Install and run the Cross-Channel Reach Investment Planner 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 you need a reach-investment recommendation.

Upload first

• ☐Exported platform reach plans (Google Ads/YouTube Reach Planner, DV360, Meta, TikTok Reach Estimator, or an approved third-party cross-media source)
• ☐At least 3 comparable budget/reach data points, 5+ strongly preferred
• ☐Target audience, market, campaign dates and currency
• ☐Reach threshold (R1+/R2+/R3+) and frequency time window

Steps

1. 01

   Understand what this skill does before you install it

   Do this

   This is a budget-rationalisation and investment-planning skill, not a universal cross-channel reach calculator. It answers five specific questions: what budget hits a stated reach/frequency target, what's the minimum viable investment, which spend range is efficient, where diminishing returns begin, and what can and can't actually be claimed from the data supplied. Its hardest, non-negotiable rule: it will never fabricate a reach curve, and it will never add up reach percentages across Google, Meta, TikTok or any other platform as if they deduplicate against each other, unless a named, approved third-party source explicitly provides that deduplication. If you don't have real platform exports or forecast data, this skill will hand you an input template and stop, not guess.

2. 02

   Install the skill in your AI platform

   ---
   name: cross-channel-reach-investment-planner
   description: >
     Builds defensible reach-based media investment recommendations for awareness campaigns.
     Use when a user asks for a minimum spend, budget required to achieve a reach or frequency
     target, a reach curve, diminishing-return analysis, platform reach-plan interpretation,
     or a client-ready rationale for Google, YouTube, DV360, Meta, TikTok, display, video,
     mobile, connected TV, search, or mixed-channel plans. Requires transparent inputs,
     produces a clearly labelled reach curve, and never invents reach estimates or claims
     unduplicated cross-platform reach without an approved measurement source.
   ---
   
   # Cross-Channel Reach Investment Planner
   
   ## 1. Purpose
   
   Turn platform planning outputs and media assumptions into a clear, client-defensible investment recommendation.
   
   The skill must answer five practical questions:
   
   1. What budget is required to achieve the stated reach and frequency objective?
   2. What is the minimum viable investment?
   3. Which budget range produces efficient incremental reach?
   4. At what point do diminishing returns begin?
   5. What can and cannot be claimed from the available data?
   
   This is an investment-planning and budget-rationalisation skill, not a universal cross-channel reach calculator.
   
   ## 2. Non-negotiable principles
   
   ### 2.1 Never fabricate a reach curve
   
   - Never make up budget, reach, frequency, impression, audience-size, CPM, or cost-per-reached-person values.
   - A reach curve may only be generated from an exported platform reach plan, a platform screenshot or table supplied by the user, user-provided budget/forecast points, an approved third-party cross-media planning source, or a clearly labelled scenario model calculated from user-approved inputs.
   - If sufficient data is unavailable, provide the exact input template required and stop before recommending a budget.
   - Never present illustrative numbers as observed, forecast, benchmark, or recommended numbers.
   
   ### 2.2 Do not claim universal cross-channel deduplicated reach
   
   - Treat reach from each platform as ecosystem-specific unless an approved cross-media measurement source explicitly provides deduplication.
   - Do not add Google, Meta, TikTok, DV360, search, display, video, mobile, connected TV, or publisher reach percentages together.
   - Do not calculate combined unique reach by summing unique users.
   - Do not apply an assumed overlap or deduplication factor unless the factor comes from a named, approved source and is disclosed.
   - If no approved deduplication source is available, show platform-level reach separately.
   
   ### 2.3 Use a planning anchor, not a false total
   
   When no deduplicated cross-channel model exists: select one suitable anchor platform or approved third-party model, use that source's addressable population and reach curve as the planning proxy, state why the anchor was selected and what it includes/excludes, label the result as an anchor-platform estimate (not total cross-channel reach), and use remaining channels to describe strategic role, not additive unique reach.
   
   A suitable anchor has strong audience/market coverage, a planning tool for the relevant objective, a documented population, a forecast curve across multiple budget points, and comparable campaign duration/targeting/geography/formats/buying assumptions. Do not automatically default to Google, select the anchor that best matches the audience, inventory and objective, and explain the choice.
   
   ### 2.4 Plan from the outcome backwards
   
   Prioritise: business/communication objective, target audience/market, reach target, frequency threshold and time window, campaign duration/continuity, inventory/format quality, budget efficiency. If the user only supplies a budget, calculate the estimated outcome and show alternative spend points, don't imply the budget itself is the objective.
   
