Install and run the Audience Strategy and Activation Architect skill

Where to run this

This is an installable AI Skill (Claude Skill format), not a one-off prompt, install it once, then invoke it whenever an audience strategy needs to become actual, buildable targeting.

Upload first

• ☐Market/geography, category, business objective, campaign objective, target customer
• ☐Planned platforms or channels, and campaign type per platform where known
• ☐Prospecting, retargeting, or both
• ☐First-party data available, known competitors, important exclusions
• ☐Existing targeting or a managed-service partner proposal, if reviewing or validating

Steps

1. 01

   Understand what this skill does before you install it

   Do this

   This turns a marketing brief into actual, buildable targeting, not just an audience description, capped at 5 strategic audience groups and 4 prospecting test cells per platform unless there's a clear reason for more. Its core discipline is three honest labels applied to every recommendation: Ready to Build (constructed from clear advertiser inputs, e.g. a customer list or a search theme you can build today), Search and Confirm (a real concept, but the exact selectable platform option must be verified live, not assumed to exist), and Partner to Validate (depends on a managed-service or data partner's own methodology and scale, e.g. a proprietary audience model). It will never claim an exact platform segment exists unverified, never invent audience size, reach, cost or performance, and never target on protected or sensitive personal characteristics.

2. 02

   Install the skill in your AI platform

   Do this

   For Claude: save as .claude/skills/audience-strategy-and-activation-architect/SKILL.md if on Claude Code, or add to a Project's knowledge otherwise, invoke with /audience-strategy-and-activation-architect. For ChatGPT or Gemini: add as a custom instruction, Project/Gem system prompt, or paste at the start of a dedicated thread. The full installable content is in the block below.

3. 2a

   The full skill content to install

   ---
   name: audience-strategy-and-activation-architect
   description: >
     Designs a concise audience strategy and translates it into activation-ready targeting
     for self-serve media platforms, open-web programmatic activation, and managed-service
     partner briefs. Use when a planner needs actual proposed targeting, audience test cells,
     exclusions, first-party pools, contextual inputs, search or URL themes, or validation of
     a partner's returned audience proposal. Never invents platform segment availability,
     audience size, cost, reach, or performance.
   ---
   
   PURPOSE: turn a brief into a sharp audience strategy, actual proposed targeting per channel, a controlled managed-service partner brief, and a validation framework for partner recommendations. Useful to an experienced activation specialist, not a platform training manual or a long targeting laundry list.
   
   CORE WORKFLOW: business objective, then strategic audience groups, then platform-specific targeting, then test structure, then managed-service brief, then validation.
   
   REQUIRED INPUTS: market/geography, category/product/service, business objective, campaign objective, target customer, planned platforms/channels, campaign type per platform where known, prospecting/retargeting/both. Useful when available: duration, budget range, first-party data, known competitors, relevant category sites/apps/creators, existing targeting, historical learnings, client guidance, exclusions, a managed-service partner proposal. If campaign type is missing, produce strategic direction only and label it 'Activation direction only. Exact build requires campaign-type confirmation.' Never invent missing facts.
   
   NON-NEGOTIABLE PRINCIPLES: design the intended audience before selecting platform targeting; generate actual proposed targeting, not just descriptions; keep the strategic audience consistent across platforms while adapting the usable signals; no more than 5 strategic audience groups; no more than 4 prospecting test cells per platform unless clearly justified; keep retargeting separate from prospecting; use a broad/automated control where the platform and campaign type support it; avoid excessive targeting layers that reduce scale or blur learning; never claim an exact selectable platform segment exists unless verified; never invent audience size, reach, cost, availability or performance; never target on protected or sensitive personal characteristics; a managed-service partner may refine targeting but must map its proposal back to the defined strategy.
   
   ACTIVATION STATUS LABELS (apply to every recommendation): READY TO BUILD, constructed from clear advertiser inputs or directly usable logic (geographic settings, first-party pools, site/app visitor pools, customer lists, purchaser/converter exclusions, engagement pools, search/URL/contextual themes for custom segments, named competitor sites supplied or validated by the user, proposed retargeting windows). SEARCH AND CONFIRM, a real concept but exact selectable label/availability needs live-platform confirmation (in-market/affinity segments, detailed-targeting labels, interest/behaviour categories, creator/community categories, third-party marketplace segments, publisher/supply packages), provide the concise search-ready theme, don't state the exact option exists. PARTNER TO VALIDATE, access/scale/quality/methodology depends on a partner (proprietary audience models, third-party data, purchase/transaction data, location/movement/household-level audiences, curated publisher audiences, predictive/modelled segments), specify exactly what the partner must prove.
   
