> ## 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 Personalized For You (PFY)

> How Glood.AI creates hyper-personalized recommendations for each customer

# Glood's Personalized For You: Individual Customer Intelligence

Glood's Personalized For You (PFY) feature is our most sophisticated recommendation algorithm, creating unique product suggestions for each individual visitor to your Shopify store. Accessible through the Glood Dashboard with just a few clicks, PFY combines deep learning with real-time personalization to maximize conversions.

## Algorithm Overview

```mermaid theme={null}
graph TD
    A[User Profile] --> B[Multi-Source Data Aggregation]
    B --> C[Behavioral History]
    B --> D[Purchase Patterns]
    B --> E[Browsing Signals]
    B --> F[Contextual Factors]
    
    C --> G[Sequence Modeling<br/>Transformer Architecture]
    D --> H[Preference Learning<br/>Deep Neural Networks]
    E --> I[Interest Prediction<br/>Graph Neural Networks]
    F --> J[Context Encoding<br/>Attention Mechanisms]
    
    G --> K[GPT-4 Intent Understanding]
    H --> L[Gemini Style Profiling]
    I --> M[Glood User Embedding]
    J --> M
    
    K --> N[Ensemble Prediction]
    L --> N
    M --> N
    
    N --> O[Personalized Rankings]
    O --> P[Real-time Filtering]
    P --> Q[Final Recommendations]
    
    style Q fill:#9f9,stroke:#333,stroke-width:2px
```

## Core AI Technologies

### 1. Deep User Profiling

#### Behavioral Embedding

Our system creates a 512-dimensional user embedding that captures:

* **Purchase History**: What, when, and how often they buy
* **Browse Patterns**: Categories explored, time spent, depth of exploration
* **Interaction Signals**: Clicks, adds to cart, wishlist saves
* **Temporal Patterns**: Shopping frequency, seasonal preferences

```python theme={null}
# Conceptual user embedding
user_embedding = concatenate([
    purchase_encoder(purchase_history),
    browse_encoder(session_data),
    temporal_encoder(time_patterns),
    preference_encoder(explicit_signals)
])
```

### 2. OpenAI GPT-4: Intent & Context Understanding

GPT-4 analyzes user behavior to understand:

#### Shopping Intent Detection

* **Browsing vs Buying**: Distinguishes research from purchase intent
* **Gift Shopping**: Identifies when users shop for others
* **Occasion Detection**: Recognizes event-driven shopping (holidays, birthdays)
* **Need Urgency**: Understands immediate vs future purchase plans

#### Natural Language Signals

* Search query understanding
* Review sentiment analysis
* Support conversation context
* Email engagement patterns

#### Persona Development

GPT-4 helps build rich user personas:

* "Tech Early Adopter"
* "Value-Conscious Family Shopper"
* "Luxury Fashion Enthusiast"
* "Sustainable Living Advocate"

### 3. Google Gemini: Visual Preference Learning

Gemini analyzes visual preferences through:

#### Style Profiling

* **Color Preferences**: Tracks color choices across purchases
* **Design Aesthetics**: Modern, vintage, minimalist, maximalist
* **Pattern Affinity**: Stripes, florals, geometric, solid
* **Brand Visual Language**: Preference for certain visual identities

#### Visual Behavior Analysis

```python theme={null}
visual_preference = gemini.analyze({
    'viewed_products': product_images,
    'purchased_products': purchase_images,
    'engagement_time': image_interaction_time,
    'zoom_behavior': detailed_view_patterns
})
```

### 4. Glood Proprietary Models: Predictive Personalization

Our custom deep learning models provide:

#### Sequential Pattern Mining

* **LSTM Networks**: Capture long-term dependencies
* **Attention Mechanisms**: Focus on relevant history
* **Temporal Convolutions**: Identify seasonal patterns

#### Collaborative Filtering Enhanced

* **Neural Collaborative Filtering**: Deep learning on user-item interactions
* **Graph Neural Networks**: Leverage user-product-category relationships
* **Matrix Factorization++**: Advanced latent factor models

