# DV360 Audience Guide ## Project Knowledge for Claude and ChatGPT **Version:** 1.0 **Purpose:** Give the DV360 Agent practical, evidence-led decision logic for first-party audiences, prospecting, remarketing, suppression, contextual targeting, marketplace data, publisher audiences, audience expansion, overlap, saturation, privacy, measurement, testing, and quality assurance. **Companion guides:** Use alongside the related operating and specialist guides in this knowledge library. --- # 1. Purpose This guide teaches the DV360 Agent how to design and govern audience strategy without treating an audience label as proof of relevance or performance. Audience strategy is the disciplined choice of: - Who the advertiser should reach - Who the advertiser should exclude - Which data source should be used - Whether audience, contextual, inventory, or broad delivery is the best mechanism - How customer lifecycle stages should be separated - How much precision the plan can afford before scale becomes impractical - How audience quality will be measured - How audience expansion and automation will be controlled - How privacy, consent, and data-use restrictions will be protected The objective is not to create the highest number of segments. The objective is to create a small, understandable audience system that improves qualified business outcomes and produces reusable learning. --- # 2. Agent Mandate When using this guide, the DV360 Agent must: 1. Start from the advertiser's business objective and DV360's assigned role. 2. Define the customer's lifecycle stage before selecting a segment. 3. Identify the origin, owner, permissions, freshness, and intended use of each audience. 4. Distinguish targeting boundaries from audience signals used for expansion or optimisation. 5. Separate prospecting, retargeting, retention, and suppression. 6. Evaluate audiences using qualified outcomes, not click-through rate alone. 7. Diagnose overlap, saturation, scale, inventory quality, creative, and measurement before blaming the audience. 8. Use contextual or broad strategies when they are more transparent, scalable, or suitable. 9. Design controlled audience tests when causality is uncertain. 10. Require human approval before audience activation, suppression changes, customer-data use, or material expansion. 11. Escalate cross-channel customer strategy and budget allocation to the appropriate human decision-owner. --- # 3. Audience Philosophy ## 3.1 Audience labels are hypotheses Names such as "in-market," "luxury traveller," "small-business owner," or "high-value customer" describe an intended segment. They do not prove that every reachable user matches the label or that the segment will create advertiser value. ## 3.2 Relevance is not the same as incrementality An audience can be highly likely to convert because it contains existing demand. This can make attributed performance strong while incremental value remains limited. ## 3.3 Precision has a cost Every additional targeting layer can reduce eligible reach, increase auction pressure, fragment learning, and concentrate frequency. Precision is useful only when the resulting audience remains large enough and commercially valuable. ## 3.4 Broad delivery is not automatically wasteful Broad or automated delivery may discover valuable users, but it requires credible conversion signals, adequate budget, suitable inventory, measurement, and guardrails. ## 3.5 Context can outperform identity When identity signals are unavailable, weak, expensive, restricted, or unnecessary, contextual targeting can offer a transparent way to align media with relevant content and moments. ## 3.6 Audience quality is a business question The right segment produces useful reach, qualified outcomes, and learning under acceptable privacy, cost, and operational constraints. --- # 4. Audience Source Hierarchy Use this hierarchy as a starting point, not a universal ranking. ## Tier 1: Advertiser first-party outcomes Examples: - Purchasers - Qualified leads - Subscribers - Loyalty members - Product users - High-value customers - Lapsed customers Strength: - Closest relationship to actual advertiser outcomes Limitations: - Scale - Matchability - Consent - Freshness - Bias toward known customers ## Tier 2: Advertiser behavioural signals Examples: - Product viewers - Form starters - Cart abandoners - Content engagers - App users - Store or journey events Strength: - Indicates observable interaction with the advertiser Limitations: - Can be shallow - May include low-quality traffic - Requires accurate event design ## Tier 3: Publisher first-party audiences and contexts Strength: - Can provide differentiated content or audience access Limitations: - Definition transparency - Portability - Measurement - Availability - Commercial cost ## Tier 4: Platform audiences and custom signals Examples may include affinity, in-market, demographic, life-event, or custom-list options where available. Strength: - Scale and operational accessibility Limitations: - Shared availability to other advertisers - Limited differentiation - Modelled definitions - Market and policy variation ## Tier 5: Marketplace or third-party audiences Strength: - May provide specialised attributes or market coverage Limitations: - Data provenance - Freshness - Accuracy - Duplication - Cost - Privacy and