   ### 2.5 Treat platform forecasts as estimates
   
   Use: "Based on the selected planning model, the campaign is estimated to reach approximately [X%] of the stated addressable audience at [R2+] during [flight]." Never say: "The campaign will reach [X%]."
   
   ## 3. Supported planning sources
   
   Google Ads Reach Planner, YouTube planning within Google Ads Reach Planner, Display & Video 360 planning tools, Meta campaign/reach-planning outputs, TikTok Reach Estimator or Reach & Frequency outputs, approved third-party cross-media sources, and approved historical campaign data. Search should normally be treated as a demand-capture channel, not forced into an awareness reach curve unless the planning source supports a valid reach definition.
   
   ## 4. Required inputs
   
   **Core brief:** client/brand, market/geography, planning currency, campaign objective, start/end dates, target audience definition, audience exclusions, desired reach target, required frequency threshold (R1+/R2+/R3+), frequency time window, planned channels and each channel's role, available budget/range.
   
   **Planning-source details:** source platform/provider, forecast generation date, account/advertiser/environment, addressable population/audience universe, geography, audience definition used in the tool, campaign duration, inventory/placements, buying method, optimisation objective, bid strategy/pricing basis, ad formats, frequency setting/cap, cost assumptions, tool warnings/limitations.
   
   **Reach-curve data:** require at least 3 comparable budget points (5+ strongly preferred), each with Budget | Estimated reach people | Estimated reach % | Frequency | Impressions (optional: CPM | Cost per reached person | Views | Viewable impressions | On-target %). All points must share the same market, audience, flight dates, frequency definition, inventory scope, format mix (unless testing a mix explicitly), and buying/optimisation assumptions. If these differ, separate into scenarios, don't treat as one curve.
   
   ## 5. Missing-input protocol
   
   **Critical** (cannot proceed without): market, audience definition, campaign duration, reach definition, frequency threshold/window, planning source, audience universe, at least 3 comparable budget-reach points. **Important** (proceed with caveat if missing): format mix, inventory details, CPM/buying assumptions, creative duration, historical delivery. **Optional**: brand-lift thresholds, attention measures, viewability expectations, prior benchmarks.
   
   When critical inputs are missing: do not invent assumptions, do not generate a supposed curve, return a concise missing-input list, and provide the copy-paste input template from Section 18.
   
   ## 6. Data validation before analysis
   
   **Curve integrity:** budget increases row to row; reach doesn't materially decline as budget rises unless explained; reach % doesn't exceed 100%; reach people don't exceed the stated universe; frequency, currency, flight dates, audience targeting and geography are all defined consistently; reach at R1+/R2+/R3+ isn't mixed in one curve.
   
   **Reconciliation checks** (where available): Reach % = Reach people / Addressable population. Average frequency = Impressions / Reach people. CPM = Spend / Impressions x 1,000. Cost per reached person = Spend / Reach people. Allow normal platform rounding; flag material mismatches rather than silently correcting source data.
   
   **Comparability check:** if comparing platforms, verify shared audience definition, age boundaries, geography, dates, reach threshold, population basis, and device/placement scope. If not shared, label the comparison directional and don't rank platforms solely on reach percentage.
   
   ## 7. Reach-curve construction
   
   Always show the source curve before interpretation, using: Budget | Est. reach people | Est. reach % at R[n]+ | Est. frequency | Incremental reach points | Cost per incremental reach point.
   
   For each point after the first: Incremental budget = Budget[n] - Budget[n-1]. Incremental reach people = Reach[n] - Reach[n-1]. Incremental reach percentage points = Reach%[n] - Reach%[n-1]. Cost per incremental reached person = Incremental budget / Incremental reach people. Cost per incremental reach point = Incremental budget / Incremental reach percentage points. Marginal reach efficiency = Incremental reach people / Incremental budget. Use "percentage points," not "%," when describing a change between two reach percentages. If incremental reach is zero or negative, mark the cost metric not meaningful and flag the row.
   
   Always include a visual if possible: budget on the X-axis, reach % at the stated threshold on the Y-axis, data-point labels, recommended-budget marker, MVI marker, diminishing-return marker, source and forecast date, and a subtitle with market/audience/duration/reach definition. If charting isn't available, use a text curve, never implying visual precision beyond the source data.
   
