# DV360 Optimization Guide ## Enterprise Operating System Guide Purpose: Enterprise framework for campaign optimization, decision governance, pacing management, bid strategy management, audience optimization, supply optimization, creative optimization, budget allocation, experimentation, reporting feedback loops, and operational decision making in Display & Video 360. # 1. Optimization Principles Optimization exists to improve business outcomes, not platform metrics in isolation. Every optimization decision must answer: - What changed? - Why did it change? - Is the change material? - What evidence supports action? - What business outcome is expected? - What risk is introduced? Optimization without diagnosis is guessing. # 2. Optimization Operating Cycle Daily: - Delivery health - Pacing - Critical errors - Supply availability - Creative status - Measurement health Weekly: - Audience analysis - Supply analysis - Creative analysis - KPI review - Budget shifts - Frequency review Monthly: - Strategy review - Incrementality review - Structure review - Vendor review - Benchmark review Quarterly: - Portfolio review - Partner review - Framework updates # 3. Optimization Hierarchy Optimization should occur in this order: 1. Measurement accuracy 2. Delivery eligibility 3. Budget pacing 4. Inventory quality 5. Audience quality 6. Creative quality 7. Bidding efficiency 8. Business outcome efficiency Never optimize conversion rates before confirming conversion tracking works. # 4. Optimization Governance Every major optimization requires: - Owner - Evidence - Expected outcome - Review date - Rollback condition Track: - What changed - When changed - Why changed - Impact observed # 5. Decision Classification Immediate: - Tracking failures - Delivery failures - Creative rejection - Deal failure Planned: - Budget shifts - Bid changes - Audience strategy - Supply strategy Experimental: - New audiences - New supply - New creative - New bidding methods # 6. Pacing Optimization Framework Review: - Spend versus plan - Spend versus available budget - Flight progress - Inventory availability Questions: - Is underdelivery structural? - Is overspend intentional? - Is pacing aligned with objective? Do not chase spend if quality is deteriorating. # 7. Underdelivery Optimization Review: - Status - Dates - Budget - Bid competitiveness - Inventory eligibility - Audience size - Frequency caps - Geography - Creative approval Potential actions: - Increase bid - Expand supply - Expand audience - Relax restrictions - Adjust pacing # 8. Overspend Optimization Review: - Budget controls - Pacing settings - Bid strategy - Learning phase Actions: - Reduce pacing - Tighten bids - Reallocate budgets - Protect priority inventory # 9. Bid Optimization Framework Bid decisions should reflect: - Objective - Attribution framework - Conversion quality - Competition - Supply conditions Avoid constant bid adjustments without evidence. # 10. Automated Bidding Governance Review: - Signal quality - Conversion volume - Learning status - Stability Optimization signals must be: - Accurate - Timely - Meaningful Poor signals create poor bidding outcomes. # 11. Audience Optimization Review: - Reach - Frequency - Conversion rate - CPA - Quality indicators Audience decisions: - Scale - Retain - Suppress - Expand - Test # 12. Prospecting Optimization Review: - Reach efficiency - Frequency distribution - Quality visits - New-customer outcomes Avoid evaluating prospecting solely through last-click outcomes. # 13. Retargeting Optimization Review: - Audience freshness - Conversion lag - Frequency pressure - Incremental value Retargeting should not consume budget without evidence of additional value. # 14. First-Party Audience Optimization Validate: - Data freshness - Match quality - Audience scale - Segmentation logic Review performance against business outcomes. # 15. Third-Party Audience Optimization Review: - Source quality - Cost - Reach - Conversion rate Do not retain expensive audiences solely due to historical usage. # 16. Supply Optimization Framework Review: - Exchange contribution - Inventory quality - Reach contribution - Cost efficiency - Conversion quality Supply optimization should focus on business contribution rather than arbitrary publisher preferences. # 17. Open Auction Optimization Review: - Scale - Pricing - Viewability - Outcome quality Open Auction provides learning opportunities but requires governance. # 18. PMP Optimization Review: - Win rates - Delivery - Quality - Incremental contribution Verify PMPs justify premium pricing. # 19. Programmatic Guaranteed Optimization Review: - Commitment delivery - Reach - Frequency - Quality Protect contractual obligations before reallocating spend. # 20. Inventory Concentration Review Review: - Publisher concentration - Exchange concentration - Deal concentration High concentration requires explanation. # 21. Creative Optimization Framework Creative is often the largest performance driver. Review: - CTR - View rate - Completion rate - Conversion rate - Engagement quality Creative optimization should occur continuously. # 22. Creative Fatigue Framework Signals: - Declining CTR - Declining engagement - Rising CPA - Stable frequency Actions: - Refresh creative - Rotate assets - Change messaging # 23. Display Creative Optimization Review: - Sizes - Messaging - CTA - Landing experience Separate creative impact from audience impact before making changes. # 24. Video Optimization Review: - View rate - Completion rate - Quartile progression - Conversions Analyze creative retention points. # 25. Native Optimization Review: - Headline performance - Image performance - Publisher performance Native performance often varies by rendering environment. # 26. Audio Optimization Review: - Reach - Frequency - Completion - Brand outcomes Audio requires broader effectiveness measurement. # 27. CTV Optimization Review: - Reach - Household exposure - Incremental reach - Completion Completion alone is insufficient evidence of effectiveness. # 28. Frequency Optimization Framework Review: - Average frequency - Distribution - High-frequency users Decision rules: - Avoid excessive repetition. - Evaluate by objective. # 29. Reach Optimization Framework Review: - Unique reach - Incremental reach - Cost efficiency Reach optimization should align with campaign role. # 30. Measurement-Led Optimization Optimization should respect: - Measurement framework - Attribution framework - Benchmark governance - Incrementality evidence Never optimize towards a broken metric. # 31. Conversion Optimization Review: - Attribution settings - Conversion quality - CRM outcomes - Revenue outcomes Platform conversions should be reconciled against business data. # 32. CPA Optimization Framework Review: - Quality - Volume - Incremental contribution Lower CPA is not always better. # 33. ROAS Optimization Framework Review: - Revenue source - Margin implications - Customer quality ROAS should not replace business profitability review. # 34. Lead Quality Optimization Review: - Lead quality - Qualification rates - Opportunity rates Optimize for qualified outcomes where possible. # 35. Budget Reallocation Framework Before shifting budget: - Confirm measurement quality - Confirm statistical relevance - Confirm scalability Document: - Source budget - Destination budget - Expected effect # 36. Marginal Investment Analysis Review: - Incremental outcomes - Diminishing returns - Capacity constraints Questions: - Where is next dollar most valuable? - Which activity is saturated? # 37. Optimization Decision Tree If KPI worsens: 1. Validate tracking. 2. Validate delivery. 3. Validate audience. 4. Validate supply. 5. Validate creative. 6. Validate bidding. If KPI improves: 1. Confirm measurement. 2. Assess scalability. 3. Evaluate sustainability. # 38. Learning Agenda Framework Document: - Hypothesis - Change - Evidence - Result - Recommendation Optimization should create institutional knowledge. # 39. Optimization Testing Framework When evidence is insufficient: - Test - Measure - Learn - Scale Do not disguise tests as proven optimizations. # 40. Automation Governance Automation should be monitored. Review: - Rules - Alerts - Automated decisions Humans remain accountable. # 41. Daily Optimization Checklist - Budget pacing - Delivery health - Inventory quality - Floodlight health - Creative approval - Critical risks # 42. Weekly Optimization Checklist - Audience review - Supply review - Creative review - KPI review - Frequency review - Budget review # 43. Monthly Optimization Checklist - Structure review - Partner review - Performance review - Measurement review - Benchmark review # 44. Optimization RACI Template Responsible: - Trader - Analyst Accountable: - Campaign owner Consulted: - Measurement - Client lead Informed: - Stakeholders # 45. Optimization Change Control Record: - Change - Owner - Evidence - Expected impact - Rollback condition # 46. Root Cause Library High CPM: - Inventory scarcity - Floor pricing - Audience restrictions Low CTR: - Creative - Audience mismatch - Weak placement quality Low CVR: - Landing page - Audience intent - Tracking High CPA: - Frequency - Audience quality - Bid strategy Low reach: - Restrictive targeting - Frequency constraints - Limited inventory # 47. Optimization Audit Review: - Structure - Measurement - Audiences - Supply - Creative - Bidding - Budgeting Classify: - Retain - Improve - Test - Retire # 48. Optimization Maturity Model Level 1 Reactive. Level 2 Metric driven. Level 3 Structured. Level 4 Measurement aligned. Level 5 Incrementality informed. # 49. Enterprise Optimization Template - Objective - KPI - Finding - Evidence - Action - Owner - Review date - Risk # 50. Final Enterprise Instruction Optimization should improve business outcomes through evidence-based decisions. The DV360 Agent must diagnose before acting, document before changing, measure after changing, and maintain alignment with measurement, attribution, QA, benchmark, and governance frameworks. --- # 51. Enterprise Optimization Decision System Every optimization begins with a decision. ## Decision 1: Is the data trustworthy? Check: - Reporting period and timezone - Currency and cost basis - Floodlight activity - Counting method - Attribution model and windows - Analytics and CRM reconciliation - Conversion lag - Recent tagging or consent changes If the data is unreliable, stop performance optimization and correct the measurement issue first. ## Decision 2: Is delivery structurally healthy? Check: - Active status - Dates - Budget and pacing - Bids and floors - Creative eligibility - Audience scale - Inventory access - Deal participation - Frequency constraints If the campaign cannot access sufficient eligible opportunities, diagnose delivery before changing the performance target. ## Decision 3: Is the comparison fair? Align: - Objective - Market - Format - Environment - Audience role - Inventory route - Creative - Attribution - Cost basis If material conditions differ, label the conclusion directional. ## Decision 4: Is the issue cost, quality, or volume? Decompose the KPI into its drivers. For example: `CPA = Cost / Conversions` A higher CPA can result from higher media cost, weaker conversion quality, lower conversion rate, less credited attribution, or a temporary lag. Each cause requires a different action. ## Decision 5: Is the action reversible? Irreversible or contractual decisions require stronger evidence and approval. ## Decision 6: Is a controlled test required? Use a test when several plausible causes remain or when the proposed change materially affects strategy, customer experience, measurement, or commercial commitment. --- # 52. KPI Deterioration Decision Tree When an approved KPI worsens: 1. **Validate measurement.** - Confirm event volume, attribution, reporting scope, and business-system reconciliation. 2. **Validate delivery.** - Check pacing, status, eligibility, market, device, and inventory access. 3. **Identify the mathematical driver.** - Separate cost, response, conversion, value, and volume effects. 4. **Segment the movement.** - Review market, line item, audience, publisher, exchange, deal, device, creative, and time. 5. **Check structural changes.** - Identify new creative, audience, inventory, bids, budgets, landing pages, offers, or tracking. 6. **Assess materiality.** - Do not react to normal volatility or insufficient evidence. 7. **Select the narrowest useful action.** - Correct the driver rather than applying a broad exclusion. 8. **Record the change.** - Include evidence, owner, approver, review trigger, and rollback. 9. **Measure after lag.** - Allow sufficient time for delivery and outcome maturation. --- # 53. KPI Improvement Decision Tree When an approved KPI improves: 1. Confirm that tracking and definitions did not change. 2. Identify which segments contributed to the improvement. 3. Check whether volume, quality, or business value was sacrificed. 4. Review frequency, reach, inventory concentration, and customer mix. 5. Determine whether the improvement is repeatable or temporary. 6. Assess remaining scalable inventory and audience headroom. 7. Estimate marginal, not average, returns. 8. Protect experiment integrity and contractual commitments. 9. Scale through a controlled budget or bid change. 10. Define a rollback condition if performance regresses. Do not scale because an average KPI is attractive if incremental volume is already becoming less efficient. --- # 54. Pacing Playbook ## Underspend with healthy quality Investigate restrictions in this order: - Status and dates - Budget and pacing setup - Creative approval - Bid and floor relationship - Inventory availability - Audience scale - Geography and device - Frequency - Deal access Potential actions: - Correct setup errors - Increase bids where additional access is valuable - Expand suitable supply - Broaden audiences with safeguards - Add compatible creative - Adjust pacing ## Underspend with weak quality Do not solve delivery by opening low-quality inventory blindly. Correct measurement, creative, audience, supply, or destination before expanding. ## Overspend risk Check: - Incorrect budget - Aggressive pacing - Duplicated structures - Automated allocation - Unexpected inventory access - Currency or date errors Use the Setup & Trafficking Guide for structural corrections. --- # 55. Automated Bidding Playbook Before using or changing automated bidding, confirm: - The selected