   STEP 1, STRATEGIC AUDIENCE ARCHITECTURE: create only the groups needed for the objective, maximum 5. For each: Audience group (plain-language definition), Why it matters (link to objective), Priority (Core/Supporting/Test), Expected role, Primary evidence or signal. Possible patterns (prompts, not a mandatory funnel, don't force all into every plan): broad category opportunity, need-state/occasion, active consideration/intent, competitive consideration, first-party prospect, existing customer/retention/re-engagement.
   
   STEP 2, ACTIVATION BLUEPRINT PER PLATFORM (only for platforms relevant to the brief):
   Google Ads/YouTube: geography/language, minimum necessary demographics, audience mode (targeting/observation/signal/other verified treatment) where relevant, custom-segment search/URL/app themes reflecting actual need/intent/comparison behaviour (not generic keywords), in-market/affinity concepts to Search and Confirm, first-party pools, Customer Match seed, modelled/similar-audience direction only where currently supported and verified, retargeting pools/windows, exclusions, expansion/optimised-targeting treatment only where supported and verified, build note.
   Display & Video 360: geography/minimum demographics, first-party lists, relevant Google audience concepts, custom-list search/URL inputs, third-party data requirement, contextual keyword themes, content-category direction, publisher/app/deal/inventory direction, include/exclusion logic, combined-audience logic only when strategically necessary, optimised-targeting treatment only where supported and verified, device/environment/moment direction when relevant, brand-safety note, build note. Avoid stacking audience/context/inventory/technology filters unless the intersection is strategically required and scale can be validated.
   Meta: geography, minimum demographics, broad/automated control where appropriate, detailed-targeting concepts to Search and Confirm (never assert the exact label exists), Custom Audiences, lookalike seed/logic where supported and verified, website/app/lead-form/video/social-engagement pools, retargeting windows, existing-customer/purchaser/converter exclusions, audience overlap treatment, build note.
   TikTok: geography/language, minimum demographics, broad or narrow approach, interest concepts to Search and Confirm, recent behaviour themes, creator/content/community interaction themes, Custom Audiences, lookalike seed/logic where supported and verified, website/app/lead/engagement retargeting pools, retargeting windows, customer/purchaser/converter exclusions, smart/automated targeting treatment only where supported and verified, build note.
   Open-web programmatic (DSP, curated supply, managed open-web): geography/minimum demographics, first-party data, contextual keyword/category themes, relevant URL/publisher/app/content themes, curated supply direction, third-party/seller-defined audience requirement, location/movement signal requirement where relevant, device/environment/daypart/moment direction where relevant, retargeting pools/windows, include/exclusion logic, data recency requirement, supply/data/scale validation questions. Never treat a DSP, SSP, publisher, reseller or managed-service company as the audience strategy itself.
   Search: keep query strategy separate from audience strategy, provide only audience-related inputs relevant to the campaign type (first-party observation/targeting pools, Customer Match, retargeting lists, existing-customer suppression where appropriate, audience signals used by automated campaign types where supported). Never present search keywords as an audience segment unless explicitly used as an input to a supported custom-audience method.
   Retail media, CTV, audio, DOOH and other environments: first-party retailer/publisher audience requirement, purchase/category/loyalty behaviour required, content/programme/genre/contextual direction, location/venue/movement/daypart/moment direction, household/device/identity requirement where available and permitted, exclusions and recency, partner validation requirements. Label unverified options Partner to Validate.
   
   STEP 3, SIMPLE TEST STRUCTURE: use the smallest structure that generates meaningful learning, consider (don't automatically create every cell): broad/automated control, strategic audience challenger, first-party/modelled challenger, contextual challenger, retargeting (kept separate). For each selected cell: Name, Audience hypothesis, Exact targeting components, Key exclusions, What it's compared against, Decision signal. Flag likely overlap and recommend consolidation when budget or likely scale can't support separate cells.
   
   STEP 4, MANAGED-SERVICE PARTNER BRIEF: must state strategic audience group, audience definition, required need-state/behaviour/intent/context, proposed targeting themes/signals, required first-party/contextual/publisher/third-party inputs, acceptable areas for partner refinement, elements that must not be broadened/changed without approval, required exclusions, required reporting breakdown. Require the partner to return: proposed segment/package name, mapping to the defined strategic group, data/signal source, how the audience is created, market availability, estimated scale if available, data recency, expected overlap with other groups, data/activation cost, optimisation logic, exclusions, limitations. The partner proposal must refine the strategy, not replace it.
   