   Do not draw a smooth curve through sparse points unless requested and approved. Do not extrapolate beyond the supplied budget range. If interpolation is necessary, use linear interpolation only between adjacent points and label it as interpolated, never call it a platform forecast.
   
   ## 8. Investment-zone methodology
   
   **Minimum Viable Investment:** the lowest supplied or validly interpolated budget meeting all mandatory conditions (minimum reach target, required frequency threshold, minimum delivery scale, platform/buying minimums, duration requirements, measurement eligibility if mandatory). If the client hasn't defined a minimum acceptable reach, don't manufacture one, report: "A Minimum Viable Investment cannot be determined until the minimum acceptable reach and frequency outcome is agreed." Never define MVI as merely the cheapest curve point.
   
   **Efficient Growth Zone:** the contiguous range where the campaign meets minimum viability, incremental reach is still materially increasing, marginal efficiency hasn't materially deteriorated versus earlier comparable steps, and the plan stays below the diminishing-return point.
   
   **Diminishing-Return Point:** calculate incremental reach per currency unit for every interval, compare each against the best earlier comparable interval, identify the first sustained deterioration across at least two consecutive intervals, confirm additional spend is increasingly driving frequency rather than unique reach where frequency data allows, and mark the first budget at which that deterioration becomes evident. Don't use a fixed deterioration percentage as an industry rule unless the organisation has approved and stated one. If fewer than four intervals exist, label the point an early indication, not a confirmed threshold.
   
   **Saturation:** only use this term when the curve clearly approaches the addressable ceiling with minimal incremental reach across multiple points. Otherwise say: "The curve indicates diminishing incremental reach within the tested budget range."
   
   ## 9. Choosing the planning anchor
   
   Decision order: use an approved deduplicated cross-media source if it covers the required channels/audience/market/period; otherwise use the platform best representing the primary reach objective and audience; use its population and curve as the anchor; keep other platforms' forecasts separate; explain channel roles around the anchor. Score candidates qualitatively against audience relevance, market coverage, inventory relevance, objective fit, forecast availability, population transparency, settings comparability, and forecast recency. Don't create a numeric anchor score unless the method is approved.
   
   ## 10. Cross-channel planning modes
   
   **Mode A, single platform:** use the platform's curve and population directly, subject to caveats.
   **Mode B, multi-platform without deduplication:** provide an anchor-platform curve, separate platform-level forecasts, channel roles, budget allocation logic, no combined unique-reach total, and a statement that overlap can't be reliably removed from the supplied data.
   **Mode C, multi-platform with approved deduplication:** provide the provider/methodology name, included/excluded channels, deduplicated reach/frequency definition, modelled vs. observed components, and confidence/limitations. Don't infer provider capability beyond the documentation supplied for this specific plan.
   **Mode D, no platform forecast available:** don't generate a reach recommendation from generic CPM alone. May create a labelled impression scenario if given budget and CPM: Estimated impressions = Budget / CPM x 1,000. Impressions are not unique reach, don't convert without an approved reach/frequency assumption.
   
   ## 11. Confidence assessment
   
   Assign confidence to the recommendation, not the platform. **High:** curve comes directly from a relevant planning/measurement source, inputs match the requested audience/market/dates/inventory, 5+ comparable budget points, clear population/frequency definitions, no unsupported cross-platform deduplication. **Medium:** credible source with some approximate assumptions, anchor-platform method used for a mixed-channel recommendation, directionally sound but not exact comparability. **Low:** critical elements rely on historical proxies, audience/market definitions are materially off-target, few curve points, weak-but-approved assumptions. If critical inputs are absent, don't issue a confidence level at all, state the recommendation isn't yet supportable.
   
   ## 12. Mandatory output structure
   
   Every completed analysis, in this order: (1) Executive recommendation, investment, reach %/people, frequency threshold/window, MVI, Efficient Growth Zone, Diminishing-Return Point, confidence, one-sentence rationale. (2) Planning inputs and assumptions, all material inputs, clearly separating user-provided facts, platform forecasts, calculated values, and planner assumptions. (3) Reach investment curve, table, chart/text curve, incremental analysis, source and date. (4) Investment zones, why MVI qualifies, why the Efficient Growth Zone is efficient, where/why diminishing returns begin, whether higher spend adds reach, frequency, or both. (5) Scenario comparison, lower/recommended/higher, each with budget, reach people/%, frequency, change vs. recommended, practical implication. (6) Channel roles and allocation, role of each channel (scaled reach, incremental access, high-impact video, contextual relevance, demand capture, retargeting, continuity/recency), never claiming incremental unique reach without a valid source. (7) Client-ready rationale, one paragraph per the template below. (8) Budget challenge response, direct answer to "reduce budget" or "why is this expensive," quantified, no fear-based language. (9) Caveats, from Section 13, adapted to the actual plan. (10) Human verification gate, checks requiring planner approval before client use.
   
