How to Analyze Match Rules in Reltio

The Reltio platform provides API and UI tooling to troubleshoot matching rule configurations]. When a tenant encounters performance issues during match processing, this tool helps identify three primary configuration problems:

• Overcollisioned tokens 

• Too many tokens generated for an entity 

• Duplicate or overlapping matching groups/rules 

Before analyzing results, run the analyzer task 

1. Open Tenant Management in the Reltio Console for your tenant 

2. In Tenant Management, open the Jobs screen and click NEW JOB 

3. Select the Analyze match rule task 

4. Configure the job parameters:

   • Entity Types: Select the entity types you want to investigate (e.g., Organization)

   • Sample Size: Select the sample size 
     (Note: Because analysis occurs in memory, large sample sets cannot be selected, and issues outside the sample window may not be captured for large tenants 

5. Click Run and monitor progress under Pending tasks until completion 

1. When the job finishes, open the completed task 

2. The initial screen displays the selected entity types and potential rule issues

3. Select the target entity type (e.g., Organization) to open its alert list 

4. Review the alerts and short descriptions, then click View details to open the in-depth analysis view 

The detailed view is broken down into two main diagnostic areas [

Token generation issues are critical because they are the main cause of match processing latency

• Token Distribution Thresholds:

  • Lower Threshold (< 5 tokens): Entities should have more than 5 tokens in order to generate enough comparison pairs 

  • Upper Threshold (> 300 tokens): Entities should have fewer than 300 tokens. A large number of tokens can cause overcollision or force unnecessary comparison pairs

• Candidate Comparison Table:

  • Identifies tokens generating abnormally large candidate counts 
    (Causes can include an overcollisioned token or a dataset with many genuine potential matches 

  • Click View more to see example entities sharing the token and identify the match rule that generated it (e.g., ExternalMatchRule3 - Exact(Country),Fuzzy(OrgName)) 

  • Verify if the token is overcollisioned, and inspect token generation details using the matchTokenDetailed endpoint 

Evaluates how match rules correlate with each other

• Red Fields (High Overlap / Duplicate Results):

  • Indicates that a pair of rules generates the same entity results.

  • Resolution: Merge the rules or use ignoreInToken
    Example: If two rules differ only by an equals block (such as ReltioCertified or ManagedByReltio), two separate token sets are created (e.g., JIiqf-JBB-0:4500 vs JIiqf-JBB-0:4500:true:true). Adding those attributes to ignoreInToken removes the trailing flags and avoids duplicating tokens 

• High Correlation Scores (e.g., Correlation = 0.67):

  • Indicates that one rule expands another 
    Example: Comparing Exact(DUNSNumber, Name, Country) with Exact(DUNSNumber). Because DUNSNumber has high cardinality, adding Name and Country to ignoreInToken cuts the token count in half without sacrificing matching power

• Blue Fields (Low / No Correlation):

  • Indicates that the rules do not correlate, confirming that they address distinct match criteria 

• Maximum Sample Size Cap: [Unknown] — The documentation states that sample sizes are constrained due to in-memory processing, but the exact upper limit allowed in the Console UI is not specified.

  • Documentation needed: UI documentation specifying the maximum integer accepted for sample size.

• Overcollision Threshold Numbers: [Unknown] — While the page references investigating overcollisioned tokens, the specific candidate volume that triggers an overcollision alert is not detailed on the source page.

  • Documentation needed: Review of the How-to check overcollisioned tokens sub-page.

Documentation links

• Match Rule Analyzer version 2 (Dynamic)
• Inspections for Tokens
• Inspections for Overcollisioned Tokens
• Overcollisioned Tokens
• Search Overcollisioned Tokens API