Content Variation Engine: Multi-Location Uniqueness via Jaccard Scoring
Multi-location businesses need unique content per location — not the same page with "[CITY]" swapped out. Jaccard scoring measures actual phrase-level uniqueness. This system targets 0.75-0.85 similarity (on-topic but not duplicate).
Why Jaccard Scoring
Jaccard similarity measures shared phrases between two documents:
Jaccard(A, B) = |A ∩ B| / |A ∪ B|
- Score of 1.0 = identical documents
- Score of 0.0 = no shared phrases
- Target: 0.75-0.85 — similar enough to be on-topic, unique enough to avoid duplicate content penalties
Scores above 0.90 = near-duplicate = Google may deindex or suppress the variant page. Scores below 0.70 = so different they may not rank for the same keyword family.
The 7 Variation Axes
Each location page variation draws uniqueness from these seven sources:
Axis 1: Geographic References
- City name, county, neighborhood names
- Local landmarks, districts, zip codes
- Regional terminology (Chicagoland vs greater Chicago area)
- Distance references ("serving [City] and surrounding [radius] mile area")
Axis 2: Local Social Proof
- Customer testimonials from that specific city
- Named projects in that area
- Local contractor/partner mentions
Axis 3: Local Context
- City-specific regulations or codes (if applicable)
- Local weather patterns affecting the service
- Area-specific pricing considerations
Axis 4: Different Examples
- Same service explained with different examples per location
- Different case study or before/after for each city
Axis 5: Different Stats
- Local stat sourced from city/county data
- Area-specific market condition
- Neighborhood-level data if available
Axis 6: Opening and Closing Paragraphs
- Intro and CTA rewritten entirely — never shared across variants
- Different opening hook angle per location
Axis 7: FAQ Variation
- 2-3 FAQ questions that are city-specific
- Local permit, code, or process questions
- City-named queries ("What do roofing companies in Denver charge?")
Rewrite Protocol
Step 1: Write the master page (anchor location — typically headquarters or primary market)
Step 2: For each variant location, apply variation to these sections in order:
| Section | Change Required | Notes | |---------|----------------|-------| | Title tag + H1 | Always | City name mandatory | | Intro paragraph | Full rewrite | New opening hook | | Social proof section | New testimonials | Location-specific | | Process/differentiator section | Light edit | Same structure, new examples | | Local context section | New content | City-specific info | | FAQ section | 2-3 unique Qs added | Keep 3-4 from master | | CTA section | Light edit | Same CTA, city name | | Meta description | Full rewrite | City + unique value prop |
Step 3: Run Jaccard check before publishing.
Jaccard Measurement (Manual Method)
For teams without automated tools:
- Extract all unique 3-word phrases from master page
- Extract all unique 3-word phrases from variant page
- Count phrases in both (intersection)
- Count total unique phrases (union)
- Divide intersection by union
If score exceeds 0.85: identify the repeated phrase blocks and rewrite those sections specifically.
Quick proxy test: If 3 consecutive sentences in the variant are word-for-word identical to the master, the section needs a rewrite.
Minimum Uniqueness by Page Section
| Section | Minimum Unique Content | |---------|----------------------| | H1 + Title | 100% (city-modified) | | Intro paragraph | 100% (full rewrite) | | Social proof | 100% (new testimonials) | | Body sections | 40-60% unique per section | | FAQ | 30%+ unique questions | | CTA | 20%+ unique (city name minimum) |
Overall page: Target 35-40% unique content at minimum. 50%+ is the safe zone.
Scale Note
At 20+ location pages, manual variation becomes unsustainable. At that scale:
- Templatize the stable sections (process, credentials, guarantees)
- Build a local data insert layer (city name, local stat, testimonial)
- Write new intro paragraphs (the highest-weight uniqueness section) for every page
- Automated Jaccard checks before publish
#content-sop #multi-location #local-seo #uniqueness #duplicate-content