google-ads-keyword-engine

Files (1)

• SKILL.md

SKILL.md (first lines)

## 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:

| Source | What to Extract | How to Use |
|--------|----------------|-----------|
| Sales call recordings | Exact phrases prospects use to describe problems | Map to high-intent keywords |
| Support tickets | Common questions, complaint language | Map to informational and comparison keywords |
| Product reviews (yours) | Feature language, benefit descriptions | Feed into RSA headlines and keyword themes |
| Competitor reviews | Complaints and unmet needs | Target competitor weakness terms |
| Reddit / Quora | Thread titles, question phrasing | Long-tail keyword discovery |
| Industry forums | Technical terminology, emerging trends | Niche keyword expansion |
| Autocomplete / "related searches" | Google's own language suggestions | Validate demand and discover variants |

## Keyword Evaluation Framework

Score every keyword candidate on four dimensions:

| Dimension | Weight | How to Assess | Scoring |
|-----------|--------|--------------|---------|
| **Intent** | 35% | Commercial > Comparison > Informational > Navigational | Commercial intent = 10, Informational = 3 |
| **Volume** | 25% | Monthly search volume from Keyword Planner | Relative to niche: top 20% = 10, bottom 20% = 2 |
| **Competition** | 20% | Keyword Planner competition level + top-of-page bid | Low competition + high intent = ideal |
| **CPC** | 20% | Estimated cost vs target CPA margin | CPC < 20% of target CPA = 10, CPC > 50% of target CPA = 3 |

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

**Intent classification guide:**

| Intent Type | Signal Words | Conversion Likelihood | Example |
|------------|-------------|----------------------|---------|
| Commercial | buy, hire, pricing, cost, quote, near me, best, top | Highest | "hire plumber near me" |
| Comparison | vs, compare, review, alternative, which | High | "hubspot vs salesforce" |
| Informational | how to, what is, guide, tutorial, tips | Low (but useful for TOFU) | "how to fix leaky faucet" |
| Navigational | [brand name], login, website | Varies | "hubspot login" |

## Match Type Strategy (2025)

| Scenario | Recommended Match Type | Why |
|----------|----------------------|-----|
| Mature account (50+ conversions/month) + Smart Bidding | **Broad match** as default | Algorithm + Smart Bidding handles relevance. This is Google's 2025 recommendation |
| Brand terms | **Exact match** | Protect brand queries from broad expansion. Control messaging |
| High-intent, high-CPC terms | **Exact match** | Precise control on expensive clicks |
| Controlled expansion | **Phrase match** | Captures variations while maintaining intent direction |
| New account (<30 conversions/month) | **Phrase + Exact** | Not enough data for broad match + Smart Bidding to work |
| Testing new keyword themes | **Phrase match** first | Gather data before committing to broad |

**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:**

| N-gram Pattern | Criteria | Action |
|---------------|----------|--------|
| High spend, 0 conversions | $50+ spend AND 0 conversions | Add as negative keyword |
| High spend, poor CPA | Spend > $100 AND CPA > 2x target | Investigate — may need negative or ad group restructure |
| Strong CPA/ROAS | CPA < target AND 3+ conversions | Promote to exact-match keyword in dedicated ad group |
| High impressions, low CTR | 500+ impressions AND CTR < 1% | Intent mismatch — investigate query relevance |
| Emerging theme | Multiple related n-grams performing well | Create new ad group around this theme |

**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:

| List Theme | Example Negatives | Apply To |
|-----------|-------------------|----------|
| Competitors | [competitor brand names] | All non-competitor campaigns |
| Jobs / Careers | jobs, salary, careers, hiring, internship, glassdoor | All campaigns |
| Free / DIY | free, DIY, template, open source, how to, tutorial | Bottom-funnel campaigns |
| Education / Info | what is, definition, wiki, course, certification | Bottom-funnel campaigns |
| Geographic exclusions | [irrelevant cities, states, countries] | Geo-targeted campaigns |
| Irrelevant industries | [industry terms not served] | All campaigns |
| Negative modifiers | cheap, used, wholesale (if premium brand) | Brand campaigns |

From narayan-metaflow/metaflow-marketing-skills, MIT.

More from narayan-metaflow/metaflow-marketing-skills