Personalization
Personalization re-ranks search results for each individual user based on their history, affinities, price sensitivity, size/color preferences, and behavioral patterns.
Overview
Personalization creates unique search experiences for each user by learning from their past behavior and preferences, then applying those learnings to rank products in a way that's most likely to resonate with them. The Marqo pixel automatically collects user interaction data (clicks, purchases, browsing behavior) from your site, so personalization works automatically without requiring any manual data submission.
Personalization Signals
User History
- Past purchases: Products the user has bought before
- Browsing history: Products the user has viewed
- Search history: Queries the user has searched
- Cart history: Products added to cart but not purchased
User Affinities
- Brand preferences: Brands the user prefers
- Category preferences: Categories the user shops most
- Price range: User's typical price range
- Style preferences: Colors, sizes, styles the user prefers
Behavioral Patterns
- Purchase frequency: How often the user purchases
- Session patterns: When and how the user shops
- Device preferences: Preferred device or channel
- Engagement level: How engaged the user is
Personalization Types
Purchase-Based Personalization
Rank products based on purchase history:
{
"personalization_type": "purchase_based",
"signals": [
"past_purchases",
"purchase_frequency",
"purchase_categories"
],
"weight": 0.4
}
Behavioral Personalization
Rank based on browsing and engagement:
{
"personalization_type": "behavioral",
"signals": [
"viewed_products",
"clicked_products",
"time_on_page",
"add_to_cart_rate"
],
"weight": 0.3
}
Preference-Based Personalization
Rank based on explicit and inferred preferences:
{
"personalization_type": "preference_based",
"signals": [
"brand_preferences",
"price_sensitivity",
"size_preferences",
"color_preferences"
],
"weight": 0.3
}
Personalization Models
Collaborative Filtering
Recommend products similar users liked:
{
"model_type": "collaborative_filtering",
"algorithm": "matrix_factorization",
"similarity_metric": "cosine",
"min_users": 10
}
Content-Based Filtering
Recommend products similar to user's history:
{
"model_type": "content_based",
"features": [
"category",
"brand",
"attributes",
"price_range"
],
"similarity_metric": "cosine"
}
Hybrid Approach
Combine collaborative and content-based:
{
"model_type": "hybrid",
"collaborative_weight": 0.6,
"content_based_weight": 0.4
}
Personalization Strength
Control how much personalization affects rankings:
{
"personalization_strength": {
"new_users": 0.1,
"returning_users": 0.3,
"vip_users": 0.5
}
}
Cold-Start Handling
Handle users with no history:
{
"cold_start_strategy": {
"use_popular_products": true,
"use_category_trends": true,
"use_demographic_segments": true
}
}
Privacy & Consent
Respect user privacy preferences:
{
"privacy_settings": {
"require_consent": true,
"allow_opt_out": true,
"anonymize_data": true
}
}
Personalization Examples
Example 1: Brand Preference
{
"user_id": "user_123",
"preferences": {
"preferred_brands": ["Nike", "Adidas"],
"boost_strength": 1.3
},
"ranking_adjustment": {
"boost_brands": ["Nike", "Adidas"],
"boost_strength": 1.3
}
}
Example 2: Price Sensitivity
{
"user_id": "user_456",
"preferences": {
"price_sensitivity": "high",
"typical_price_range": [50, 150]
},
"ranking_adjustment": {
"boost_price_range": [50, 150],
"boost_strength": 1.2
}
}
Example 3: Size Preference
{
"user_id": "user_789",
"preferences": {
"preferred_size": "large",
"preferred_color": "blue"
},
"ranking_adjustment": {
"boost_attributes": {
"size": "large",
"color": "blue"
},
"boost_strength": 1.15
}
}
Personalization Analytics
Track personalization effectiveness:
- Personalization lift: How much does personalization improve conversion?
- User satisfaction: Do users prefer personalized results?
- Diversity impact: Does personalization reduce diversity too much?
- Segment performance: How does personalization perform for different user segments?
Best Practices
- Start subtle: Begin with light personalization, increase gradually
- Balance relevance: Don't sacrifice relevance for personalization
- Maintain diversity: Ensure personalized results aren't too narrow
- Respect privacy: Always get user consent for personalization
- Test continuously: A/B test personalization approaches
- Monitor performance: Track how personalization affects key metrics
Related Topics
- Recommendations - Personalized recommendations