ad-budget-optimizer

Reallocate ad budget across campaigns and channels using ROAS and marginal return analysis

SKILL.md

GitHub

Ad Budget Optimizer

Analyzes campaign performance data (spend, revenue, ROAS, CPA, conversion rate) across campaigns and channels, identifies top performers and underperformers, calculates optimal budget allocation using marginal ROAS analysis, and recommends specific budget shifts. Outputs a reallocation plan with projected impact on overall ROAS and lead volume.

Prerequisites

  • Campaign performance data (can be pasted as CSV, JSON, or described verbally)
  • Current budget allocation across campaigns/channels
  • Business goals (target CPA, target ROAS, or lead volume target)
  • Optional: agency.config.json for client context
  • Optional: historical performance data for trend analysis

Capabilities Used

  1. google-ads-builder -- for restructuring underperforming Google campaigns
  2. meta-ads-builder -- for restructuring underperforming Meta campaigns
  3. landing-page-auditor -- for diagnosing conversion issues on high-spend, low-conversion campaigns

Phase 0: Read Config

  1. Read agency.config.json from the project root (if available).
  2. Extract budget.monthly_total for total budget context.
  3. Extract kpis.target_cpa and kpis.target_roas for benchmark goals.
  4. Check tools.websearch availability for benchmarking.
  5. Accept parameters:
    • performance_data -- (required) campaign performance data in any format (CSV, JSON, or natural language description)
    • budget_total -- (optional) total monthly budget. Default: sum from performance data
    • goal -- (optional) minimize_cpa | maximize_roas | maximize_leads | maximize_revenue. Default: maximize_roas
    • constraints -- (optional) minimum/maximum spend per campaign or channel
    • timeframe -- (optional) data time period for analysis
    • risk_tolerance -- (optional) conservative | moderate | aggressive. Default: moderate

Phase 1: Data Ingestion and Normalization

Parse Performance Data

Accept data in any format and normalize into a standard structure:

Campaign/Channel Spend Revenue Conversions CPA ROAS CTR CVR
Campaign A $X $Y N $Z X:1 X% X%

Calculate Missing Metrics

If not all metrics are provided, derive them:

  • ROAS = Revenue / Spend
  • CPA = Spend / Conversions
  • CVR = Conversions / Clicks
  • CTR = Clicks / Impressions
  • Revenue per conversion = Revenue / Conversions
  • Cost per click = Spend / Clicks

Data Quality Check

  • Flag campaigns with insufficient data (< 100 clicks or < 5 conversions) -- not statistically significant
  • Flag campaigns with no conversion tracking
  • Flag anomalies (sudden spikes/drops that may indicate tracking issues)
  • Note data freshness: when was this data collected?

Phase 2: Performance Analysis

Tier Classification

Sort campaigns into performance tiers:

Tier 1: Winners (scale these)

  • ROAS > target ROAS (or CPA < target CPA)
  • Conversion volume is meaningful (not just 1-2 lucky conversions)
  • Consistent performance over the data period
  • Still has room to scale (not at audience saturation)

Tier 2: Promising (optimize these)

  • ROAS within 20% of target
  • Good leading indicators (high CTR, decent CVR) but ROAS not yet optimal
  • Recently launched campaigns still in learning phase
  • Campaigns with fixable issues (landing page, ad copy, targeting)

Tier 3: Underperformers (fix or cut)

  • ROAS significantly below target (< 50% of goal)
  • High CPA with no improvement trend
  • Low CTR and low CVR combined
  • No clear path to profitability

Tier 4: Dead Weight (kill these)

  • Zero or near-zero conversions despite meaningful spend
  • CPA > 3x target with no improvement trend
  • ROAS < 0.5:1 for more than 30 days
  • Audiences exhausted (frequency > 3 on Meta, low impression share on Google)

Channel-Level Analysis

Aggregate by channel (Google Search, Google Display, Meta, LinkedIn, etc.):

  • Channel ROAS and CPA
  • Channel contribution to total revenue
  • Channel efficiency trend (improving or declining?)
  • Cross-channel attribution considerations

Phase 3: Marginal ROAS Calculation

Diminishing Returns Analysis

For each campaign, estimate the marginal ROAS curve:

At current spend levels:

  • What is the incremental ROAS of the last 10% of spend?
  • Is the campaign on the rising, flat, or declining portion of its returns curve?
  • Signs of saturation: increasing CPA over time, declining CTR, rising frequency (Meta), declining impression share improvement (Google)

Marginal ROAS Ranking

Rank all campaigns by estimated marginal ROAS:

  1. Campaigns with rising ROAS at current spend (scale opportunity)
  2. Campaigns with stable ROAS at current spend (maintain or modest scale)
  3. Campaigns with declining ROAS at current spend (at or past optimal spend)
  4. Campaigns with consistently negative ROAS (cut)

Optimal Allocation Model

Using marginal ROAS estimates:

  • Shift budget from low-marginal-ROAS campaigns to high-marginal-ROAS campaigns
  • Stop increasing when marginal ROAS of the scaled campaign equals the marginal ROAS of the reduced campaign
  • Apply constraints (minimum spend floors, maximum spend caps)

Phase 4: Budget Reallocation Plan

Recommended Changes

For each campaign, specify:

  • Current spend
  • Recommended spend
  • Change amount and percentage
  • Rationale for the change
  • Expected impact on ROAS/CPA
  • Risk level of the change

Reallocation Rules by Risk Tolerance

Conservative

  • Maximum 15% shift per campaign per period
  • No campaign reduced below 50% of current spend
  • Only shift to campaigns with proven ROAS > target
  • Monitor for 2 weeks before additional shifts