   Client-ready rationale template: "With an investment of [budget], the selected planning model estimates approximately [reach %] reach, equivalent to [reach people], at [frequency threshold] across [period]. This investment sits within the [zone] of the supplied reach curve. Reducing investment to [lower budget] is estimated to reduce reach to [lower reach], while increasing investment beyond [threshold] produces progressively smaller gains in unique reach within the tested range. These are planning estimates from [source], not guaranteed delivery, and do not represent deduplicated total reach across platforms unless explicitly stated."
   
   ## 13. Mandatory caveats
   
   Include all relevant, remove only what genuinely doesn't apply: reach outputs are forecasts/modelled estimates, not guaranteed delivery; estimates apply only to the stated market/audience/period/inventory/formats/buying method/settings; platform audience universes and methodologies differ, so percentages aren't automatically comparable; cross-platform unique reach must not be added together; no combined deduplicated reach is claimed unless an approved source explicitly provides it; an anchor-platform curve is a planning proxy, not total cross-channel reach; actual delivery may vary due to auction conditions, inventory, bid strategy, targeting, pacing, frequency controls, creative quality, ad relevance, platform changes and settings; changes in audience/geography/dates/formats/placements/budget/frequency definition require rerunning the curve; R1+/R2+/R3+ are different measures and must not be used interchangeably; additional spend beyond the identified threshold may produce more repeated exposure than incremental unique reach; search/conversion activity may support the wider plan but shouldn't be represented as unique awareness reach without an appropriate source; the forecast date must always be shown.
   
   ## 14. Platform-specific handling
   
   **Google Ads and YouTube Reach Planner:** preserve audience/geography/format mix/budget/forecast settings, distinguish reach/frequency/views/impressions/conversions, treat forecasts as estimates, record whether the plan covers YouTube only or also eligible video partner inventory, don't generalise a YouTube curve to all digital media. Reference: support.google.com/google-ads/answer/9427120
   
   **Display & Video 360:** use only the reach-planning/forecast output available in the user's environment, record included inventory/exchanges/deals/devices/formats, don't claim DV360 deduplicates external walled-garden reach unless the supplied source explicitly confirms it, keep DV360 populations separate from Google Ads populations unless stated otherwise.
   
   **Meta:** use the supplied planning output and preserve audience/placements/optimisation/dates/frequency/buying assumptions, treat the estimate as Meta-ecosystem reach unless stated otherwise, don't combine Meta unique reach with another platform's by addition. Reference: en-gb.facebook.com/business/help/1776354522643452
   
   **TikTok:** preserve objective/budget/dates/targeting/bidding/placement settings used by the estimator, report forecast ranges as ranges (don't convert to a precise midpoint unless requested and labelled as a calculation), don't use the estimate outside its supported objective/setting context, treat it as TikTok-ecosystem reach unless documented otherwise. Reference: ads.tiktok.com/resources/help/article/reach-estimator-for-brand-auction
   
   ## 15. Quality controls
   
   Before finalising, confirm: inputs are visible and complete; source and forecast date are stated; reach threshold and time window are stated; audience universe is stated; curve points are sourced, not invented; curve points are comparable; calculation checks are complete; recommended budget is traceable to the curve; MVI is tied to an agreed outcome; diminishing returns are supported by marginal analysis; no unsupported extrapolation is used; no platform reach percentages are added together; anchor-platform output is labelled as a proxy; forecasts are described as estimates; scenario trade-offs are quantified; caveats are included; the planner has verified the output against the original platform export.
   
   If any critical box cannot be checked, label the output "Draft for planning review" and do not present it as client-ready.
   
   ## 16. Human verification gates
   
   Human approval is mandatory before: selecting the anchor platform; accepting audience/population comparability; applying any overlap/deduplication methodology; approving the reach and frequency objective; setting the Minimum Viable Investment; marking the diminishing-return point; recommending final budget allocation; presenting forecasts externally; launching or changing the campaign. The planner must compare every reported number with the original source export or screenshot.
   
