Performance Marketing Knowledge Pack Files

Keyword Research Methodology

Structured process for building and maintaining a keyword universe for paid search campaigns.

Keyword Universe Building (5-Step Process)

Step 1: Plant Seeds (10-20 starter keywords)

Generate seeds from three angles:

Angle	What to Ask	Example Seeds
Product	What do you sell?	"crm software", "project management tool"
Problem	What pain does the customer have?	"manage customer relationships", "track sales pipeline"
Solution	What outcome do they want?	"close more deals", "automate follow-ups"

Sources for seeds: product pages, sales call transcripts, competitor homepages, customer support FAQs, Google Search Console queries.

Step 2: Expand (200-2,000 keywords)

Method	What	Expected Yield
Keyword tools (Google Keyword Planner, etc.)	Volume, competition, CPC data for seeds	50-200 per seed
Competitor analysis	Keywords competitors bid on	100-500 unique
Search term reports (existing accounts)	Actual queries already triggering ads	50-500
"People Also Ask" mining	Question-format long-tail queries	20-50
Autocomplete suggestions	Google autocomplete from seeds	10-30 per seed
Related searches	Bottom of SERP related queries	5-10 per search

Step 3: Classify Intent

Intent Level	Signals	PPC Value	Match Type
Transactional	buy, price, cost, order, near me, [brand]+[product]	Highest	Exact, Phrase
Commercial Investigation	best, review, comparison, vs, alternative, top 10	High	Phrase, Exact
Navigational	[brand name], [product name], login, support	Medium (brand campaigns)	Exact
Informational	how to, what is, guide, tutorial, definition	Low — usually exclude from PPC	Negative or Broad (with caution)

Keyword Opportunity Score (KOS)

Score every keyword to prioritize the list:

KOS = (Monthly Volume × Intent Multiplier × Est. CVR) / (Competition Score × Est. CPC)

Factor	How to Score	Scale
Monthly Volume	From keyword tool	Raw number
Intent Multiplier	Transactional=3, Commercial=2, Informational=0.5	0.5-3
Est. CVR	From historical data or industry avg (2-5% search)	0.01-0.10
Competition Score	Low=1, Medium=2, High=3	1-3
Est. CPC	From keyword tool	Raw number

Higher KOS = better opportunity. Focus on top quartile for initial campaign build.

Step 4: Group into Clusters

Clustering Rule	Guideline
One intent per group	Don't mix "buy crm" and "what is crm"
10-20 keywords per ad group	Tight enough for relevant RSAs
Similar CPC range	Prevents bid mismatch within group
Shared landing page	All keywords in group should map to same page

Naming: {theme}_{intent}_{match-type} — e.g., crm-software_transactional_exact

Step 5: Maintain

Cadence	Action
Weekly (new campaigns)	Review search terms, add negatives, promote converters
Biweekly (mature)	Search term review, keyword performance check
Monthly	Refresh universe with new seeds from search terms, competitor changes
Quarterly	Full re-evaluation of keyword list vs business goals

Search Term Classification: Negative Keyword Scoring

Score each search term 0-100 for exclusion priority:

Dimension	Points	What It Measures
Semantic Relevance	0-30	How far from business offerings. Larger gap = higher score
Conversion Probability	0-25	Historical performance + predicted likelihood. Lower = higher score
Cost Impact	0-20	Frequency × CPC. Higher projected waste = higher score
Brand Safety	90-100 auto	Profanity, adult, illegal content. Auto-max regardless
Keyword Conflict	-15 to -25	Reduces score when negative would block active keywords

Action by score:

Score	Classification	Action	Timeline
0-20	Relevant	No action	—
21-40	Monitor	Weekly review	1-2 weeks
41-60	Likely irrelevant	Batch upload as negative	1 week
61-80	High priority	Add negative	48 hours
81-100	Critical	Immediate exclusion	Same day

When confidence is below 85%, require human review. The 40-70 range is where most judgment calls happen.

Quick Decision Rules

Condition	Action
Cost > 2× target CPA, 0 conversions	Add as negative
Clicks > 60, 0 conversions	Strong negative candidate
CTR < 1% with 100+ impressions	Review ad relevance; likely negative
Converting at or below target CPA	Promote to keyword in matching ad group
High volume, good CTR, no conversion data yet	Monitor 30-60 days before deciding
Search term matches existing keyword exactly	Already covered; no action needed

N-gram Analysis

For large search term reports (500+ terms), run n-gram analysis:

1. Break all search terms into 1-gram, 2-gram, and 3-gram tokens
2. Count frequency of each token
3. Calculate total cost and conversions per token
4. Flag tokens appearing in 5%+ of terms with 0 conversions and significant cost
5. Add flagged tokens as campaign-level negatives

Example: If "free" appears in 47 search terms, cost $340 total, 0 conversions → add "free" as campaign negative.