google-ads-keyword-engine

Process

For new keyword research:

1. Define the business, products/services, and target customer
2. Mine customer language from VOC sources
3. Build seed keyword list by intent category
4. Expand using tools and techniques
5. Evaluate and prioritize using the scoring framework
6. Assign match types and organize into ad groups
7. Build negative keyword lists

For search term analysis (existing campaigns):

1. Pull search term report (minimum 90 days, or 30 days for new campaigns)
2. Run n-gram analysis (1-gram, 2-gram, 3-gram)
3. Identify waste (high spend, zero conversions)
4. Identify winners (strong CPA/ROAS)
5. Take action: negate, promote, or investigate
6. Update negative keyword lists

Customer Language Mining

Start with real customer language, not internal jargon. These sources give you the words actual buyers use:

Keyword Evaluation Framework

Score every keyword candidate on four dimensions:

Priority Score = (Intent × 0.35) + (Volume × 0.25) + (Competition × 0.20) + (CPC × 0.20)

Intent classification guide:

Match Type Strategy (2025)

Critical requirement for broad match: Must be paired with Smart Bidding (tCPA or tROAS) and have 50+ conversions/month. Broad match with manual bidding or low conversion volume will waste budget on irrelevant queries.

ECPC is deprecated (March 2025). If someone asks about broad match + ECPC, redirect to Manual CPC or Maximize Conversions as the bidding pair.

N-Gram Analysis Process

N-gram analysis breaks search queries into word-level components to find patterns invisible at the query level.

What it is:

• 1-gram: single words ("shoes", "running", "waterproof")
• 2-gram: word pairs ("running shoes", "waterproof shoes")
• 3-gram: word triplets ("best running shoes", "waterproof trail shoes")

How to run it:

1. Export search term report (90 days minimum, queries with 3+ impressions)
2. Break every query into its n-grams
3. Aggregate metrics by n-gram: total impressions, clicks, spend, conversions, CPA, ROAS
4. Sort by spend descending to find biggest waste, or by CPA to find best performers

Action framework:

Fractional conversion warning: Data-driven attribution creates fractional conversions (e.g., 0.4 conversions). A keyword showing 0.3 conversions contributed to a conversion elsewhere. Filter to conversions > 1 before evaluating CPA to avoid false negatives.

Tools for n-gram analysis:

• Brainlabs Search Query Mining Script (updated for GAQL by Nils Rooijmans, March 2025) — automated, runs in Google Ads
• Adalysis — auto-generated n-gram reports
• PEMAVOR Search Term Miner — free, no-code
• BigQuery SQL — enterprise-scale analysis
• For script implementation, use the google-ads-scripts skill

Negative Keyword Strategy

Themed Negative Lists

Build shared negative keyword lists by category and apply across relevant campaigns:

Cross-Campaign Negatives

• Add exact-match negatives to prevent campaigns from competing against each other
• PMax cannibalization: if PMax search term insights show it's serving on queries your Search campaigns should own, add those as PMax negatives AND add them as exact-match keywords in Search

Maintenance Cadence

• Week 1-4 of new campaigns: Review search terms weekly
• Stable campaigns: Bi-weekly review
• Always check: Negatives aren't blocking desired traffic (Brainlabs script has checkNegatives parameter)

Tool Recommendations

Output Format

For keyword research:

## Keyword Research: [Business/Product]

### Keyword List
| Keyword | Match Type | Intent | Volume | CPC | Competition | Priority Score | Ad Group |
|---------|-----------|--------|--------|-----|-------------|---------------|----------|
| [keyword] | Exact | Commercial | 1,200 | $4.50 | Medium | 8.2 | [group] |

### Ad Group Structure
- Ad Group: [Theme]
  - Keywords: [list with match types]
  - Negative keywords: [list]

### Negative Keyword Lists
- Competitors: [list]
- Jobs/Careers: [list]
- Free/DIY: [list]
- [Custom theme]: [list]

### Match Type Rationale
[Why these match types were chosen based on account maturity and conversion volume]

For search term analysis:

## Search Term Analysis: [Date Range]

### N-Gram Waste Report
| N-gram | Impressions | Clicks | Spend | Conversions | CPA | Action |
|--------|------------|--------|-------|-------------|-----|--------|
| [n-gram] | 2,400 | 89 | $312 | 0 | ∞ | Negate |

### Promotion Candidates
| Search Term | Clicks | Conversions | CPA | Current Match | Action |
|------------|--------|-------------|-----|---------------|--------|
| [term] | 45 | 12 | $18 | Broad trigger | Promote to exact in new ad group |

### Negative Keyword Additions
| Keyword | Match Type | List | Reason |
|---------|-----------|------|--------|
| [keyword] | Phrase | Jobs | Zero conversions, $89 spend |

### Summary
- Total waste identified: $[X]
- Promotion candidates: [N] terms
- New negatives added: [N] keywords across [N] lists

Example

Input: "I run a SaaS project management tool competing with two large incumbents. We're spending $20K/month on Google Ads with a $120 target CPA. We have 65 conversions/month. I need help with our keyword strategy and a search term cleanup."

Output:

Keyword Strategy & Search Term Analysis

Match Type Recommendation

With 65 conversions/month and active Smart Bidding, you qualify for broad match + tCPA as your primary strategy. Keep exact match for brand terms and highest-intent commercial keywords.

Keyword List (Top Priority Additions)

N-Gram Waste Report (from search term analysis)

Promotion Candidates

Negative Keyword Lists

• Competitors (protect from brand overlap): asana, monday.com, clickup, notion, trello, basecamp, jira, wrike (applied as exact match to non-competitor campaigns)
• Jobs/Careers: jobs, salary, careers, hiring, internship, glassdoor, project manager salary
• Free/DIY: free, open source, template, free trial hack, cracked
• Education: what is project management, certification, PMP, course, degree, tutorial

PMax Cannibalization Check

Review PMax search term insights for queries like "asana alternative" and "project management software" — these should be served by Search exact-match campaigns, not PMax. Add as PMax negatives if PMax is stealing them.

Guidelines

• Don't recommend broad match without Smart Bidding + 50 conversions/month. This is a hard requirement. Broad match with manual bidding or low volume wastes budget on irrelevant queries.
• Don't judge keywords on fractional conversions. Data-driven attribution creates fractions. A keyword with 0.4 conversions isn't failing — it contributed. Filter to > 1 conversion before evaluating CPA.
• Don't build negative lists once and forget them. Search term review is ongoing: weekly for new campaigns, bi-weekly for stable ones.
• Don't add negatives without checking for conflicts. A negative keyword can block a desired positive keyword. Always cross-reference.
• Don't rely solely on Keyword Planner. It shows what people search, not how they think. VOC mining from sales calls, reviews, and forums reveals language Keyword Planner misses.
• ECPC is gone (March 2025). Don't reference it as a bidding option. If someone's account still shows ECPC, flag it and recommend migration.
• PMax search term visibility is here (2025). Use PMax search term insights and campaign-level negatives to manage cannibalization between PMax and Search.
• Cross-references: For n-gram automation scripts, use the google-ads-scripts skill. For campaign building with these keywords, use the google-ads-campaign-builder skill. For ongoing keyword optimization, use the google-ads-optimizer skill. For full account audit including keyword health, use the google-ads-audit skill.
• Confidence: Volume estimates from Keyword Planner are ranges, not exact numbers. CPC estimates can vary significantly from actuals. Present keyword data as directional, not precise.