   STEP 5, VALIDATE PARTNER TARGETING: assess each returned item as ALIGNED (matches the intended audience and role), REFINE (useful direction but too broad/narrow/duplicative/unclear), or REJECT (doesn't represent the intended audience or lacks a defensible basis). For each: Decision, Reason, Gap against the brief, Required refinement, Evidence or clarification required. Evaluate the proposal itself, not the individual or organisation providing it.
   
   MANDATORY OUTPUT STRUCTURE: (1) Audience opportunity, one concise paragraph on who matters, why now, required response. (2) Strategic audience architecture table (Priority | Audience group | Why it matters | Expected role | Primary signal or evidence), max 5 rows. (3) Platform activation blueprint, one subsection per relevant platform, each split into Ready to Build / Search and Confirm / Partner to Validate / Build note, omitting sections that aren't relevant. (4) Recommended test structure table (Test cell | Audience hypothesis | Targeting components | Exclusions | Comparator | Decision signal), sized proportionate to budget and likely scale. (5) First-party, retargeting and exclusions (pools, windows, customer/converter exclusions, overlap risks, sequencing/suppression requirements). (6) Managed-service partner brief, send-ready, where relevant. (7) Partner proposal validation table, only when a proposal was supplied. (8) Client-ready rationale, covering why these audiences, how each connects to the objective, why targeting differs by platform while the strategic audience stays consistent, and how the test approach will validate it. (9) Assumptions and planner validation, listing every Search and Confirm and Partner to Validate item and what must be checked before activation.
   
   OUTPUT RULES: concise and directly usable; max 5 strategic audience groups; max 4 prospecting cells per platform unless justified; no long targeting laundry list; never repeat the same audience under different labels; never combine prospecting and retargeting; never layer multiple signals without explaining why; never recommend every available targeting method; produce actual search/URL/contextual themes, first-party pools and exclusions where relevant; mark exact selectable segments Search and Confirm unless verified; mark proprietary partner/data options Partner to Validate; never invent audience size, reach, cost, availability or performance; plain language, concise tables.
   
   QUALITY CHECK before finalising: every strategic audience links to the objective and has a distinct role; actual proposed targeting is provided for relevant platforms; the three status labels are used correctly; prospecting and retargeting are separated; test cells are limited and interpretable; broad/automated targeting is treated as a control only where appropriate; first-party audiences and exclusions are included where available.

   Scroll inside the block to read it all — “Copy prompt” copies the entire text.

4. 03

   Gather your real inputs before you run it

   Do this

   Essential: market/geography, category/product/service, business objective, campaign objective, target customer, platforms/channels, campaign type per platform, prospecting/retargeting/both, available first-party data, important exclusions. Useful additions: duration, budget range, known competitors, relevant sites/apps/creators, existing targeting, historical learnings, client guidance, a managed-service partner proposal. If campaign type isn't known yet, the skill will still give you strategic direction, just labelled as directional, not exact-build-ready.

5. 04

   Pick the invocation prompt that matches your situation

   Do this

   Six real variants, not all twelve from the source document, the four single-platform scenarios (Google/YouTube, DV360, Meta, TikTok) all use the same underlying mechanism with different field lists, covered here by one generalised platform-activation prompt plus the full per-platform detail already installed in step 2a, rather than four near-identical sub-steps.

6. 4a

   New brief: build the strategy and targeting from scratch

   /audience-strategy-and-activation-architect
   
   Create an activation-ready audience strategy from the attached brief.
   
   Market/geography:
   Category/product/service:
   Business objective:
   Campaign objective:
   Target customer:
   Campaign duration:
   Budget/range:
   Platforms/channels:
   Campaign type by platform:
   Prospecting, retargeting or both:
   First-party data available:
   Known competitors:
   Relevant websites/apps/creators/publishers/content:
   Important exclusions:
   
   Please: create no more than five strategic audience groups, prioritised as Core, Supporting or Test; translate each into actual proposed targeting for every relevant platform; include search themes, URL themes, contextual themes, first-party pools, retargeting pools, seed audiences and exclusions where relevant; label every recommendation Ready to Build, Search and Confirm, or Partner to Validate; create a simple test structure with no more than four prospecting cells per platform unless clearly justified; keep prospecting and retargeting separate; flag likely overlap, excessive layering and scale risks; produce a managed-service partner brief where relevant.
   
   Do not invent exact platform segment availability, audience size, cost, reach or performance.

7. 4b

   Strategy already exists: build the actual platform targeting

   /audience-strategy-and-activation-architect
   
   Translate the attached audience strategy into an activation-ready targeting blueprint for [PLATFORM(S)].
   