event matches the business objective. - Event firing is accurate. - Values are meaningful and current. - Conversion lag is understood. - The campaign has enough usable signal for the intended strategy. - Major structural changes are not occurring simultaneously. - The target is commercially realistic. ## When delivery is limited Check whether the target, bid, inventory, audience, budget, or measurement signal is constraining participation. ## When efficiency improves but volume falls Assess whether the strategy is avoiding valuable opportunities to protect the reported average. ## When volume improves but quality falls Review event quality, audience expansion, inventory mix, view-through attribution, and CRM outcomes. ## Governance rules - Avoid repeated target changes before the effect of the previous change can be assessed. - Record the before state. - Preserve conversion definitions during evaluation. - Do not treat a platform recommendation as automatic approval. - Human owners remain accountable for the business outcome. --- # 56. Value-Based Bidding Playbook Value-based optimization requires a governed value signal. Review: - Value source - Gross versus net revenue - Currency - Refunds and cancellations - Margin - Product mix - New-customer value - Offline latency - Missing or default values ## Decision rules - Do not send arbitrary values solely to activate a strategy. - Do not optimize to gross revenue when margin materially changes the decision. - Check whether a few high-value records dominate learning. - Validate that values arrive within a useful timeframe. - Compare platform value with finance or commerce systems. - Protect minimum volume and customer-quality guardrails. --- # 57. Audience Optimization Playbook ## Step 1: Confirm the audience role Separate prospecting, remarketing, retention, exclusion, and seed use. ## Step 2: Validate delivery Review spend, reach, frequency, inventory, device, market, and creative. ## Step 3: Validate business quality Review qualified sessions, leads, customers, revenue, margin, and new-customer contribution. ## Step 4: Diagnose overlap and saturation Check duplicated eligibility, internal competition, repeated exposure, and slowing marginal reach. ## Step 5: Select action - Retain when the role remains useful. - Scale when marginal headroom exists. - Consolidate when segments are operationally redundant. - Suppress when users are ineligible or already completed the outcome. - Refresh when data is stale. - Test when quality or expansion is uncertain. - Retire when the audience lacks value or permitted use. Do not declare an audience poor when creative, placement, destination, or measurement differs materially. --- # 58. Prospecting Playbook Prospecting should be judged on its ability to create valuable new demand or customer outcomes. Review: - New eligible reach - Frequency - Qualified engagement - New-customer rate - Assisted contribution - Inventory and creative quality - Incremental evidence Potential actions: - Broaden targeting when precision is limiting reach. - Add contextual or publisher-based routes. - Use approved first-party seeds for controlled expansion. - Consolidate overlapping audiences. - Refresh creative for broader customer motivations. - Protect existing-customer suppression. Do not force prospecting to match remarketing CPA without considering role and incrementality. --- # 59. Remarketing Playbook Review: - Inclusion event - Membership duration - Customer stage - Completed-outcome suppression - Pool size - Reach and frequency - Creative sequence - Conversion lag - View-through share Potential actions: - Split genuinely different intent stages. - Shorten or lengthen recency only with evidence. - Suppress completed users promptly where appropriate. - Reduce exposure when marginal outcomes flatten. - Introduce stage-specific creative. - Test incrementality where remarketing receives substantial budget. Strong attributed performance may reflect captured existing demand. Do not scale without checking additional value. --- # 60. Supply Optimization Playbook ## Review dimensions - Exchange or SSP - Publisher - Domain or app - Deal - Direct or reseller path - Device and environment - Viewability and invalid traffic - Reach contribution - Qualified outcomes ## Decision rules - Do not optimize solely to CPM. - Do not exclude solely for low CTR. - Distinguish publisher from supply route. - Protect unique reach and useful context. - Review long-tail spend and unresolved supply. - Compare measured quality using compatible methodologies. - Apply exclusions at the narrowest effective level. ## Actions - Consolidate duplicate paths. - Shift toward more efficient authorised routes. - Add high-quality supply where scale is constrained. - Negotiate or test private access. - Exclude persistently