Adding Product AI Recommendations to Your Email

Product AI Recommendations Overview

With Yotpo Email’s advanced algorithms and Artificial Intelligence (AI) technology, you can show each customer highly personalized product recommendations, best sellers, new arrivals, and more. Fine-tune these recommendations by including or excluding certain collections, excluding certain products, defining the price range, and applying more filters to ensure they are on target. Use the Product AI Recommendations block in email campaigns and flows, especially in upsell flows, so customers who purchased a product in the past can automatically receive an email with tailored product recommendations.

Setting up Your Product AI Recommendations Block

Getting Started

To add the Product Recommendations blocks to your email, first open the email editor.

• Email campaigns

• Multi-message campaigns

• Email templates

• Flows

After opening the editor:

1. Navigate to the Content tab

2. Expand the Blocks menu and drop the Product Recommendations block into your email.

3. Choose the type of your recommendation block:

   • Personalized recommendations:We recommend products to shoppers based on what they’ve bought in the past. Our system looks at how often people who bought one product also bought another, and gives each product pair a score. Then, based on the shopper’s purchase history, we use those scores to suggest products they’re most likely to want next.

   • Non-personalized recommendations: These recommendations highlight popular products in your store, based on overall shopping trends and buying patterns. Every shopper sees the same list, making it a great way to showcase your store’s top-performing items.

     • Best sellers: Products with the most purchases over the last year

     • Newest products: Recently added items in your store

     • Highest-priced products: Premium products with the highest prices in the last year

     • Lowest-priced products: Budget-friendly items with the lowest prices in the last year

     • Most-reviewed products: Items with the highest number of reviews over the past year

     • Most-liked products: Products with the most positive review sentiment in the last year

4. Turn on the toggle next to Customize recommendations to create rule-based filters to refine the product selection.

   • Click Create filter, and in the modal that appears, enter the filter’s name.

   • Add the filter’s rules. There are five types of rules (Collection, Tag, Product, Price, and Stock level), and each can be used in a filter only once.

     • Include / exclude collections

     • Exclude products

     • Include / exclude tags

     • Price range

     • Stock level range

5. Click Save and apply to email to complete the setup.  

You will see a summary of the newly created filter in the block settings.

Note

You can add more than one Product Recommendations block per email (however, only 1 block per any type is allowed). Each block can have a different user-defined filter.

Understanding Product AI Recommendations

Our Product AI Recommendations are split into two main types: Personalized Recommendations and Non-personalized Recommendations.

Personalized Recommendations

We recommend products to shoppers based on what they’ve bought in the past. Our system looks at how often people who bought one product also bought another, and gives each product pair a score. Then, based on the shopper’s purchase history, we use those scores to suggest products they’re most likely to want next.

Non-personalized Recommendations

These recommendations highlight popular products in your store, based on overall shopping trends and buying patterns. Every shopper sees the same list, making it a great way to showcase your store’s top-performing items.

• Best sellers: Products with the most purchases over the last year

• Newest products: Recently added items in your store

• Highest-priced products: Premium products with the highest prices in the last year

• Lowest-priced products: Budget-friendly items with the lowest prices in the last year

• Most-reviewed products: Items with the highest number of reviews over the past year

• Most-liked products: Products with the most positive review sentiment in the last year

Applying Filters to the Recommendations

To make sure your recommendations are on target, you can fine-tune them by leveraging the following filters/criteria:

• Including or excluding certain collections

• Exclude certain products

• Include or exclude certain tags of products

• Define the price range of selected products

• Define the stock level of selected products

When you apply filters to your personalized product recommendations, here’s what happens behind the scenes:

1. We start by pulling a list of the top 100 most purchased products from your catalog that match your filters.

2. Then, we send the following information to our system :

   • The 100 filtered products

   • How many products you'd like to show

   • The type of recommendation you're using (like personalized, best sellers, most liked, or most reviewed)

   • The time range (last year)

Generating Recommendations

Our model gets to work by first creating a broader list—about 3x more products than needed. This gives it room to refine the results and make them more relevant.