regional availability ## Tier 6: Contextual, inventory, and broad delivery These are not inferior simply because they do not rely on a named audience. They may be preferable when they offer greater transparency, scale, contextual relevance, or privacy durability. --- # 5. Required Audience Record For every audience, document: - Audience name - Business purpose - Lifecycle stage - Data owner - Data source - Collection method - Eligibility criteria - Exclusion criteria - Creation date - Refresh method - Membership duration - Geographic availability - Estimated or observed scale - Consent and permitted use - Sensitive-data review - Activation platforms - Cost where applicable - Measurement approach - Known limitations - Approval owner - Retirement date or review trigger Do not activate an audience solely because it already exists in the platform. --- # 6. Lifecycle Architecture Build the audience system around customer states. ## Prospect People without a known completed outcome. Possible subgroups: - Broad eligible prospects - Category-interested prospects - High-intent prospects - Contextually relevant prospects - Publisher audiences - Modelled or expanded prospects ## Engaged prospect People who have interacted but not completed the priority outcome. Examples: - Product viewers - Video or content engagers - Form starters - Cart abandoners - App visitors ## Lead or applicant People who have submitted an inquiry or application but whose qualification or outcome may still be unknown. ## Customer People who have completed the commercial outcome. Possible subgroups: - New customer - Existing customer - High-value customer - Product-specific customer - Lapsed customer - Ineligible customer ## Suppressed user A user excluded because reaching that user would be wasteful, inappropriate, restricted, or inconsistent with the campaign role. --- # 7. First-Party Audience Governance First-party data can be highly valuable because it reflects actual advertiser relationships or behaviour. It is not automatically accurate, complete, consented, or suitable for every use. ## Readiness checks - Business definition is clear - Source system is known - Data owner is accountable - Consent and permitted use are documented - Data transfer is approved and secure - Identifiers are handled according to policy - Records are fresh enough for the use case - Duplicates and invalid records are addressed - Customer status is accurate - Audience size is viable - Activation and retention periods are appropriate ## Use cases - Suppression - Retargeting - Retention and cross-sell - High-value customer modelling - Lapsed-customer reactivation - Seed for approved expansion - Outcome-quality feedback ## Decision rules - Do not use all customers as one audience when value, product, geography, or lifecycle materially differs. - Do not over-segment beyond viable scale. - Do not assume a large customer list will match at the same size in DV360. - Do not use first-party data outside its approved purpose. - Do not confuse list match rate with audience quality. --- # 8. CRM Audience Framework CRM audiences should connect media to business-qualified customer states. ## Recommended CRM states - Lead created - Qualified lead - Sales-qualified opportunity - Customer - High-value customer - Product owner - Lapsed customer - Not contactable - Invalid or fraudulent lead - Ineligible user ## CRM use decisions Ask: - Is the status current? - Is the status consistently applied? - Are important markets or customer groups missing? - Does the sales process create selection bias? - How often will the list refresh? - What does suppression remove from prospecting? - Can qualified outcomes be returned to measurement and optimisation? ## Quality warning If CRM status is delayed or inconsistently maintained, the agent must lower confidence and avoid aggressive use in bidding, modelling, or suppression. --- # 9. Remarketing Framework Remarketing should reconnect eligible users with a clear unfinished customer task. It should not endlessly expose everyone who visited the advertiser. ## Recommended structure Separate audiences where intent and messaging differ, such as: - Product or service viewer - Journey starter - Form or application starter - Cart abandoner - Prior engager - Recent purchaser - Lapsed customer ## Required decisions - What event qualifies the user? - Which completed event removes the user? - How long is the user relevant? - What message matches the stage? - What frequency is responsible? - What outcome is expected? - When does retargeting stop being useful? ## Decision rules - Separate retargeting and prospecting when their roles, economics, optimisation, and frequency differ. - Suppress completed outcomes promptly where appropriate. - Do not use one generic remarketing pool for unrelated customer states. - Do not interpret strong remarketing attribution as automatic incrementality. - Avoid overexposure to small pools. --- # 10. Suppression Framework Suppression protects budget, customer experience, eligibility, and measurement clarity. ## Potential suppression groups - Completed purchasers or applicants - Existing customers when prospecting for new customers - Employees or internal users where permitted - Invalid or fraudulent leads - Users outside product eligibility - Recent converters during a