Product Recommendations Playbook

A complete guide to setting up AI-powered recommendations. Learn how to configure the wizard, implement interaction tracking, and deploy high-converting recipes for every page type.

Start Now
Start Here
Wizard & Interaction Tracking
Widget Design
Badges & Dynamic Content
Advanced Filtering
Dynamic Rules & Logic
Homepage
Smart & Personalized
Category & Discovery
Contextual Browsing
Product Page
Cross-Sell & Recent
Cart & Checkout
Cart-Based Upsell
Retention
New Items & Reorder
A/B Testing
Compare Algorithms
Performance
Revenue & Analytics

Start with the Recommendation Wizard

From the Personyze dashboard, click "Product Recommendation" to launch the setup wizard.

Step 1: Product Interaction Tracking

You must tell Personyze when a user views, buys, or adds an item to the cart. You have 3 methods:

  • Site Containers: Use the Visual Selector to click on the SKU or Price element on your page.
  • Google Tag Manager: Set a trigger for Data Layer events (e.g., event equals "add_to_cart").
  • Programmatic JS: Use the exact code below. You can use Internal ID or SKU as the key.
Tracking Code (Copy & Paste):
(self.personyze=self.personyze||[]).push(['Product Viewed', "PRODUCT_ID"]);

(self.personyze=self.personyze||[]).push(['Product Added to cart', "PRODUCT_ID", 'Quantity', QUANTITY]);

(self.personyze=self.personyze||[]).push(['Product Removed from cart', "PRODUCT_ID"]);

(self.personyze=self.personyze||[]).push(['Products Removed from cart']); (Clear all)

Step 2: Targeting (Optional)

Define *where* this widget appears. You can set it to show only on the Homepage, or only for visitors from North America.

Widget Design & Customization

Customize the look and feel to match your brand exactly using the "Customize Look and Feel" step.

Dynamic Badges

Add visual labels to products to increase CTR:

  • Top Seller: Highlights popular items.
  • New: For recently added products.
  • Low in Stock: Creates urgency.
  • Recently Discounted: Shows price drops.

Advanced Content Logic

  • VIP Pricing: Display a specific price variable (e.g., vip_price) for logged-in VIP users.
  • Multi-Language: Load different Product Titles and Descriptions based on the user's selected language or geo-location.

Advanced Filtering

In the "Final Touches" step, apply rules to ensure recommendations are business-smart.

Dynamic Exclusion Rules

  • Exclude Purchased: Don't show items the user already bought.
  • Exclude Added to Cart: Prevent redundancy.
  • Exclude Current Page: Don't recommend the product they are currently viewing.

Smart "Match" Filters

Filter products based on User Data variables:

  • Brand Match: Only recommend products from a Brand the user has purchased in the past.
  • Gender/Age Match: Ensure product tags match the user's demographic profile.
  • Category Match: "If user is viewing Laptops, only recommend Laptop Bags."

Recipe 1: The Smart Homepage

Your homepage must appeal to everyone—both first-time visitors and loyal customers.

Smart Homepage (Hybrid)

Use the "Personalized Recommendations" algorithm. It automatically switches logic:

  • For Anonymous: Uses Most Popular (Crowd Wisdom) to show bestsellers.
  • For Known Users: Uses Deep Personalization based on:
    • User Data: Past purchases and category affinity.
    • Collaborative Filtering: "Users like you bought..."

This ensures relevance from the first click to the hundredth.

Recipe 2: Category & Discovery

Help users find what they are looking for by leveraging context.

1. Inspired by Last Viewed Category

Shows items from the category the user engaged with most recently. If they looked at "Drills" yesterday, show "Power Tools" on the homepage today.

2. Best Seller from Category

Show the top-selling items within the current category. If the user is browsing "Laptops", show the best-selling laptops, not shoes.

3. Most Popular from Category

Similar to Best Seller, but based on Views rather than Purchases. Good for trending items that haven't converted yet.

Recipe 3: Product Page Cross-Sell

Optimize the product detail page to keep users engaged and browsing.

1. Others Who Viewed Also Viewed

Collaborative Filtering: "People who looked at this shirt also looked at these." Great for providing alternatives if the main product isn't a perfect match.