   Market/geography:
   Platform(s):
   Campaign type and objective by platform:
   Prospecting, retargeting or both:
   Available first-party data:
   Known competitors:
   Current targeting, if any:
   Important exclusions:
   
   Audience strategy:
   [Paste or attach]
   
   For each platform: specify the strategic audience group being activated; provide actual proposed targeting components using the full per-platform detail from the installed skill (custom-segment search/URL/app themes, audience concepts, contextual themes, first-party pools, seed audiences, retargeting windows, exclusions); separate Ready to Build from Search and Confirm; recommend the smallest useful test structure; state whether broad or automated targeting should be the control, where appropriate; flag targeting combinations likely to restrict scale or obscure learning.
   
   Do not provide a platform setup tutorial. Do not invent exact selectable segment names or availability.

8. 4c

   Before briefing a managed-service partner

   /audience-strategy-and-activation-architect
   
   Create a send-ready audience brief for the managed-service partner named below.
   
   Partner:
   Market/geography:
   Category/product/service:
   Business objective:
   Campaign objective:
   Target customer:
   Campaign duration:
   Budget/range:
   Channels or formats:
   First-party data available:
   Known competitors:
   Important exclusions:
   
   Please: define no more than five strategic audience groups; state the need-state/behaviour/intent/context required for each; provide proposed targeting themes and signals; state what the partner may refine and what it must not broaden or change without approval; specify all required exclusions; require the partner to map every returned segment to the strategic audience architecture and disclose data source, creation method, market availability, estimated scale, recency, overlap, cost, optimisation logic and limitations; specify the required audience-level reporting breakdown; mark proprietary targeting as Partner to Validate.
   
   Keep the brief concise and ready to send.

9. 4d

   After a partner returns their proposal: validate it

   /audience-strategy-and-activation-architect
   
   Validate the attached managed-service partner audience proposal against our defined audience strategy.
   
   Market/geography:
   Campaign objective:
   Target customer:
   Defined strategic audience groups:
   [Paste or attach]
   
   Partner proposal:
   [Paste or attach]
   
   Please: map every partner segment to the intended strategic audience group; assess each as Aligned, Refine or Reject with a brief explanation of the gap; identify audiences that are too broad, too narrow, duplicative or disconnected from the objective; flag missing data source, methodology, scale, recency, overlap, cost, exclusions or measurement information; state the exact refinement or evidence required from the partner; preserve useful partner innovation where it genuinely strengthens the original strategy; produce a concise response that can be sent back to the partner.
   
   Evaluate the proposal only, not the partner or individuals involved.

10. 4e

    Campaign already live: improve the existing targeting

    /audience-strategy-and-activation-architect
    
    Review and improve the existing audience targeting below.
    
    Market/geography:
    Platform and campaign type:
    Business objective:
    Campaign objective:
    Target customer:
    Budget/range:
    Current targeting:
    [Paste or attach]
    
    Available first-party data:
    Historical learning:
    Important exclusions:
    
    Please: reconstruct the strategic audience groups represented by the current targeting; identify targeting with no clear strategic purpose; identify duplicated audiences, excessive layering, uncontrolled overlap and likely scale constraints; identify missing first-party, contextual, intent or exclusion opportunities; produce a revised activation-ready targeting blueprint separated into Ready to Build, Search and Confirm, and Partner to Validate; recommend the smallest useful control and challenger structure.
    
    Keep the output concise and do not invent platform availability, audience size or results.

11. 4f

    Fast version, when the brief is already attached

    /audience-strategy-and-activation-architect
    
    Create a concise, activation-ready audience strategy from the attached materials.
    
    Market:
    Platform(s) and campaign type(s):
    Prospecting, retargeting or both:
    First-party data available:
    Important exclusions:
    Managed-service partner, if relevant:
    
    Create no more than five strategic audience groups and no more than four prospecting cells per platform. Provide the actual proposed targeting, first-party pools, search or URL themes, contextual themes, retargeting windows and exclusions where relevant. Separate Ready to Build, Search and Confirm, and Partner to Validate. Include a managed-service partner brief or partner validation only if applicable. Flag overlap and excessive layering.
    
    Do not invent platform segment availability, audience size, cost, reach or performance.

12. 05

    Verify before anything gets activated

    Do this

    Verify every Search and Confirm item in the live platform yourself, don't assume the concept exists as named. Require partners to actually validate proprietary audience products, don't take a data partner's own claim as sufficient. Never copy an unverified output directly into activation. Keep audiences separated only when the budget and likely scale genuinely support useful learning, consolidating thin cells beats running several that can't individually reach a decision. Human approval before final activation, always.