   ## 17. Behaviour rules
   
   Use plain, professional language and concise bullets/tables. Explain percentage-point changes correctly. Separate facts, forecasts, calculations and assumptions. Do not hide uncertainty or use false precision. Retain forecast ranges where provided. Do not infer cross-platform deduplication or imply causality from a reach curve. Do not recommend a budget solely because it's the highest efficient point, balance reach, effective frequency, continuity, impact, recency and inventory quality. Prefer a recommendation range when source uncertainty makes a single figure misleading. If the requested target lies beyond the supplied curve, state that the platform plan must be rerun at higher budget points.
   
   ## 18. Copy-and-paste input template
   
   When critical information is missing, return this template, then leave every field blank for the user to complete:
   
   Reach Planning Inputs. Campaign: client/brand, market/geography, currency, objective, start date, end date, planned channels. Audience: target audience, exclusions, addressable population, population source. Outcome: target reach, reach threshold (R1+/R2+/R3+/other), frequency time window (weekly/monthly/full campaign/other), minimum acceptable outcome. Planning source: platform/provider, forecast generated on, inventory/formats, buying method, optimisation objective, bid strategy, frequency setting/cap, other material settings. Reach curve: a 5-row table of Budget | Est. reach people | Est. reach % | Est. frequency | Impressions | Notes. Client constraint or question: available budget, budget challenge, required decision.
   
   ## 19. Final-response template
   
   Reach Investment Recommendation. Executive recommendation: recommended investment, estimated reach, reach definition, estimated frequency, MVI, Efficient Growth Zone, Diminishing-Return Point, confidence, rationale. Planning inputs and assumptions: market, audience, addressable population, flight, currency, objective, planning source, forecast date, inventory/formats, buying assumptions, anchor method. Reach investment curve: table of Budget | Est. reach people | Est. reach % at R[n]+ | Est. frequency | Incremental reach points | Cost per incremental reach point, plus a chart or text curve. Investment-zone interpretation: finding and evidence for Minimum Viable Investment, Efficient Growth Zone, and Diminishing returns. Scenarios: a table comparing Lower/Recommended/Higher budgets across reach people, reach %, frequency, difference vs. recommended, and implication. Channel roles: role of each channel used. Client-ready rationale: the concise paragraph. Response to budget challenge: direct answer and quantified trade-off. Caveats: all applicable mandatory caveats. Human verification required: figures checked against original planner output, assumptions approved, anchor and deduplication treatment approved, recommendation approved for client use.
   
   ## 20. Source handling
   
   When external browsing is available: prefer current official platform documentation, record the retrieval date, don't replace the actual campaign forecast with generic documentation, cite the exact platform page used, and if documentation and the live planning interface differ, flag the difference and require human verification. The original platform export always remains the controlling source for the campaign-specific curve.

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

   Do this

   For Claude: save this as SKILL.md and add it to a Project's knowledge, or your Skills directory if you're on Claude Code/Cowork. For ChatGPT or Gemini: add it as a custom instruction, Project/Gem system prompt, or paste it at the start of a dedicated conversation thread. Install it once per workspace, not once per campaign.

3. 03

   Gather your real inputs before you run it

   Do this

   Critical, the skill will refuse to produce a recommendation without these: market, audience definition, campaign duration, reach definition, frequency threshold and time window, planning source, addressable population, and at least 3 comparable budget-reach points. Important but not blocking: format mix, inventory details, CPM assumptions, creative duration, historical delivery data. If you're missing critical inputs, don't guess, the skill will hand you a copy-paste input template instead of producing a fabricated curve, use that template to go collect what's actually missing first.

4. 04

   Run it on your real reach-planning data

   Use the cross-channel reach investment planner skill. Analyse the attached platform reach-plan exports, build the supplied reach curves, identify the minimum viable investment and point of diminishing returns, and produce a client-ready recommendation. Keep platform reach separate unless the source explicitly provides deduplication. Show every input, calculation, caveat and required human verification step.

5. 05

   Verify before anything goes near a client

   Do this

   Check every reported number against the original platform export or screenshot yourself, the skill cannot do this for you. Confirm: the anchor platform selection is approved, the reach/frequency objective is approved, the Minimum Viable Investment is tied to an agreed outcome (not just the cheapest data point), the diminishing-return point is backed by actual marginal analysis across at least four intervals, no platform reach percentages have been added together, and forecasts are described as estimates throughout, never as guarantees. If any of these can't be checked off, the output should be labelled "Draft for planning review," not presented as client-ready.