harmful or valueless inventory with evidence. - Maintain an exception and reinstatement process. --- # 61. PMP Optimization Playbook Assess whether the deal provides value through: - Unique inventory - Priority - Delivery certainty - Audience - Context - Quality - Commercial terms ## Underperforming PMP Check: - Deal acceptance and assignment - Floor and bid - Dates - Audience and targeting conflicts - Creative compatibility - Seller availability - Frequency - Measurement ## High-cost PMP Compare the premium with the specific advantage delivered. Do not judge on CPM alone. ## Renewal decision Choose: - Renew - Renegotiate - Restructure - Retest - Reduce - Exit Document evidence, alternative supply, commitments, and review conditions. --- # 62. Creative Optimization Playbook ## Diagnose at four levels - Concept - Execution - Format adaptation - Element ## Review - Reach and frequency - Inventory and device mix - Viewability or completion - Landing continuity - Qualified outcomes - Brand outcomes - Rights and expiry ## Actions - Expand a proven concept through meaningful executions. - Refresh stale executions. - Adapt to missing formats or sizes. - Correct rendering or destination issues. - Retire expired or persistently weak assets. - Test materially different customer ideas. Do not call a creative the winner when rotation, audience, inventory, or budget conditions differ materially. --- # 63. Frequency and Reach Playbook ## Rising frequency and flat reach Check audience size, inventory restrictions, deal concentration, overlapping line items, device constraints, and budget growth. ## Low frequency and weak recall or action Check whether the message requires repeated exposure, whether creative is recognisable, and whether the campaign reaches the intended audience. ## High frequency and stable outcomes Stable average performance does not prove additional exposures remain valuable. Review marginal outcomes and heavy-frequency groups. ## Actions - Broaden suitable reach. - Consolidate overlapping structures. - Coordinate caps. - Refresh creative. - Reduce spend where marginal reach is exhausted. - Test alternative frequency strategies. Do not apply one universal cap across formats and objectives. --- # 64. Budget Reallocation Playbook Before moving budget, compare the source and destination on: - Business role - Measurement confidence - Qualified outcome - Marginal headroom - Reach and frequency - Inventory quality - Creative readiness - Contractual commitments - Test requirements ## Reallocation sequence 1. Validate data. 2. Protect guaranteed and approved commitments. 3. Estimate the opportunity lost from the source. 4. Estimate the marginal opportunity at the destination. 5. Define the change size. 6. Obtain approval. 7. Implement and record. 8. Review after sufficient delivery and lag. 9. Roll back if the guardrail fails. Do not move all budget from a weaker average performer when the source provides unique reach, learning, or incremental value. --- # 65. Marginal Investment Framework Average efficiency describes historical output. Marginal efficiency estimates the value of the next investment. ## Evidence inputs - Spend progression - Outcome progression - Reach progression - Frequency progression - Inventory availability - Audience size - Bids and floors - Conversion lag - Business capacity ## Saturation signs - Spend rises faster than qualified outcomes. - Reach growth slows. - Frequency concentrates. - CPM or CPA increases. - Lower-quality supply or audiences absorb incremental spend. ## Decision rules - Scale in controlled increments. - Use scenarios when evidence is sparse. - Protect test cells and strategic reach. - Express uncertainty. - Do not claim a precise marginal return without sufficient data. --- # 66. Format-Specific Optimization Playbooks ## Display Review size coverage, placements, viewability, audience, creative, frequency, destination, and qualified outcomes. ## Video Review format, duration, view definition, quartile retention, completion, audience, inventory, reach, frequency, and business or brand outcomes. ## Connected television Review app and publisher transparency, household or device reach, frequency, completion, incremental reach, creative, deals, and cross-device measurement. ## Audio Review publisher and content type, reach, frequency, completion where supported, creative recall, companion, search or brand outcomes, and measurement limitations. ## Native Review component rendering, publisher context, viewability coverage, click quality, landing continuity, audience, and qualified outcomes. Use the dedicated format guide before implementing material changes. --- # 67. Root-Cause Scenarios ## High CPM and stable conversion rate Possible causes include auction pressure, higher-quality supply, seasonality, floors, audience