Next, it narrows things down by:

• Removing products that are out of stock

• Applying filters on the result, such as remove out of stock items, remove excluded items

• Taking out products the shopper has purchased any time in the past

• Retaining products only from the initial eligible list

This ensures your final recommendations are relevant and aligned with your strategy.

Generating Recommendations without Filters

When no filters are applied, the recommendation process follows a straightforward approach. We send the following inputs to our system :

• The number of products to display

• How many products you'd like to show

• The type of recommendation you're using (like personalized, best sellers, most liked, or most reviewed)

• The time range (last year)

Recommendation and Filtering Process

The model generates a preliminary list that includes approximately 3x more products than needed.

This list is then refined through several filtering steps:

• Removing out-of-stock products

• Applying filters on the result, such as remove out of stock items, remove excluded items

• Removing items the shopper has already purchased

Smart Product Recommendations

To help you boost conversions and grow your revenue, we make sure shoppers don’t see the same product recommended twice within 6 months. This keeps your recommendations fresh, relevant, and more likely to drive clicks and purchases.

Good to Know

You don’t need to do anything to activate this, our system handles it automatically, no matter which filters you use.

Recommendation Fallbacks

In the event our system can’t find any recommendations, shoppers will still see great products.

We’ve built in several fallbacks to keep your product suggestions flowing:

• Too few recommendations: the system automatically adds products  your best sellers  from the last 30 days.

• No results are returned or an error occurs:  the system will show active products from your store, sorted by their average review scores.

In both cases, any filters you’ve set will still apply. This way, your shoppers continue to see relevant, high-quality items, no matter what.

Customizing the Product AI Recommendations Block

The Product AI Recommendations block offers the same customization options as the Products block, with a few additional settings:

• Stack on mobile:

  • Turned on ( default option): Products will be arranged vertically on mobile devices.

  • Turned off: Products will be arranged side-by-side on mobile devices (in a single row)

• Orientation: Define if your products display side-by-side or vertically

• Products in one row: Define how many products you display in a row (up to 4)

• Max products to show: You can choose to show up to 20 products

• Exclude recent purchases: The block will not show any products that the customer has ever purchased, this toggle is enabled by default

• Background color: Define the background color of the block

• Border: Change or remove the border of the block

• Product card composition

  • Image: Display or hide the product image

  • Name: Display or hide the product name, and define the text's color, style, alignment, etc.

  • Price: Display or hide the product price, and define the price text's color, style, alignment, etc.

  • Show original price: Select this option to include the product's original (pre-discount) price. This is the Compare-at price value in Shopify. (by default turned on)

  • Show 'Sale' label: Adds a 'Sale' label to the product. (by default turned on)

• Rating: Display or hide the product rating. Select the minimum rating for displaying the rating, define the text's color, style, alignment, etc. By default the minimum rating is set to 4.

• Button: Adds a Buy button to the product that links to your store (customize button label, text font style and size, button color, text color, alignment and adjust to width. Adjust to Width is enabled by default)

Sending Test Emails

If a filter is defined, the visualization of the Product Recommendations block in test emails and email previews should follow the selected rules. However, recommendations will differ from the actual recommendations because test emails use sample data that is not based on actual shopper data.

Email Previews (Beta)

If you would like to view an email as a specified shopper, you can do so in an email campaign by previewing the email as the shopper you wish to see.

Note

The recommendations block  will appear in flow email previews with random data and differ from actual emails being sent to your shoppers.

Recommendation Block Duplication

Inside the store (as part of a campaign/flow/template duplication)

When you duplicate a campaign, flow, or template containing the product recommendations block, the settings of the block, including Customize recommendations settings, will be duplicated.

Across Stores

When cloning between stores, only the recommendation type will be cloned as designed. However, if the Customize recommendations setting is turned on and filters are defined,  the setting will not be automatically turned on,   and the defined filters will not be duplicated.