AI Dashboards - Scoring models

Scoring Models

1. Overview

The Scoring Models module is part of the AI Dashboards system used to adjust ranking signals for search, recommendation, or discovery results.

It allows you to define boosting weights for specific brands, categories, or behavioral signals to influence result ranking performance.


2. Navigation

Left Sidebar (Scoring Models):

  • Brand Booster (selected) → Adjust ranking by brand strength
  • Category Booster → Adjust ranking by product/category type
  • Shopping Insights → Analyze shopping behavior signals
  • Metadata Manager → Manage supporting metadata for scoring logic

Top Right:

  • Website Selector (e.g., DEF) → Switch environment/site
  • Save Button → Save configuration changes

3. Brand Booster Configuration

The current view shows a JSON-based configuration editor for brand scoring weights.

Example Structure

{
  "Apple": 1.3,
  "Officeo": 1.4
}

4. How It Works

Each key-value pair represents:

Field Description
Brand Name The brand being scored
Score Weight Multiplier applied to ranking relevance

5. Scoring Logic

  • A higher score (>1.0) increases visibility in rankings
  • A lower score (<1.0) decreases visibility
  • A neutral score (1.0) means no ranking influence

Example Interpretation:

  • Apple: 1.3   → Apple-related items are boosted by 30%
  • Officeo: 1.4   → Officeo-related items are boosted by 40%

6. Use Cases

Brand Booster

Used to:

  • Promote preferred brands
  • Improve visibility for strategic partners
  • Boost high-performing or high-converting brands

Category Booster (Module Purpose)

Although not shown in the screenshot, this module typically:

  • Boosts product categories (e.g., Electronics, Furniture)
  • Helps align ranking with business priorities

Shopping Insights (Module Purpose)

Used for:

  • Analyzing user shopping behavior
  • Identifying trending products/categories
  • Feeding signals into ranking systems

Metadata Manager (Module Purpose)

Used for:

  • Managing additional attributes used in scoring
  • Supporting filters and enrichment data
  • Maintaining consistency across models

7. Configuration Rules

  • JSON must be valid (no trailing commas)
  • Brand names must match system taxonomy
  • Values must be numeric (float or integer)
  • Recommended range: 0.0 – 2.0  
  • Changes only take effect after clicking Save

8. Impact on System

Once applied, scoring values influence:

  • Search result ranking
  • Product listing order
  • Recommendation systems
  • Personalized discovery feeds

9. Best Practices

  • Avoid excessive boosting (>1.5) unless necessary
  • Regularly review brand performance
  • Balance boosts across categories to avoid bias
  • Test changes in staging before production deployment

10. Summary

The Brand Booster under Scoring Models is a lightweight but powerful configuration tool that allows dynamic adjustment of brand importance in ranking systems using simple numeric multipliers.

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.