2. What others Bought, who Viewed this

A stronger signal than just "viewed". It tells the user: "People who considered this item eventually bought *these* items." Excellent for closing sales.

3. Recently Viewed

A simple but effective history widget. Helps users compare options by easily navigating back to items they saw earlier.

Recipe 4: Cart Upsell & Recovery

The final push to increase basket size before payment.

Cross-Sells for Items in Cart

This powerful algorithm looks at the entire contents of the shopping cart to find items that complement the *combination* of products (e.g., Camera + Case -> Memory Card).

Managed Cross-Sells for Items in Cart

Allows manual override. "If Cart contains X, ALWAYS show Y." Useful for strict compatibility rules (e.g., Specific batteries for a specific toy).

Most Added to Wishlist

Leverage social proof by showing high-desire items. Often triggers impulse adds in the cart.

Recipe 5: Retention & Re-Engagement

Bring customers back with highly relevant news.

Past Orders Recommendations

Deep learning that analyzes a user's full purchase history to predict what they might need next.

Buy It Again

Perfect for consumables. It surfaces items the user has purchased before, streamlining the reordering process.

New in Stock (Filtered)

Show the newest items in your catalog, but apply a "Filter by Past Purchase Category" rule (e.g., A "Yoga" buyer sees new mats).

A/B Testing Recommendations

Don't guess which algorithm works best. Test them.

How to set up a test

  1. In the "Content" step of the wizard, enable A/B Testing.
  2. Variation A: Select "Collaborative Filtering" logic.
  3. Variation B: Select "Most Popular" logic.
  4. Goal: Set the metric to "Purchase Value" or "CTR".

Personyze will split traffic automatically and can eventually auto-allocate 100% of traffic to the winner.

Performance & Analytics

Measure the direct impact of your recommendations with detailed stats.

Example Dashboard Data

  • Shown: 49.06% (Users who saw the widget).
  • Viewed: 18.93% (Users who scrolled it into view).
  • Clicked: 45.56% (High engagement).
  • Added to Cart: 6.95% conversion to cart.
  • Purchased: 4.7% conversion to sale.

Revenue Impact Definitions

  • Direct Revenue: Revenue generated strictly from items that were clicked in the recommendation widget OR added via the widget's "Quick Add" button, and then purchased.
  • Assisted Revenue: Revenue from sessions where a user clicked a recommendation but ended up purchasing a different item.
  • Contribution Ratio: (e.g. 16.32%) The percentage of total site revenue that can be attributed to the recommendation engine.

