standards before implementation. --- # 43. Final Enterprise Instruction Troubleshooting is a controlled evidence process, not reactive editing. The DV360 Agent must confirm the symptom, validate data, inspect hierarchy and eligibility, identify recent changes, test competing hypotheses, correct the narrowest verified cause, protect mandatory controls, and validate recovery. When evidence is insufficient, the Agent must state the uncertainty and define the next diagnostic check rather than manufacture a root cause. --- # 44. Resolution Verification Checklist A correction is verified only when: - The intended entity shows the corrected state. - Delivery or measurement resumes under the expected conditions. - No unrelated entity changed. - Budget, pacing, targeting, frequency, deals, and creative remain aligned. - Floodlight and business-system signals remain coherent. - Reporting and billing impact have been assessed. - The monitoring period reflects the relevant delivery and conversion lag. - The accountable owner accepts the evidence. If the symptom improves but the suspected cause was not demonstrated, classify the issue as mitigated rather than conclusively resolved. Continue monitoring or design a stronger diagnostic check. --- # 45. Cross-Campaign Risk Review After a confirmed incident, search for the same risk across other live or planned campaigns. Review: - Shared templates - Duplicated line items - Common Floodlight activities - Shared audiences - Shared deals - Shared creative - Common landing pages - Bulk-upload source files - Global exclusions and verification settings This review prevents a local fix from leaving the same defect elsewhere. Record which campaigns were checked, what evidence was reviewed, and whether preventive action was applied.