restriction, device mix, or deals. Determine whether the additional cost creates sufficient value before acting. ## Stable CPM and falling conversion rate Review audience mix, creative, landing page, offer, product availability, placement mix, consent, and conversion definition. ## CTR rises while qualified outcomes fall Review clickbait, accidental clicks, intrusive placements, audience quality, landing mismatch, and tracking. ## Viewability rises while CPA worsens Review cost inflation, supply shift, reduced reach, frequency concentration, audience relevance, and whether viewability was over-prioritised. ## Completion rises while brand outcome is flat Review duration, forced exposure, branding, message, audience, inventory, study power, and contamination. ## Platform CPA improves while CRM quality falls Review optimisation event, audience expansion, supply, creative promise, lead validation, and feedback latency. ## ROAS improves while profit falls Review gross versus net value, margin, discounting, refunds, product mix, fulfilment cost, and customer type. ## Reach improves while frequency also rises Check whether expansion reached new users broadly or concentrated additional exposure in overlapping segments. ## Prospecting has weak last-click performance Review assisted contribution, new-customer outcomes, qualified engagement, reach, creative, and incrementality before reducing investment. ## Remarketing has excellent CPA and high frequency Review existing demand, pool size, completed-outcome suppression, view-through attribution, marginal conversions, and customer experience. ## One publisher dominates conversions Check spend, audience, creative, route, attribution, device, and conversion quality before consolidating further. ## Open Auction outperforms PMP Check comparability and the deal's role. PMP may provide certainty, context, quality, or unique reach not captured by average CPA. ## Automated bidding underdelivers Review target realism, event quality, conversion volume, lag, inventory, audience, budget, and learning disruption. ## Budget increase causes immediate efficiency loss Check marginal headroom, frequency, inventory quality, audience expansion, bid pressure, and conversion lag. ## Performance changes after creative launch Separate creative effect from changes in allocation, audience, inventory, device, landing page, and seasonality. ## Conversions fall after privacy implementation Separate real demand from reduced observability. Review CRM, analytics, modelled reporting, consent, and implementation dates. --- # 68. Optimization QA Framework Before implementing an optimization, confirm: - The business objective is unchanged. - Measurement is healthy. - The comparison is fair. - Conversion lag is considered. - The root cause is supported by evidence. - Contractual commitments remain protected. - Experiment integrity remains protected. - Privacy, policy, suitability, and rights remain protected. - The action is approved at the correct level. - A review trigger and rollback condition exist. After implementing, confirm: - The correct entities changed. - The before and after states are recorded. - No unintended budget, targeting, deal, creative, or Floodlight change occurred. - Delivery remains healthy. - Guardrails remain within approved limits. - Business outcomes are reviewed after appropriate lag. --- # 69. Optimization QA Template ```markdown # DV360 Optimization QA - Optimization ID: - Advertiser and campaign: - Market: - Decision: - Business objective: - Primary KPI: - Measurement status: - Root-cause evidence: - Affected entities: - Current state: - Proposed state: - Expected direction: - Delivery guardrail: - Quality guardrail: - Business guardrail: - Contractual dependencies: - Experiment dependencies: - Responsible owner: - Accountable approver: - Review trigger: - Rollback condition: - Implementation evidence: - Outcome after lag: - Final decision: ``` --- # 70. Enterprise Optimization RACI ```markdown # DV360 Optimization RACI | Activity | Responsible | Accountable | Consulted | Informed | |---|---|---|---|---| | Validate measurement health | | | | | | Monitor pacing and delivery | | | | | | Diagnose KPI movement | | | | | | Recommend bid change | | | | | | Approve bid or target change | | | | | | Recommend budget reallocation | | | | | | Approve budget reallocation | | | | | | Optimise audiences | | | | | | Optimise supply and deals | | | | | | Optimise creative rotation | | | | | | Change frequency controls | | | | | | Protect experiment integrity | | | | | | Validate CRM or revenue impact | | | | | | Complete optimization QA | | | | | | Approve rollback | | | | | | Record and share learning | | | | | ``` RACI rules: - Assign one accountable owner per material decision. - Separate recommendation from approval where risk is material. - Consult measurement, commercial, creative, privacy, supply, and client owners where relevant. - Update the RACI when the service model or campaign ownership changes. --- # 71. Optimization Change Control Framework Material changes include: - Budget or pacing - Bid strategy, bid, or target - Primary optimization event - Attribution or window - Audience inclusion, exclusion, or expansion - Inventory source, allowlist, blocklist, or deal - Frequency - Creative rotation or destination - Experiment cell ## Decision sequence 1. Is this a correction, optimization, or experiment? 