> ## Documentation Index
> Fetch the complete documentation index at: https://docs.glood.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Glood Similar Products

> How Glood.AI finds perfect alternatives with visual and semantic matching

# Glood's Similar Products: Smart Alternative Discovery

Glood's Similar Products feature helps customers find alternatives when their first choice isn't quite right—whether it's out of stock, wrong price, or not exactly what they wanted. Configure it in seconds through the Glood Dashboard, with powerful controls over how similarity is determined.

## Algorithm Architecture

```mermaid theme={null}
graph TD
    A[Product Being Viewed] --> B[Feature Extraction]
    
    B --> C[Text Analysis]
    B --> D[Visual Analysis]
    B --> E[Attribute Matching]
    B --> F[Behavioral Patterns]
    
    C --> G[GPT-4<br/>Semantic Embedding]
    D --> H[Gemini Pro<br/>Visual Embedding]
    E --> I[Glood Model<br/>Attribute Scoring]
    F --> J[Collaborative Filtering]
    
    G --> K[Multi-Modal Fusion]
    H --> K
    I --> K
    J --> K
    
    K --> L[Similarity Scoring]
    L --> M[Ranking & Filtering]
    M --> N[Final Recommendations]
    
    style N fill:#9f9,stroke:#333,stroke-width:2px
```

## AI Model Specializations

### OpenAI GPT-4: Semantic Understanding

GPT-4 analyzes product similarity through:

#### Natural Language Processing

* **Description Embedding**: Converts product descriptions into high-dimensional vectors
* **Feature Extraction**: Identifies key product attributes from text
* **Intent Matching**: Understands the core purpose and use case
* **Specification Parsing**: Extracts technical details for comparison

#### Semantic Similarity Scoring

```python theme={null}
# Conceptual representation
similarity = cosine_similarity(
    gpt4_embed(product_A_description),
    gpt4_embed(product_B_description)
)
```

#### Cross-lingual Capabilities

* Matches products across different languages
* Understands regional terminology variations
* Handles technical jargon and colloquialisms

### Google Gemini Pro: Visual Intelligence

Gemini provides advanced visual analysis:

#### Image Understanding

* **Style Detection**: Identifies design patterns, aesthetics, and visual themes
* **Color Analysis**: Extracts dominant colors and palettes
* **Shape Recognition**: Understands product form factors
* **Texture Identification**: Recognizes materials and finishes

#### Multi-modal Fusion

* Combines image and text understanding
* Validates text claims with visual evidence
* Identifies discrepancies between descriptions and images

#### Fashion & Design Expertise

* Pattern matching in clothing and textiles
* Furniture style classification
* Artistic and decorative element recognition

### Glood Proprietary Models: E-commerce Optimization

Our custom models focus on:

#### Behavioral Similarity

* **Co-view Patterns**: Products frequently viewed together
* **Switch Rates**: How often customers switch between products
* **Comparison Behavior**: Items compared in the same session
* **Purchase Substitution**: Actual replacement patterns

#### Commercial Factors

* **Price Band Matching**: Keeps recommendations in similar price ranges
* **Brand Affinity**: Considers brand preferences and loyalty
* **Quality Tier Alignment**: Matches premium with premium
* **Availability Scoring**: Prioritizes in-stock alternatives

## Similarity Dimensions

### 1. Functional Similarity

Products that solve the same problem:

* Running shoes � Other running shoes
* Coffee makers � Espresso machines
* Phone cases � Protective covers

### 2. Visual Similarity

Products that look alike:

* Similar color schemes
* Matching design aesthetics
* Comparable styles or patterns

### 3. Contextual Similarity

Products used in similar situations:

* Beach towels � Beach umbrellas
* Yoga mats � Yoga blocks
* Camping tents � Sleeping bags

### 4. Attribute Similarity

Products with matching specifications:

* Same size or dimensions
* Compatible features
* Similar performance metrics

## Advanced Techniques

### Hierarchical Similarity

```mermaid theme={null}
graph TB
    A[Exact Match<br/>Same SKU, Different Color] --> B[Variant Match<br/>Same Product, Different Size]
    B --> C[Model Match<br/>Same Brand, Similar Model]
    C --> D[Category Match<br/>Same Subcategory]
    D --> E[Purpose Match<br/>Similar Use Case]
    E --> F[Style Match<br/>Similar Aesthetic]
```

