Wasted Spend Finder

Identify and categorize wasted ad spend across Google Ads and Meta platforms. Produces uploadable exclusion lists with thematic categorization.

Install

git clone https://github.com/thatrebeccarae/claude-marketing.git && cp -r claude-marketing/skills/wasted-spend-finder ~/.claude/skills/

Core Capabilities

Google Ads Waste Detection

1. Search Term Waste — zero-conversion terms above spend threshold
2. Placement Waste — Display Network sites/apps and YouTube channels with no conversions
3. Keyword-Level Waste — low conversion rate keywords relative to spend

Meta Ads Waste Detection

1. Placement Waste — Audience Network apps, low-quality inventory
2. Audience Segment Waste — lookalike/interest targeting with poor performance
3. Creative Waste — high frequency + zero conversions, creative fatigue

Waste Thresholds

Spend Threshold Formula

Waste threshold = Average CPA x Multiplier

Multiplier by campaign type:
- Search (branded): 2x CPA
- Search (non-brand): 3x CPA
- Display: 3x CPA
- Meta (all): 2.5x CPA

A search term, placement, or audience segment is flagged as waste when:

• Spend exceeds the threshold AND
• Conversions = 0 AND
• Minimum click threshold is met

Minimum Click Thresholds

Campaign Type	Min Clicks Before Flagging
Search (branded)	50
Search (non-brand)	25
Display / GDN	15
YouTube	20
Meta Feed	20
Meta Stories	15
Meta Audience Network	10

Conversion Rate Minimums

Flag keywords/placements performing below these floors:

Campaign Type	Min Conversion Rate
Search (branded)	8%
Search (non-brand)	1.5%
Display	0.3%
YouTube	0.2%
Meta Feed	0.5%
Meta Stories	0.3%

Negative Keyword Categorization

Classify wasted search terms into themes for organized exclusion:

Theme	Pattern	Match Type
Competitor terms	Brand names, product names	Exact / Phrase
Informational intent	how to, what is, tutorial, guide, definition	Phrase
Wrong product/service	Related but non-target products	Phrase / Exact
Geographic irrelevance	Wrong cities, states, countries	Exact
Job/career searches	jobs, salary, hiring, career, resume	Phrase
Free/cheap intent	free, cheap, discount, coupon, deal	Phrase
Irrelevant modifiers	DIY, homemade, kids, old, used	Phrase

Placement Exclusion Categories

Google Display Network

• Low-quality apps (game apps, utility apps, kids apps)
• Parked domains and MFA (made-for-advertising) sites
• Irrelevant content categories

YouTube

• Music videos and lyrics channels
• Kids and animation content
• Gaming content (unless relevant)
• Clickbait and low-quality channels

Meta Audience Network

• Game apps
• Utility and calculator apps
• Foreign language apps
• Apps with under 100 reviews

Workflow

1. Data Collection

Pull data for the analysis period (typically 30-90 days):

Google Ads:

• Search terms report
• Placement report (Display + YouTube)
• Keyword performance report
• Campaign/ad group settings for CPA benchmarks

Meta Ads:

• Placement breakdown report
• Audience breakdown report
• Ad-level report with frequency data

2. Apply Thresholds

Calculate waste threshold per campaign using the formula above. Flag items exceeding thresholds.

3. Categorize

Assign each flagged item to a theme category. Group for batch exclusion.

4. Generate Exclusion Lists

Output CSV files ready for upload:

• negative-keywords.csv — with match type and campaign/ad group scope
• placement-exclusions.csv — with reason and spend wasted
• audience-exclusions.csv — with segment details and waste amount

5. Validate

• Cross-check that exclusions do not remove converting terms
• Verify match types will not over-exclude
• Estimate savings from the proposed exclusions

Safety Rules

1. Analysis only. Never make live account changes without human review.
2. Projections are estimates. Label all savings projections clearly.
3. Check for false positives. Some high-spend zero-conversion terms may be in a consideration window.
4. Consider attribution. View-through and assisted conversions may justify some "wasted" spend.
5. Preserve data. Export current state before applying any exclusions.

For GAQL query patterns and CSV formats, see REFERENCE.md.