Fuzzy Matching

Understanding Fuzzy Matching

Fuzzy matching is a technique that finds text that approximately matches your brand name, even when there are misspellings, typos, or slight variations. This ensures you catch mentions that would be missed by exact matching alone.

How Fuzzy Matching Works

SearchRank AI uses string similarity algorithms to compare text against your brand name and aliases. The fuzzy threshold is a percentage that determines how similar a string must be to count as a match.

Similarity Calculation

Brand: SearchRank AI
Text Found: SearchRanc AI

Similarity Score: 92%
Threshold Set: 85%

Result: ✓ Match (92% >= 85%)

Threshold Levels

95-100%: Very Strict

• Nearly exact matches only
• Minimal false positives
• May miss common typos
• Best for: Short brand names or when precision is critical

85-94%: Strict (Recommended Default)

• Catches minor typos
• Low false positive rate
• Good balance of precision and recall
• Best for: Most brands, especially those with unique names

75-84%: Moderate

• Catches more variations
• Increased false positive risk
• Good for complex brand names
• Best for: Brands with common misspellings

60-74%: Loose

• Very permissive matching
• High false positive risk
• May catch unrelated terms
• Best for: Discovery phase or very long brand names

Configuring Your Threshold

1. Navigate to domain settings
2. Find "Fuzzy Matching Threshold"
3. Adjust the slider or enter a percentage
4. Save and test with sample prompts

Factors Affecting Optimal Threshold

Brand Name Length

Short names (3-5 chars): Use 90-95%
"Nike" - one character difference is significant

Medium names (6-12 chars): Use 85-90%
"SearchRank" - can tolerate minor typos

Long names (13+ chars): Use 75-85%
"SearchRank AI Platform" - more room for variation

Brand Name Uniqueness

Unique names: Can use lower threshold (80-85%)
"Xerox" - unlikely to have similar words

Common words: Need higher threshold (90-95%)
"Best Buy" - many similar phrases exist

Industry Context

Consider if similar terms exist in your industry that could cause false positives.

Testing Your Threshold

Step 1: Create Test Cases

Exact Match: "SearchRank AI" → Should match
Minor Typo: "SearchRanc AI" → Should match
Different Brand: "DataRank AI" → Should NOT match
Similar Term: "Search and Rank" → Should NOT match

Step 2: Run Tests

Execute prompts that return these variations and check detection accuracy.

Step 3: Adjust Threshold

If missing typos, lower threshold. If catching false positives, raise threshold.

Common Issues and Solutions

Issue: Missing Obvious Typos

Problem: "SearchRanc AI" not detected
Current Threshold: 95%

Solution: Lower to 85-90%

Issue: False Positives

Problem: "PageRank AI" detected as your brand
Current Threshold: 75%

Solution: Raise to 85-90% or use exact matching

Issue: Similar Brand Confusion

Problem: Competitor "SearchTrack AI" matches
Current Threshold: 80%

Solution: Raise to 90%+ or add as excluded term

Advanced: Combining with Aliases

Fuzzy matching applies to your primary brand name AND all aliases. This means:

• Misspellings of aliases are also caught
• Very specific aliases can use lower thresholds safely
• Generic aliases should use higher thresholds

Strategy: Layered Approach

Primary Brand: "SearchRank AI" → 85% threshold
Exact Aliases (no fuzzy needed):
- "SRAI"
- "SearchRankAI"

Allow fuzzy on aliases:
- "Search Rank" → catches "Serch Rank"
- "SR AI" → catches "SR-AI"

Monitoring Threshold Performance

Regularly review your detection results:

• True Positives: Correct brand detections
• False Positives: Incorrectly matched unrelated terms
• False Negatives: Missed brand mentions

Optimization Workflow

1. Start with 85% threshold (safe default)
2. Run initial prompts and review results
3. Identify false positives and false negatives
4. Adjust threshold in 5% increments
5. Test again and iterate
6. Document your optimal setting

Best Practices

1. Start conservative (85-90%) and lower if needed
2. Test with known variations before going live
3. Review detection reports weekly initially
4. Consider brand name characteristics when setting threshold
5. Use aliases for known variations rather than very low thresholds

With fuzzy matching configured, you're ready to track competitors in Competitor Tracking.