Frequently Asked Questions

What is Collaborative Filtering?
Collaborative Filtering is the core technology behind recommendations like "People who bought this also bought that." It analyzes patterns across thousands of user sessions to find hidden correlations between products without needing manual rules.
How quickly does it learn?
Personyze starts collecting data immediately. Basic algorithms like "Most Popular" work instantly. Advanced algorithms like "Personalized Recommendations" typically need a few days of traffic (or a historical data upload) to build accurate user profiles.
Can I manually pin products to the top?
Yes. You can use "Managed Recommendations" or "Pins" to force specific items (like high-margin goods or overstock) to appear first, while letting the algorithm fill the remaining slots dynamically.
Does it work for content (articles/blogs)?
Absolutely. The engine works identically for content. Instead of "Products," it recommends "Articles" based on reading history, tags, and category affinity. You just need to sync your content feed instead of a product feed.
Can I exclude out-of-stock items?
Yes. By default, Personyze filters out items where stock_status = 0. You can customize this rule in the filtering step to also exclude items with low stock or specific tags.
How do I handle new users with no history?
This is the "Cold Start" problem. Personyze handles this by falling back to Crowd Logic (e.g., "Most Popular" or "Trending Now") or using Contextual Data (e.g., "Popular in your City") until the user generates their own history.
Can I design the widget myself?
Yes. You have full control over the HTML/CSS. You can use our visual editor to tweak existing templates or paste your own custom HTML structure to match your site's branding perfectly.
What is "Direct Revenue" vs "Assisted Revenue"?
Direct Revenue counts sales where the user clicked the specific recommended item and bought it. Assisted Revenue counts sales where the user clicked a recommendation but ended up buying *something else* in the same session.
Can I show different logic on Mobile?
Yes. You can create separate campaigns for Mobile and Desktop. For example, show a horizontal slider of 4 items on Mobile, but a grid of 8 items on Desktop.
Does it support multiple languages?
Yes. If your product feed contains localized fields (e.g., title_en, title_es), you can configure the widget to dynamically display the correct language based on the user's browser settings or site selection.
Can I recommend bundles?
Yes. The "Cross-Sells for Items in Cart" algorithm effectively creates bundles by suggesting complementary items (e.g., suggesting a Lens when a Camera is in the cart).
How often does the feed update?
Feeds typically sync once every 24 hours. However, you can configure more frequent updates (e.g., hourly) or use our API to push real-time stock/price changes instantly.
Can I test Algorithm A vs Algorithm B?
Yes. The A/B testing feature allows you to split traffic between two different logic sets (e.g., "Collaborative Filtering" vs. "Most Popular") to see which one generates more revenue.
Is it GDPR compliant?
Yes. Personyze anonymizes user data and respects "Do Not Track" requests. Personalization relies on pseudonymous IDs rather than PII unless explicitly provided by the user.
What if I have thousands of products?
Personyze scales effortlessly. Our engine is designed to handle catalogs with millions of SKUs and high-traffic sites without latency, ensuring recommendations load instantly.
Start Here
Wizard & Interaction Tracking
Widget Design
Badges & Dynamic Content
Advanced Filtering
Dynamic Rules & Logic
Homepage
Smart & Personalized
Category & Discovery
Contextual Browsing
Product Page
Cross-Sell & Recent
Cart & Checkout
Cart-Based Upsell
Retention
New Items & Reorder
A/B Testing
Compare Algorithms
Performance
Revenue & Analytics

Start with the Recommendation Wizard

From the Personyze dashboard, click "Product Recommendation" to launch the setup wizard.

Step 1: Product Interaction Tracking

You must tell Personyze when a user views, buys, or adds an item to the cart. You have 3 methods:

  • Site Containers: Use the Visual Selector to click on the SKU or Price element on your page.
  • Google Tag Manager: Set a trigger for Data Layer events (e.g., event equals "add_to_cart").
  • Programmatic JS: Use the exact code below. You can use Internal ID or SKU as the key.
Tracking Code (Copy & Paste):
(self.personyze=self.personyze||[]).push(['Product Viewed', "PRODUCT_ID"]);

(self.personyze=self.personyze||[]).push(['Product Added to cart', "PRODUCT_ID", 'Quantity', QUANTITY]);

(self.personyze=self.personyze||[]).push(['Product Removed from cart', "PRODUCT_ID"]);

(self.personyze=self.personyze||[]).push(['Products Removed from cart']); (Clear all)

Step 2: Targeting (Optional)

Define *where* this widget appears. You can set it to show only on the Homepage, or only for visitors from North America.

Widget Design & Customization

Customize the look and feel to match your brand exactly using the "Customize Look and Feel" step.

Dynamic Badges

Add visual labels to products to increase CTR:

  • Top Seller: Highlights popular items.
  • New: For recently added products.
  • Low in Stock: Creates urgency.
  • Recently Discounted: Shows price drops.

Advanced Content Logic

  • VIP Pricing: Display a specific price variable (e.g., vip_price) for logged-in VIP users.
  • Multi-Language: Load different Product Titles and Descriptions based on the user's selected language or geo-location.

Advanced Filtering

In the "Final Touches" step, apply rules to ensure recommendations are business-smart.

Dynamic Exclusion Rules

  • Exclude Purchased: Don't show items the user already bought.
  • Exclude Added to Cart: Prevent redundancy.
  • Exclude Current Page: Don't recommend the product they are currently viewing.

Smart "Match" Filters

Filter products based on User Data variables:

  • Brand Match: Only recommend products from a Brand the user has purchased in the past.
  • Gender/Age Match: Ensure product tags match the user's demographic profile.
  • Category Match: "If user is viewing Laptops, only recommend Laptop Bags."

Recipe 1: The Smart Homepage

Your homepage must appeal to everyone—both first-time visitors and loyal customers.