Moderate

  • Maximum 30% shift per campaign per period
  • Campaigns can be reduced to 25% of current spend
  • Shift to campaigns with ROAS > 80% of target
  • Monitor for 1 week before additional shifts

Aggressive

  • Maximum 50% shift per campaign per period
  • Kill underperformers entirely (reallocate 100%)
  • Shift to campaigns showing positive marginal ROAS trends
  • Monitor daily, iterate quickly

Implementation Timeline

  • Week 1: Apply budget shifts, maintain ad creative
  • Week 2: Monitor performance, adjust bids
  • Week 3: Second round of optimization based on new data
  • Week 4: Full review, set new baseline

Phase 5: Projected Impact

Before vs After Projections

Metric Current Projected Change
Total Spend $X $X (same) 0%
Total Revenue $Y $Y' +X%
Blended ROAS X:1 Y:1 +X%
Blended CPA $X $Y -X%
Total Conversions N N' +X%

Confidence Level

  • High confidence: based on > 1000 clicks and > 50 conversions per campaign
  • Medium confidence: based on > 100 clicks and > 10 conversions per campaign
  • Low confidence: limited data, projections are directional only

Scenario Modeling

Provide three scenarios:

  • Best case: all winners scale efficiently, no saturation effects
  • Base case: expected diminishing returns on scaled campaigns
  • Worst case: scaled campaigns hit saturation quickly, some regression

Phase 6: Diagnostic Recommendations

For Underperforming Campaigns (before killing)

Diagnose the root cause:

Low CTR + Low CVR -- wrong audience or wrong message

  • Recommendation: rebuild targeting and ad creative

High CTR + Low CVR -- landing page problem

  • Recommendation: run landing-page-auditor on the destination URL

Low CTR + High CVR -- message resonates with buyers but doesn't attract clicks

  • Recommendation: improve ad creative, test new hooks

High CPA + High ROAS -- working but expensive

  • Recommendation: maintain if LTV justifies, optimize for efficiency

For Cross-Channel Shifts

If recommending shifting budget between channels:

  • Use google-ads-builder to structure new Google campaigns
  • Use meta-ads-builder to structure new Meta campaigns
  • Account for platform learning periods (Meta needs 50 conversions to exit learning phase)

Phase 7: Output

Return structured JSON:

{
  "analyzed_at": "2024-01-15T14:30:00Z",
  "goal": "maximize_roas",
  "risk_tolerance": "moderate",
  "total_budget": "100000 INR",
  "current_performance": {
    "total_spend": "100000 INR",
    "total_revenue": "250000 INR",
    "blended_roas": "2.5:1",
    "blended_cpa": "1250 INR",
    "total_conversions": 80
  },
  "campaign_tiers": {
    "winners": [
      {
        "campaign": "Shopify Dev - Google Search",
        "current_spend": "30000 INR",
        "roas": "4.2:1",
        "cpa": "600 INR",
        "tier": "Winner",
        "marginal_roas": "Rising",
        "recommendation": "Scale by 30%",
        "new_spend": "39000 INR",
        "rationale": "Highest ROAS, still under target impression share, room to scale"
      }
    ],
    "promising": [],
    "underperformers": [
      {
        "campaign": "Brand Awareness - Meta Display",
        "current_spend": "20000 INR",
        "roas": "0.8:1",
        "cpa": "2500 INR",
        "tier": "Underperformer",
        "marginal_roas": "Declining",
        "recommendation": "Reduce by 50%, diagnose landing page",
        "new_spend": "10000 INR",
        "rationale": "CPA 2x target, declining ROAS trend, landing page audit needed"
      }
    ],
    "dead_weight": []
  },
  "reallocation_plan": [
    { "campaign": "Shopify Dev - Google Search", "current": "30000", "new": "39000", "change": "+9000 (+30%)" },
    { "campaign": "Brand Awareness - Meta Display", "current": "20000", "new": "10000", "change": "-10000 (-50%)" },
    { "campaign": "CRO Services - Meta Lead Gen", "current": "25000", "new": "26000", "change": "+1000 (+4%)" }
  ],
  "projected_impact": {
    "scenario_base": {
      "total_revenue": "295000 INR",
      "blended_roas": "2.95:1",
      "blended_cpa": "1050 INR",
      "total_conversions": 95,
      "improvement": "+18% revenue, +18% ROAS, -16% CPA"
    },
    "scenario_best": {
      "improvement": "+25% revenue"
    },
    "scenario_worst": {
      "improvement": "+8% revenue"
    },
    "confidence_level": "Medium"
  },
  "diagnostic_actions": [
    { "campaign": "Brand Awareness - Meta Display", "diagnosis": "High CTR + Low CVR = landing page problem", "action": "Run landing-page-auditor on destination URL" }
  ],
  "implementation_timeline": {
    "week_1": "Apply budget shifts, monitor daily",
    "week_2": "Evaluate initial impact, adjust bids",
    "week_3": "Second optimization round",
    "week_4": "Full review, set new baselines"
  }
}

Example Usage

Trigger phrases:

  • "Optimize my ad budget allocation"
  • "Reallocate budget based on ROAS"
  • "Which campaigns should I scale or cut?"
  • "Ad budget optimization for [business]"
  • "Analyze campaign performance and recommend budget shifts"
User: Here's my campaign data: [pastes CSV]. Optimize for ROAS.
Assistant: [parses data, classifies campaigns into tiers, calculates marginal ROAS, recommends specific budget shifts, projects impact]
User: I'm spending $5K/month across Google and Meta. Which campaigns should I scale?
Assistant: [analyzes performance by campaign, identifies winners and losers, recommends reallocation with projected ROI improvement]
User: My Meta ads CPA is too high. Should I shift budget to Google?
Assistant: [compares cross-channel performance, analyzes marginal returns, recommends optimal channel mix with implementation timeline]