## Multi-Model Fusion Strategy

### Ensemble Architecture

```mermaid theme={null}
graph LR
    A[User Event] --> B[Model Router]
    B --> C[Cold Start?]
    C -->|Yes| D[Popularity + Gemini Visual]
    C -->|No| E[Full Model Stack]
    
    E --> F[GPT-4<br/>30% weight]
    E --> G[Gemini<br/>25% weight]
    E --> H[Glood DNN<br/>45% weight]
    
    F --> I[Weighted Combination]
    G --> I
    H --> I
    D --> I
    
    I --> J[Final Scores]
```

### Dynamic Weight Adjustment

Weights adjust based on:

* **Data Availability**: More history increases Glood model weight
* **Product Type**: Visual products increase Gemini weight
* **User Segment**: New users rely more on GPT-4 understanding
* **Context**: Mobile vs desktop, time of day, location

## Advanced Personalization Techniques

### 1. Real-time Session Personalization

The algorithm adapts within a single session:

* **Interest Drift Detection**: Recognizes when users explore new categories
* **Micro-intent Modeling**: Updates recommendations after each click
* **Attention Decay**: Recent actions weighted higher than older ones

### 2. Cross-Device Personalization

Unified experience across devices:

* **Device Preference Learning**: Different behaviors on mobile vs desktop
* **Context Switching**: Work computer vs home browsing
* **Session Continuation**: Picks up where user left off

### 3. Temporal Dynamics

Time-aware personalization:

* **Seasonal Adjustments**: Summer clothes in spring, holiday items in November
* **Purchase Cycles**: Predicts when consumables need replenishment
* **Life Event Detection**: Identifies major life changes affecting preferences

### 4. Exploration vs Exploitation

Balancing familiar and novel:

```python theme={null}
recommendations = (
    exploitation_weight * known_preferences +
    exploration_weight * diverse_discoveries
)
```

## Personalization Levels

### Level 1: Anonymous Visitors

* Popularity-based recommendations
* Session-based personalization
* Geographic and device signals

### Level 2: Known Non-Purchasers

* Browse history personalization
* Email engagement signals
* Wishlist and cart analysis

### Level 3: Single Purchase Customers

* Purchase-based profiling
* Category expansion recommendations
* Complementary product suggestions

### Level 4: Repeat Customers

* Full behavioral modeling
* Predictive replenishment
* Loyalty-aware recommendations

### Level 5: VIP Customers

* Hyper-personalized experiences
* Early access to relevant products
* Exclusive recommendation sets

## Privacy-Preserving Personalization

### Data Protection Measures

* **Differential Privacy**: Adding noise to protect individual data
* **Federated Learning**: Models trained without centralizing data
* **Encryption**: All personal data encrypted at rest and in transit
* **Anonymization**: Behavioral patterns separated from PII

### User Control

* Opt-out capabilities
* Recommendation transparency
* Data deletion rights
* Preference management

## Performance Metrics

### Engagement Metrics

* **CTR Improvement**: 3.5x higher than non-personalized
* **Conversion Lift**: 42% increase in purchase probability
* **Session Duration**: 65% longer engagement time
* **Return Visits**: 2.3x more likely to return

### Quality Metrics

* **Relevance Score**: 4.6/5 user rating
* **Diversity Index**: 0.73 (optimal range 0.7-0.8)
* **Novelty Factor**: 23% new discoveries
* **Satisfaction Rate**: 89% positive feedback

## Real-World Examples

### Fashion Personalization

```yaml theme={null}
User Profile: 28-year-old urban professional
Recent Activity: Viewed sustainable fashion brands
Personalized Recommendations:
  - Eco-friendly work attire
  - Sustainable accessories from viewed brands
  - Similar items in preferred color palette
  - New arrivals matching style profile
```

### Electronics Enthusiast

```yaml theme={null}
User Profile: Early adopter, premium segment
Recent Activity: Researched smart home devices
Personalized Recommendations:
  - Latest smart home innovations
  - Compatible ecosystem products
  - Premium alternatives to viewed items
  - Pre-orders for upcoming releases
```