cooling-off period - Users who withdrew consent - Audiences restricted by policy or law ## Suppression QA - Correct event or CRM state - Correct geographic and product scope - Correct recency - Refresh cadence - Inclusion and exclusion conflict check - Minimum viable audience and platform eligibility - Cross-line-item consistency - Approved exception handling ## Risk Over-suppression can remove valuable cross-sell, repeat-purchase, loyalty, or household reach. Define the business purpose before excluding. --- # 11. Prospecting Framework Prospecting should identify new eligible users with a credible path to value. ## Prospecting approaches - Broad eligible delivery - Platform in-market or affinity audiences - Custom lists or intent-oriented segments where available - Modelled expansion from first-party seeds - Contextual targeting - Publisher audiences - Marketplace audiences - Curated inventory packages ## Selection questions - What evidence connects the audience to the business objective? - Is the segment differentiated? - Is scale viable? - Is the definition transparent? - Is the cost justified? - Can existing customers be suppressed? - Can qualified outcomes be measured? - Is contextual targeting a cleaner alternative? ## Rules - Do not stack many prospecting segments in one line item if the purpose is to learn which audience works. - Do not fragment every segment into a separate line item without enough budget. - Do not judge prospecting only by immediate last-click outcomes. - Monitor new-customer contribution where possible. --- # 12. Contextual Targeting Framework DV360 targeting includes content-related controls alongside audience, inventory, and technology targeting. Official DV360 guidance explains that line items bid on impressions matching their targeting settings and that positive targeting can automatically exclude other items of the same type. Any layering decision must therefore consider its effect on available supply. ## Contextual inputs - Keywords - Categories - Topics - URLs or placements - Publisher sections - Content or programme metadata where available - Language - Brand suitability ## Appropriate uses - Privacy-durable reach - New product or category contexts - Hard-to-identify audiences - Controlled content adjacency - Market-specific cultural or editorial environments - Testing audience labels against relevant environments ## Decision rules - Contextual relevance is not proof that the user is in-market. - Avoid keyword lists that are vague, overly restrictive, or policy-sensitive. - Consider content volume and available inventory. - Evaluate environment quality and qualified outcomes. - Do not add audience and contextual restrictions simultaneously without understanding scale loss. --- # 13. Platform Audience Framework Where available, platform audience options can support reach across interest, intent, demographic, and behavioural dimensions. ## Evaluation Assess: - Role in the journey - Market availability - Segment size - Definition transparency - Cost - Inventory access - Qualified outcome - Overlap with first-party and other platform segments - Suitability for awareness, consideration, or performance ## Custom-list caution Official DV360 documentation describes custom affinity and custom intent as distinct custom-list types and notes that forecasting and inventory-availability reporting can be limited for custom lists. The agent must not invent audience reach where the system does not provide it. ## Rules - Affinity-style audiences normally require awareness or broad consideration logic. - Intent-oriented audiences may support more active consideration but remain modelled signals. - Demographics should be used only when legally, ethically, and strategically justified. - Sensitive-keyword and targeting-policy restrictions must be respected. - Platform segments are not differentiated simply because they sound specific. --- # 14. Marketplace Audience Framework Third-party or marketplace audiences require additional scrutiny. ## Qualification questions - Who owns and supplies the data? - How is the segment built? - Is it observed, declared, inferred, modelled, or combined? - What is the geographic and temporal coverage? - How recent is the data? - What identifiers are used? - How is consent handled? - What is the expected matchable scale? - Is the audience exclusive? - What does the audience cost? - What validation evidence exists? ## Decision rules - Do not buy a segment based on its name alone. - More expensive data is not automatically more accurate. - Cheap data is not automatically efficient. - Evaluate media cost plus data cost against qualified value. - Where transparency is insufficient, treat the audience as low confidence and test it against a credible baseline. --- # 15. Publisher Audience Framework Publisher audiences may provide differentiated access to readers, viewers, listeners, subscribers, or customer groups. ## Required evidence - Publisher and inventory included - Audience definition - Data source - Recency - Match or activation method - Consent and permitted use - Expected scale - Pricing - Placement restrictions - Reporting available - Whether the audience is available elsewhere ## Testing requirement Compare the publisher audience against: - The same publisher without the audience - A contextual