Skip to main content

Business rules not overriding Boost with Signals document scoring

Updated

Issue

A document configured for priority ranking through a Business Rule does not appear at the top of search results. Even with extreme boost values applied using a Boost List rule or Boost Documents stage, the document remains outranked by others when the Boost with Signals stage is active in the query pipeline.

 

Diagnosis

To verify this condition, inspect the query pipeline behavior:

  • Run the query with debug=true and examine the responseHeader and debug sections of the Solr response.

  • Review the document scores for both manually boosted and signal-boosted documents.

  • Temporarily disable the Boost with Signals stage to confirm whether the business rule takes effect in isolation.

If the document boosts work correctly when the Boost with Signals stage is disabled, it confirms the issue is due to the weight of signal-based boosts exceeding that of the business rule.

 

Environment

Fusion 4.2.6 SP3 and Fusion 5.x.

Cause

Boost with Signals uses the Final Boost Weight Expression to compute the score contribution from user behavior signals. These computed boosts can overwhelm static boosts configured via Business Rules or manual rules if not properly weighted. This behavior is working as designed but may require tuning for the intended outcome.

 

Resolution

Adjust the balance between Boost with Signals and Business Rules. The following are effective approaches:

Option 1: Use a Boost Attributes rule instead of a Boost List rule

Boost Attributes rules allow you to set an explicit boost value. This provides greater control and easier adjustment compared to Boost List rules. When creating the rule:

  • Use a Boost Query expression similar to:
    id:<target-doc-id>^2000000

  • Adjust the numeric value to achieve the desired ranking effect.

This method avoids needing a custom JavaScript stage.

Option 2: Decrease the boost strength from Boost with Signals

Modify the Final Boost Weight Expression in the Boost with Signals stage to reduce its influence on document scoring. This will affect all boosts using this query pipeline. For example:

Default expression:

math:log(weight_d + 1) + 10 * math:log(score + 1)

Adjusted expression:

math:log(weight_d + 1) + 5 * math:log(score + 1)

Reducing the multiplier (e.g., from 10 to 5) scales down signal-based boost strength, allowing static business rules to assert greater influence over ranking.

This approach is preferred when overall signal strength is too dominant and a consistent, system-wide recalibration is desired.

Relevancy testing recommended

After applying any of these changes, test the queries affected to ensure the changes order documents the way you want them to.

Powered by Zendesk