### Home Decorator

```yaml theme={null}
User Profile: DIY enthusiast, modern aesthetic
Recent Activity: Purchased minimalist furniture
Personalized Recommendations:
  - Complementary decor items
  - Similar style from other brands
  - Seasonal updates to existing themes
  - Project-based product bundles
```

## Continuous Learning

### Feedback Loops

1. **Implicit Signals**: Clicks, time spent, scroll depth
2. **Explicit Feedback**: Ratings, reviews, returns
3. **Purchase Validation**: What actually converts
4. **Long-term Value**: Repeat purchase patterns

### Model Evolution

* **Online Learning**: Real-time model updates
* **A/B Testing**: Continuous algorithm improvement
* **Transfer Learning**: Leverage learnings across segments
* **Meta-Learning**: Learning how to learn from new users faster

## Glood Dashboard Configuration

### Setting Up PFY in Glood

1. **Navigate to AI Recommendations**
   * Log into your Glood Dashboard
   * Click "AI Recommendations" → "Create Section"
   * Select "Personalized For You" as the recommendation type

2. **Configure PFY Settings**
   ```yaml theme={null}
   Section Name: "Recommended For You"
   Number of Products: 8
   Personalization Level: High
   Fallback Strategy: Trending Products
   Filter Out of Stock: Yes
   Price Range: Adaptive
   ```

3. **Advanced Controls**
   * **Exploration Rate**: Balance between familiar and new (0-100%)
   * **Category Diversity**: Enforce variety across categories
   * **Brand Mix**: Control single-brand dominance
   * **Recency Bias**: Weight recent interactions higher

### Manual Overrides

Even with PFY, you maintain control:

* **Pinned Products**: Force specific products to always appear
* **Blacklisted Items**: Never show certain products
* **Boost Categories**: Increase weight for specific collections
* **Seasonal Adjustments**: Schedule different strategies

## Glood's Technical Infrastructure

### How Glood Processes PFY

* **Real-time Processing**: Sub-100ms personalization
* **Global CDN**: Recommendations served from edge locations
* **Smart Caching**: Balance between fresh and fast
* **Auto-scaling**: Handles Black Friday traffic automatically

### Glood API Response

```json theme={null}
{
  "shop_id": "your-store.myshopify.com",
  "customer_id": "cust_123456",
  "section_id": "pfy_main",
  "recommendations": [
    {
      "product_id": "7234567890",
      "title": "Organic Cotton T-Shirt",
      "score": 0.92,
      "reason": "Based on your sustainable fashion preferences",
      "glood_confidence": "high",
      "personalization_signals": [
        "viewed_similar",
        "category_affinity",
        "price_match"
      ]
    }
  ],
  "generated_at": "2024-01-15T10:30:00Z",
  "cache_ttl": 300,
  "fallback_used": false
}
```

## Glood Performance Metrics

### PFY Success Stories

* **Fashion Retailer**: 47% increase in conversion rate
* **Electronics Store**: 3.2x higher AOV with PFY
* **Home Goods**: 62% of revenue from PFY recommendations
* **Beauty Brand**: 89% customer satisfaction score

### Monitor in Glood Analytics

* Real-time conversion tracking
* Customer segment performance
* Revenue attribution by personalization level
* A/B test results dashboard

## Best Practices for Glood PFY

1. **Quick Setup**
   * Use Glood's one-click PFY template
   * Start with default settings, optimize later
   * Enable on high-traffic pages first

2. **Data Quality**
   * Ensure Glood pixel is properly installed
   * Sync your full product catalog
   * Enable customer account tracking

3. **Optimization**
   * Use Glood's built-in A/B testing
   * Monitor the analytics dashboard weekly
   * Adjust exploration rate based on results

4. **Customization**
   * Match PFY design to your theme
   * Use Glood's template editor
   * Configure mobile-specific settings

For step-by-step setup, see our [Glood PFY Implementation Guide](/guides/create-sections/how-to-create-pfy-section).