### Dynamic Weight Adjustment

The algorithm adjusts similarity weights based on:

* **Product Category**: Fashion emphasizes visual, electronics emphasizes specs
* **Price Point**: Luxury items weight brand higher
* **Customer Segment**: Tech-savvy users get spec-heavy matches
* **Session Context**: Search queries influence weight distribution

### Negative Sampling

We explicitly learn what's NOT similar:

* Products with high return rates when bought together
* Items frequently removed from comparison
* Negative review mentions of alternatives

## Real-time Processing

### Feature Extraction Pipeline

1. **Image Processing** (50ms)
   * Resize and normalize
   * Extract visual features
   * Generate embeddings

2. **Text Processing** (30ms)
   * Tokenize descriptions
   * Generate semantic embeddings
   * Extract entities and attributes

3. **Similarity Computation** (20ms)
   * Calculate multi-modal distances
   * Apply category-specific weights
   * Generate ranked list

### Caching Strategy

* **Embedding Cache**: Pre-computed embeddings for all products
* **Similarity Matrix**: Pre-calculated top-N similar items
* **Real-time Adjustment**: Dynamic re-ranking based on context

## Quality Metrics

### Relevance Metrics

* **Click-through Rate**: 45% on similar product carousels
* **Dwell Time**: 3.2x longer on recommended products
* **Conversion Rate**: 22% purchase similar items
* **Return Rate**: 15% lower for AI-recommended alternatives

### Diversity Metrics

* **Category Coverage**: Recommendations span appropriate subcategories
* **Price Range**: �20% of original product price
* **Brand Mix**: Balance between same-brand and alternatives
* **Visual Variety**: Different colors/styles when appropriate

## Use Case Examples

### Fashion Retail

```yaml theme={null}
Viewing: Blue Denim Jacket
Similar Products:
  - Light Wash Denim Jacket (color variation)
  - Canvas Jacket (material alternative)
  - Bomber Jacket (style alternative)
  - Designer Denim Jacket (premium option)
```

### Electronics

```yaml theme={null}
Viewing: 65" 4K Smart TV
Similar Products:
  - 55" 4K Smart TV (size variation)
  - 65" OLED TV (technology upgrade)
  - 65" QLED TV (brand alternative)
  - 70" 4K Smart TV (size upgrade)
```

### Home Decor

```yaml theme={null}
Viewing: Modern Floor Lamp
Similar Products:
  - Arc Floor Lamp (style variation)
  - Tripod Floor Lamp (design alternative)
  - Smart Floor Lamp (feature upgrade)
  - Industrial Floor Lamp (aesthetic alternative)
```

## Continuous Improvement

### Learning Mechanisms

1. **Click Feedback**: Learn from which similarities users explore
2. **Purchase Analysis**: Understand actual substitution patterns
3. **Return Analysis**: Identify poor similarity matches
4. **Session Analysis**: Learn from comparison behavior

### Model Updates

* **Weekly Retraining**: Behavioral models updated weekly
* **Daily Embeddings**: New products get embeddings daily
* **Real-time Adjustments**: Weights adjust based on performance
* **Quarterly Reviews**: Major algorithm improvements

## Glood Dashboard Configuration

### Setting Up Similar Products

1. **Quick Setup**
   ```yaml theme={null}
   Section Name: "You May Also Like"
   Algorithm: Similar Products
   Number of Products: 6
   Layout: Grid (2x3)
   Similarity Type: Balanced
   ```

2. **Similarity Controls**
   * **Visual Weight** (0-100%): How much appearance matters
   * **Price Weight** (0-100%): Keep alternatives in budget
   * **Brand Weight** (0-100%): Same brand vs competitors
   * **Category Strictness**: Same category or explore related

