Overview
While Exact Match rules are ideal for clear-cut identifiers (like Social Security Numbers or Email Addresses), real-world data is often messy. Names may have typos (e.g., "Jon" vs. "John"), and addresses may vary in formatting.
Fuzzy Match Rules (often managed through Relevance Based Actions) allow the UnifyApps MDM to identify duplicate records even when the data is not identical.
Instead of a simple "Yes/No" result, the system calculates a similarity score for record pairs, allowing you to define different actions based on how closely the records match.
The Scoring Mechanism
UnifyApps employs sophisticated comparator classes to generate a Match Score for every comparison.
• The Scale: The match score ranges from 0 to 1, where '1' represents a perfect, identical match and '0' indicates no similarity at all.
• Comparator Classes: Behind the scenes, the system uses algorithms (such as Levenshtein distance or Jaro-Winkler) to analyze the fields and compute this probability.
Configuring Relevance Based Actions
To implement fuzzy matching logic, you must configure the Rule Settings ane the Define Merge Policy step of your Match Rule to use Relevance Based Action.
This setting allows you to apply a scoring algorithm to determine whether records should be merged automatically or sent for manual review.
1. Select Action Strategy
In the "Define Merge Policy" section, select Relevance Based Action.
Unlike "Automatically Merge" (which acts blindly on a match) or "Flag for Review" (which queues everything), this option gives you granular control based on the match confidence.
2. Define Match Score Thresholds
Once selected, you can configure thresholds to categorize matches into different buckets:
• Auto-Merge Threshold: Set a high score (e.g., 0.9 to 1.0) where the system is confident enough to merge records without human intervention.
• Flag For Review Threshold: Set a mid-range score (e.g., 0.6 to 0.89) where the records look similar but require a Data Steward to verify the match.
• No Action: Records falling below the lower threshold are considered unique and ignored.
The interface provides a slider to visually adjust these ranges, ensuring you can fine-tune the sensitivity of your matching logic.
3. Data Steward Review
Records that fall into the "Flag for Review" bucket are queued in the Potential Matches section.
This allows your team to manually inspect edge cases—resolving ambiguities that automated logic cannot handle safely.