package - Another relevant audience - Broad eligible inventory Control for format, creative, frequency, market, dates, and measurement where possible. --- # 16. Modelled and Expanded Audience Governance Audience expansion or optimised targeting may extend delivery beyond selected audience inputs when available and enabled. ## Readiness gate - Accurate primary conversion - Sufficiently useful outcome signal - Clear prospecting role - Existing-customer suppression where appropriate - Budget to learn - Suitable inventory - Business-quality reporting - Test or control plan ## Decision rules - Selected audience inputs may act as starting signals rather than strict boundaries when expansion is active. - Do not describe expanded delivery as if every user belongs to the seed audience. - Do not enable expansion in pure remarketing when the business requires a strict retargeting boundary. - Do not judge expansion on platform CPA alone. - Review new-customer, CRM, and incremental quality where possible. - Verify current product naming, availability, and behaviour before implementation. --- # 17. Combined Audience Logic Use include, exclude, AND, OR, and other available logical combinations only when the business purpose is clear. ## AND logic Can increase precision but reduce scale significantly. ## OR logic Can expand reach but make it harder to understand which component created value. ## Exclusion logic Can protect eligibility and customer experience but may over-restrict reach. ## Rules - Document the logic in plain language. - Forecast or assess scale where possible. - Avoid complex combinations that traders and reviewers cannot explain. - Where learning matters, keep segment logic interpretable. - Check for contradictory inclusions and exclusions. --- # 18. Audience Overlap Framework Overlap can create duplicated reach, internal competition, confusing attribution, and excessive frequency. ## Diagnose overlap across - First-party lists - Remarketing windows - Prospecting segments - Publisher audiences - Marketplace data - Contextual and audience line items - DV360 and other platforms ## Review questions - Are the same users eligible for multiple line items? - Is overlap intentional? - Are bids and budgets competing? - Is frequency managed at the correct level? - Are audiences receiving different messages? - Can outcome reporting distinguish roles? ## Actions - Consolidate similar audiences when no distinct role exists. - Use exclusions to preserve test integrity or lifecycle separation. - Maintain overlap when different inventory, format, or message roles justify it. - Escalate cross-platform overlap to the appropriate human decision-owner. Do not assume overlap is always waste. Some overlap is necessary for cross-format reach and sequence. --- # 19. Saturation Framework Audience saturation occurs when additional investment produces limited useful reach or increasingly repetitive exposure. ## Indicators - Reach growth slows while impressions continue - Frequency rises - Qualified outcome growth slows - Marginal cost increases - Creative response weakens - Small audience pools dominate delivery - Exclusions or narrow targeting limit new supply ## Do not use one threshold There is no universal saturation frequency. Consider format, objective, recency, creative rotation, customer journey, audience size, and outcome lag. ## Potential actions - Expand eligible inventory - Broaden audience logic - Add contextual reach - Refresh creative - Adjust frequency - Separate lifecycle stages - Reduce budget - Test expansion Identify the actual constraint before acting. --- # 20. Reach Versus Precision Use this decision model. ## Favour greater precision when - Eligibility is strict - Product relevance is narrow - Customer value varies materially - The audience definition is credible - Scale remains viable - Waste has a high commercial or regulatory cost ## Favour broader reach when - The objective is awareness or discovery - Available demand is larger than known segments - Conversion signals are credible - Context and inventory are suitable - Narrow definitions create high cost or low scale - Learning requires wider exploration ## Balanced design Use a controlled prospecting structure that protects essential exclusions while allowing enough scale for useful delivery and learning. --- # 21. Frequency Governance Frequency should reflect audience size, objective, format, creative, flight, recency, and customer stage. ## Review levels - Advertiser or campaign - Insertion order - Line item - Audience - Creative - Publisher or inventory environment - Cross-platform exposure where measurable ## Questions - Is the cap applied at the right level? - Is frequency concentrated in a small audience? - Is the campaign gaining new reach? - Does more exposure improve or weaken qualified outcomes? - Are different formats contributing to the same exposure burden? - Is reporting based on users, devices, households, or another method? ## Rules - Do not apply a universal cap. - Retargeting usually requires stricter recency and relevance controls than broad awareness. - Connected television or shared-device frequency needs careful interpretation. - Cross-platform frequency may not be fully observable; state the limitation. --- # 22. Audience Quality Scorecard Score each audience on evidence, not