3. **Manual Similarity Mapping**
   ```csv theme={null}
   product_id,similar_1,similar_2,similar_3,priority
   SHOE001,SHOE002,SHOE003,SHOE004,high
   BAG001,BAG002,BAG003,BAG004,medium
   ```

### Glood's Similarity Dashboard

```javascript theme={null}
// Glood similarity configuration
{
  "similarity_config": {
    "algorithm": "hybrid",
    "weights": {
      "visual": 0.4,
      "semantic": 0.3,
      "behavioral": 0.2,
      "price": 0.1
    },
    "constraints": {
      "price_variance": 0.3,  // ±30% price
      "min_score": 0.6,       // Minimum similarity
      "max_results": 12,      // Pool size
      "diversity": 0.3        // Result variety
    }
  }
}
```

## Glood-Specific Features

### Visual Similarity Toggle

Enable/disable visual matching per category:

* **Fashion**: High visual weight (70%)
* **Electronics**: Spec-based (visual 20%)
* **Furniture**: Balanced (visual 50%)
* **Consumables**: Feature-based (visual 10%)

### Smart Filtering Rules

Set up in Glood Dashboard:

1. **Stock Status**: Hide out-of-stock alternatives
2. **Price Bands**: Define acceptable price ranges
3. **Brand Rules**: Same-brand only or mix
4. **Exclusions**: Never show certain products
5. **Promotions**: Boost sale items

### A/B Testing Framework

```yaml theme={null}
Test Variations:
  Control: AI-only similarity
  Variant A: Manual overrides + AI
  Variant B: Visual-heavy matching
  Variant C: Price-conscious alternatives
  
Metrics Tracked:
  - Click-through rate
  - Conversion rate
  - Time to purchase
  - Customer satisfaction
```

## Glood Analytics for Similar Products

### Performance Metrics

Monitor in your Glood Dashboard:

* **Alternative Click Rate**: 45% average
* **Cross-sell Success**: 23% buy alternatives
* **Out-of-stock Recovery**: 67% find substitute
* **Session Extension**: +4.2 pages per visit

### Popular Use Cases

#### Out-of-Stock Management

```yaml theme={null}
Scenario: Product unavailable
Glood Action: Auto-show similar in-stock items
Result: 72% cart recovery rate
```

#### Price Sensitivity

```yaml theme={null}
Scenario: Customer views premium product
Glood Action: Include budget alternatives
Result: 34% convert to lower price point
```

#### Style Discovery

```yaml theme={null}
Scenario: Fashion browsing
Glood Action: Show different colors/patterns
Result: 2.8x items viewed per session
```

## Implementation Best Practices

### With Glood Dashboard

1. **Initial Setup** (5 minutes)
   * Install Glood from Shopify App Store
   * Create "Similar Products" section
   * Choose display locations
   * Customize appearance

2. **Optimization** (Ongoing)
   * Review weekly analytics
   * Adjust similarity weights
   * Add manual mappings for key products
   * Test different layouts

3. **Advanced Configuration**
   * Set up category-specific rules
   * Create seasonal similarity adjustments
   * Configure mobile vs desktop differences
   * Implement personalized similarity

### Common Glood Configurations

#### Fashion Store

```yaml theme={null}
Visual Weight: 70%
Brand Weight: 20%
Price Weight: 10%
Show: Different colors, similar styles
```

#### Electronics Store

```yaml theme={null}
Spec Weight: 60%
Price Weight: 25%
Brand Weight: 15%
Show: Compatible alternatives
```

#### Home Decor

```yaml theme={null}
Visual Weight: 50%
Style Weight: 30%
Price Weight: 20%
Show: Complementary pieces
```

## Success with Glood Similar Products

### Case Studies

**Fashion Retailer**

* Increased page views by 156%
* Reduced bounce rate by 42%
* Generated \$2.3M additional revenue

**Electronics Store**

* 89% out-of-stock recovery rate
* 34% higher AOV
* 4.5/5 customer satisfaction

**Furniture Store**

* 67% explored similar styles
* 23% bought multiple items
* 3x longer session duration

For step-by-step implementation, see our [Glood Similar Products Setup Guide](/guides/create-sections/how-to-create-similar-products-section).