# Budget Reallocation Model

> Build a Google Ads budget reallocation plan that shifts spend from low-performing to high-performing campaigns. Models multiple scenarios with projected impact before implementation. Use this skill when a user wants to optimize budget allocation, asks how to reallocate spend, mentions "budget shifts" or "reallocation", asks where to put more budget, has a fixed total budget across campaigns, or asks "where should I cut spend". Trigger on phrases like "budget reallocation", "shift budget", "cut from X put into Y", "where should my budget go", "budget optimization", or any question about distributing spend across campaigns.

- Skill: `kochellenk-afk/budget-reallocation-model` (Agent Skill)
- Install (CLI): `npx skillmds@latest add kochellenk-afk/budget-reallocation-model`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kochellenk-afk/budget-reallocation-model/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: kochellenk-afk (https://skillmd.com/u/kochellenk-afk)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kochellenk-afk/budget-reallocation-model

---


# Budget Reallocation Model

A skill for modeling budget shifts across Google Ads campaigns and projecting the impact of three reallocation scenarios.

## What this skill does

Given current campaign performance and a fixed total budget, this skill:

1. Ranks all campaigns by efficiency (conversions per dollar)
2. Identifies donor campaigns (high CPA, low ROAS, declining performance)
3. Identifies recipient campaigns (low CPA, high ROAS, IS lost to budget)
4. Models 3 reallocation scenarios - Conservative (10%), Moderate (20%), Aggressive (35%)
5. Projects total conversions, blended CPA, and blended ROAS for each scenario
6. Recommends one scenario with justification

## Required inputs

1. **Total monthly budget** for Google Ads (the constraint)
2. **Per-campaign performance data** (last 30–90 days)
3. **Campaign-level goals** - different campaigns may have different CPA targets

### Required columns

- Campaign name
- Spend
- Conversions, conv. value (if revenue tracking)
- CPA, ROAS
- Search IS, IS lost to budget
- Conv. rate
- Trend (if available - last 30 vs prior 30)

If IS lost to budget is missing, ask for it. Recipients without that signal can't be confidently identified.

## Workflow

### Step 1: Define efficiency

For each campaign, compute the efficiency score:

```
efficiency = conversions_per_dollar = conversions / spend
```

For revenue-tracking accounts:
```
efficiency = revenue_per_dollar = conv_value / spend  # i.e., ROAS
```

Rank campaigns top to bottom.

### Step 2: Identify donors

Donors are campaigns to take budget FROM. Criteria:

- Bottom quartile by efficiency, OR
- CPA > 1.5× target CPA, OR
- ROAS < 0.8× target ROAS, OR
- Declining trend: 30-day CPA up >25% vs. prior 30 days, with no fixable cause

Don't designate brand campaigns as donors (they're often "inefficient" by ROAS but defensive).

Don't designate campaigns with <30 days of data as donors (not stable enough to judge).

### Step 3: Identify recipients

Recipients are campaigns to put budget INTO. Criteria:

- Top quartile by efficiency
- IS lost to budget > 10% (signal that more budget would convert to more volume)
- CPA < target CPA (room to scale before efficiency degrades)
- Conv. rate stable (not a recently-spiking outlier)

The intersection of "high efficiency" and "IS lost to budget" is the gold zone. These are campaigns Google can convert MORE budget into MORE conversions almost linearly.

### Step 4: Model 3 scenarios

For each scenario, compute the new budget per campaign and project the resulting performance.

**Scenario A - Conservative (shift 10% of total budget)**
- Take 10% of total budget from donors (proportionally to their over-allocation)
- Distribute to recipients (proportionally to their IS lost to budget × efficiency)
- Project new conversions per campaign, blended CPA, blended ROAS

**Scenario B - Moderate (shift 20%)**
- Same logic, larger redistribution

**Scenario C - Aggressive (shift 35%)**
- Same logic, more aggressive

For projections, read `references/projection-formulas.md`. Key principle: projections must apply diminishing-returns assumptions, not naive linear scaling.

### Step 5: Risk assessment per scenario

For each scenario, surface 1–3 risks:

- Conservative: minimal disruption, may leave money on the table
- Moderate: balanced - usually the recommended choice
- Aggressive: highest upside but highest risk if recipient projections are wrong

Specific risks to call out:
- "If [recipient campaign]'s conv. rate doesn't hold at higher volume, blended CPA could rise 8%"
- "Aggressive cuts to [donor campaign] may permanently lose the audience pool - hard to recover"
- "Recipient X is brand-new; less than 90 days of data means projections have wider error bars"

### Step 6: Recommend ONE scenario

Don't be wishy-washy. Pick one and explain why. Default toward Moderate unless data clearly favors a different choice.

## Output format

A markdown response with:

1. **Current state table** - every campaign ranked by efficiency
2. **Donor / recipient classification** - which goes in which bucket and why
3. **Three scenarios** as side-by-side comparison:

| Metric | Current | Conservative | Moderate | Aggressive |
|---|---|---|---|---|
| Total spend | | | | |
| Total conversions | | | | |
| Blended CPA | | | | |
| Blended ROAS | | | | |
| # campaigns gaining budget | | | | |
| # campaigns losing budget | | | | |

4. **Per-scenario detail** - table showing each campaign's before/after budget for the recommended scenario
5. **Recommendation** - which scenario, why, when to implement
6. **Implementation order** - which campaigns to adjust first (start with smallest changes to verify before bigger shifts)

If user asks for a "deliverable" or wants to share the plan, generate an Excel workbook using the xlsx skill with one tab per scenario.

## What this skill must NOT do

- Don't apply linear scaling. Doubling a campaign's budget rarely doubles conversions - diminishing returns are real.
- Don't designate brand campaigns as donors based purely on ROAS efficiency.
- Don't project past current IS = 90% (auctions cap practical IS).
- Don't recommend cutting budget to zero on any campaign - phasing reductions matter for relearning Smart Bidding strategies.
- Don't recommend the Aggressive scenario as default; it has the highest variance.
- Don't ignore campaign-level goals. A B2B lead-gen campaign with a $200 CPA target shouldn't be compared head-to-head with an ecommerce campaign with a $20 CPA target. Score against each campaign's own target.

## Reference files

- `references/projection-formulas.md` - diminishing returns math with worked examples
- `references/donor-recipient-rules.md` - full classification logic

## Scripts

- `scripts/budget_model.py` - runs the 3-scenario model from a CSV