labels. ## Strategic fit Does it solve the defined customer problem? ## Provenance Is the source and construction transparent? ## Permission Is the use approved and consented? ## Freshness Is the audience current enough? ## Scale Can it deliver useful reach without excessive frequency? ## Differentiation Is access unique or widely available? ## Business quality Does it generate qualified outcomes and value? ## Incremental value Does it reach or influence customers beyond existing demand? ## Cost Are data and media costs justified? ## Operability Can the audience be refreshed, measured, explained, and governed? Use a 1-to-5 scale only when it improves transparency. Explain the evidence behind every score. --- # 23. Audience Audit Workflow ## Step 1: Define the business decision ## Step 2: Validate measurement - Floodlight activities - CRM or analytics outcomes - New versus existing customer logic - Attribution and conversion lag ## Step 3: Inventory the audiences - Source - Lifecycle stage - Size - Cost - Freshness - Permissions - Current activation ## Step 4: Map line-item use - Included audiences - Excluded audiences - Expansion setting - Inventory - Creative - Budget - Frequency ## Step 5: Review delivery - Spend - Reach - Frequency - Inventory mix - Cost - Pacing ## Step 6: Review quality - Qualified outcomes - Revenue or value - Landing engagement - New-customer mix - Media quality ## Step 7: Diagnose - Overlap - Saturation - Scale loss - Stale data - Weak definitions - Suppression failures - Audience-expansion effects - Inventory or creative confounders ## Step 8: Prioritise Create no more than five material actions under Now, Next, and Test. --- # 24. Root-Cause Playbook ## Audience CPM rises, conversion rate remains stable Investigate scale, auction pressure, inventory, frequency, seasonality, data cost, and target restriction before changing the audience. ## Audience CPA rises because conversion rate falls Investigate audience mix, recency, customer eligibility, creative-message fit, landing experience, offer, and measurement. ## Retargeting has outstanding platform CPA Check: - Existing demand - View-through contribution - Conversion lag - Exclusion logic - Incremental value - Frequency - New versus existing customer mix ## Marketplace audience has strong CTR but weak outcomes Possible causes: - Weak underlying accuracy - Click-oriented inventory - Creative curiosity without product fit - Landing mismatch - Attribution limits Do not scale on CTR. ## Contextual audience outperforms modelled audience Check comparability of inventory, creative, market, bidding, frequency, and attribution before concluding that context caused the improvement. ## Broad expansion lowers CPA Validate qualified outcomes, customer mix, reach, inventory quality, and incrementality before scaling. ## Audience underdelivers Check: - List size and matchability - Geography - Inventory - Recency - Membership duration - Inclusion/exclusion conflicts - Additional targeting layers - Frequency - Bidding - Creative eligibility - Market availability --- # 25. Audience Testing Framework Every test must define: - Decision to inform - Hypothesis - Audience source and definition - Control - Treatment - Main variable - Inventory, creative, and bidding controls - Primary qualified outcome - Reach and frequency guardrails - Media-quality guardrails - Budget and duration logic - Conversion-lag allowance - Overlap and contamination risks - Decision rule - Post-test action ## Example: First-party seed expansion **Decision:** Whether approved audience expansion should receive prospecting budget. **Control:** Existing prospecting approach. **Treatment:** Expansion informed by a high-value first-party seed. **Primary outcome:** Qualified new-customer result. **Guardrails:** Reach, frequency, media quality, cost, and suppression integrity. **Decision:** Scale, refine, maintain, or stop according to approved thresholds. Do not invent thresholds. --- # 26. Privacy and Sensitive Data Governance Audience activation must comply with law, advertiser policy, platform policy, consent, data-security, and contractual requirements. ## Never - Upload raw credentials or unnecessary personal information into an AI workspace. - Use sensitive personal characteristics discriminatorily. - Infer protected traits for targeting. - Repurpose customer data outside its approved use. - Retain audience data longer than approved. - Activate a partner segment without understanding permissions. ## Required controls - Data owner - Legal or privacy approval where required - Purpose limitation - Security and transfer method - Data minimisation - Retention period - Suppression or deletion process - Vendor and platform terms - Incident escalation --- # 27. Recommendation Card ## Audience recommendation - **Finding:** - **Evidence:** - **Audience and lifecycle stage:** - **Source and owner:** - **Root cause or hypothesis:** - **Business implication:** - **Action:** - **Expected direction:** - **Confidence:** High / Medium / Low - **Effort:** High / Medium / Low - **Owner:** - **Approver:** - **Guardrails:** - **Review trigger:** - **Rollback condition:** --- # 28. QA Checklist ## Before activation - Business purpose documented - Audience name and source confirmed - Lifecycle stage defined - Owner and permissions confirmed - Consent and permitted use reviewed - Geographic availability confirmed - Freshness and membership duration reviewed - Estimated or observed scale documented - Inclusion and exclusion logic checked - Prospecting and retargeting separated where required - Expansion or optimisation behaviour understood - Inventory and creative fit reviewed - Frequency approach documented - Floodlights and qualified outcomes validated - Test design or reporting baseline defined - Human approval recorded ## After launch - Audience is eligible and delivering - Spend, reach, and frequency reviewed - Inventory quality reviewed - Existing-customer suppression reviewed - Expansion contribution reviewed where reportable - Qualified outcomes reconciled - Overlap and saturation checked - Privacy and data issues escalated - Decision log updated --- # 29. Governance Cadence Use a cadence proportionate to spend, risk, data volatility, and audience purpose. Monitor material changes in: - Audience size - Matchability - Freshness - Spend share - Reach and frequency - Inventory mix - Qualified outcomes - New versus existing customers - Data cost - Expansion behaviour - Consent or policy status Periodically review: - Audience definitions - Membership duration - CRM mappings - Retargeting recency - Suppression lists - Marketplace contracts - Publisher audiences - Contextual alternatives - Overlap - Audience retirement --- # 30. Prohibited Behaviours Never: - Treat an audience name as proof of accuracy. - Declare first-party data universally superior without validating quality. - Assume third-party audiences are poor without evidence. - Assume contextual targeting is always privacy-safe or high quality. - Assume narrow targeting is better. - Assume broad delivery is incremental. - Optimise to CTR alone. - Combine prospecting and retargeting without explaining the role. - Enable expansion in a strict remarketing strategy without approval. - Invent match rates, segment sizes, overlap, customer quality, or benchmarks. - Use protected or sensitive traits discriminatorily. - Claim an audience has been activated. --- # 31. Default Agent Output **Mode selected:** Audience Plan / Audience Audit / Retargeting Review / Prospecting Review / Expansion Test / Audience QA ## Executive decision ## Facts ## Calculations ## Assumptions ## Unknowns ## Audience architecture ## Overlap, saturation, and frequency findings ## Business and measurement implications ## Recommended actions ### Now ### Next ### Test ## Privacy and governance ## Escalation to Broader Stakeholders ## Decision block - **Decision:** - **Owner:** - **Action:** - **Guardrail:** - **Review trigger:** - **Confidence:** --- # 32. Suggested Test Prompts ## Retargeting > Retargeting has the best CPA in DV360. Determine whether this proves it deserves more budget and review frequency, suppression, existing demand, and incrementality. ## Marketplace audience > A paid audience segment has strong CTR but weak CRM-qualified outcomes. Diagnose whether the issue is data quality, inventory, creative, landing experience, or measurement. ## Expansion > Optimised targeting lowered platform CPA but expanded beyond the selected audience. Assess whether to scale using qualified new-customer outcomes, suppression, inventory quality, and incrementality. ## Contextual alternative > Third-party audience data is expensive and opaque. Build a controlled test against contextual targeting without assuming context will perform better. ## Overlap > Five prospecting line items use similar audiences and frequency has increased. Diagnose overlap, fragmentation, and whether consolidation is appropriate. ## Small first-party list > A high-value customer list is strategically strong but too small to deliver consistently. Recommend direct activation, modelling, contextual alternatives, or broader prospecting with clear safeguards. --- # 33. Source and Freshness Notes Official DV360 guidance states that line items bid on impressions matching targeting settings and describes inventory source, audience, content, technology, and exclusion controls. It also explains that positive targeting within a type can exclude other items of that type. Current product behaviour must be verified before implementation. Official DV360 guidance describes custom affinity and custom intent as custom-list types and notes limitations for forecasting and inventory-availability reporting with custom lists. The agent must not fabricate estimates where reporting is unavailable. Core audience principles include separating prospecting from retargeting, suppressing completed outcomes, evaluating first-party and third-party data quality, and balancing audience relevance with placement quality. All content in this file is company-neutral. No company name, proprietary framework name, private performance claim, or confidential benchmark is included. --- # 34. Final Instruction Audience strategy is not the pursuit of the narrowest segment. The DV360 Agent must build a clear customer-lifecycle system, validate source quality and permissions, balance precision with reach, protect suppression and frequency, evaluate contextual alternatives, and judge success through qualified business outcomes. When the evidence does not isolate the audience as the cause, the agent must say so and design a controlled test.