Smart Homepage (Hybrid)

Use the "Personalized Recommendations" algorithm. It automatically switches logic:

  • For Anonymous: Uses Most Popular (Crowd Wisdom) to show bestsellers.
  • For Known Users: Uses Deep Personalization based on:
    • User Data: Past purchases and category affinity.
    • Collaborative Filtering: "Users like you bought..."

This ensures relevance from the first click to the hundredth.

Recipe 2: Category & Discovery

Help users find what they are looking for by leveraging context.

1. Inspired by Last Viewed Category

Shows items from the category the user engaged with most recently. If they looked at "Drills" yesterday, show "Power Tools" on the homepage today.

2. Best Seller from Category

Show the top-selling items within the current category. If the user is browsing "Laptops", show the best-selling laptops, not shoes.

3. Most Popular from Category

Similar to Best Seller, but based on Views rather than Purchases. Good for trending items that haven't converted yet.

Recipe 3: Product Page Cross-Sell

Optimize the product detail page to keep users engaged and browsing.

1. Others Who Viewed Also Viewed

Collaborative Filtering: "People who looked at this shirt also looked at these." Great for providing alternatives if the main product isn't a perfect match.

2. What others Bought, who Viewed this

A stronger signal than just "viewed". It tells the user: "People who considered this item eventually bought *these* items." Excellent for closing sales.

3. Recently Viewed

A simple but effective history widget. Helps users compare options by easily navigating back to items they saw earlier.

Recipe 4: Cart Upsell & Recovery

The final push to increase basket size before payment.

Cross-Sells for Items in Cart

This powerful algorithm looks at the entire contents of the shopping cart to find items that complement the *combination* of products (e.g., Camera + Case -> Memory Card).

Managed Cross-Sells for Items in Cart

Allows manual override. "If Cart contains X, ALWAYS show Y." Useful for strict compatibility rules (e.g., Specific batteries for a specific toy).

Most Added to Wishlist

Leverage social proof by showing high-desire items. Often triggers impulse adds in the cart.

Recipe 5: Retention & Re-Engagement

Bring customers back with highly relevant news.

Past Orders Recommendations

Deep learning that analyzes a user's full purchase history to predict what they might need next.

Buy It Again

Perfect for consumables. It surfaces items the user has purchased before, streamlining the reordering process.

New in Stock (Filtered)

Show the newest items in your catalog, but apply a "Filter by Past Purchase Category" rule (e.g., A "Yoga" buyer sees new mats).

A/B Testing Recommendations

Don't guess which algorithm works best. Test them.

How to set up a test

  1. In the "Content" step of the wizard, enable A/B Testing.
  2. Variation A: Select "Collaborative Filtering" logic.
  3. Variation B: Select "Most Popular" logic.
  4. Goal: Set the metric to "Purchase Value" or "CTR".

Personyze will split traffic automatically and can eventually auto-allocate 100% of traffic to the winner.

Performance & Analytics

Measure the direct impact of your recommendations with detailed stats.

Example Dashboard Data

  • Shown: 49.06% (Users who saw the widget).
  • Viewed: 18.93% (Users who scrolled it into view).
  • Clicked: 45.56% (High engagement).
  • Added to Cart: 6.95% conversion to cart.
  • Purchased: 4.7% conversion to sale.

Revenue Impact Definitions

  • Direct Revenue: Revenue generated strictly from items that were clicked in the recommendation widget OR added via the widget's "Quick Add" button, and then purchased.
  • Assisted Revenue: Revenue from sessions where a user clicked a recommendation but ended up purchasing a different item.
  • Contribution Ratio: (e.g. 16.32%) The percentage of total site revenue that can be attributed to the recommendation engine.
AI Product Recommendations

Suggest the perfect product, every time

Stop guessing. Use Personyze’s intelligent recommendation engine to show the right products to the right shoppers. Boost AOV with "Frequently Bought Together" and AI-driven cross-selling strategies.

  • Live Video Demo
  • 30 Minutes
  • Expert Q&A
Expert Avatar
Dan from Personyze Online now • Replied just now
Hi there! 👋 Ready to see how AI recommendations can increase your cart size?
I can walk you through our different algorithms (like Upsell & Cross-sell) in about 30 minutes.
| Yes, schedule my walkthrough...