2. Is there sufficient evidence? 3. Does it affect business objective, measurement, commercial commitment, or test integrity? 4. Who approves the decision? 5. What dependencies must be notified? 6. Is the action reversible? 7. What will be monitored? 8. When will the decision be reviewed? 9. What triggers rollback? Urgent corrections should still be documented after immediate risk containment. --- # 72. Optimization Change Control Template ```markdown # DV360 Optimization Change Request - Change ID: - Request date: - Requestor: - Campaign scope: - Change category: - Current state: - Proposed state: - Finding and evidence: - Root cause or hypothesis: - Expected outcome: - Expected trade-off: - 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 date or trigger: - Rollback condition: - Rollback steps: - Final outcome: - Learning captured: ``` --- # 73. Optimization Decision Log ```markdown ## Optimization Decision: [ID] - Date: - Decision: - Facts: - Calculations: - Assumptions: - Unknowns: - Options considered: - Selected action: - Reason: - Owner: - Approver: - Guardrails: - Review trigger: - Result: - Follow-up: ``` Maintain the log so decisions become reusable knowledge. --- # 74. Enterprise Operating Rhythm ## Daily control Focus on critical risk, trafficking health, pacing, measurement, creative approval, deal participation, and material anomalies. ## Weekly decision review Focus on business KPI, audience, supply, creative, frequency, bids, budget, and the change log. ## Monthly strategic review Focus on marginal returns, structure, partner and deal value, creative pipeline, measurement reconciliation, experiments, and benchmark status. ## Quarterly learning review Focus on reusable decisions, market differences, saturation, incrementality evidence, operating maturity, and guide updates. Cadence should reflect spend, risk, campaign duration, conversion lag, contractual obligations, and stakeholder requirements. Do not apply one universal schedule. --- # 75. Optimization Audit Template ```markdown # DV360 Optimization Audit - Business objective: - Primary source of truth: - Measurement health: - Campaign structure findings: - Pacing findings: - Bid-strategy findings: - Audience findings: - Supply and deal findings: - Creative findings: - Reach and frequency findings: - Business-outcome findings: - Incrementality evidence: - Critical risks: - Now actions: - Next actions: - Tests: - RACI gaps: - Change-control gaps: - Owners and approvers: ``` --- # 76. Expanded Prohibited Behaviours Never: - Optimize toward a broken or ambiguous event. - Change several major variables and attribute the result to one. - Reallocate budget solely from average platform CPA. - Scale remarketing without reviewing saturation and incrementality. - Exclude supply solely for low CTR. - Judge awareness activity solely by last click. - Loosen suitability, privacy, fraud, or rights controls to spend budget. - Break an experiment through unapproved mid-test changes. - Ignore guaranteed commitments. - Use an unapproved benchmark as a hard target. - Claim an optimization caused improvement without suitable evidence. - Claim a live change was implemented without platform evidence. --- # 77. Final Agent Output **Mode selected:** Delivery Recovery / KPI Diagnosis / Audience Optimization / Supply Optimization / Creative Optimization / Budget Reallocation / Marginal Investment / Optimization Audit ## Executive decision ## Facts ## Calculations ## Assumptions ## Unknowns ## Root cause ## Options considered ## Recommended action ## Expected trade-off ## QA status ## RACI ## Change control ## Review trigger ## Rollback condition ## Classification ### Now ### Next ### Test --- # 78. Final Enterprise Instruction Optimization is a controlled business decision process, not a sequence of reactive platform edits. The DV360 Agent must validate measurement, diagnose the mathematical and operational driver, compare fair segments, protect quality and commitments, select the narrowest useful action, record ownership and approval, and assess results after the appropriate lag. When evidence cannot isolate the cause, the Agent must design a controlled test instead of manufacturing certainty.