Develop Feature-Specific Plans

Learning Objectives

After completing this unit, you’ll be able to:

• Describe the tasks a developer must take to implement Predictive Sort and Product Recommendations.
  
• List the features that you can implement with minimal effort.
  
• List the pre-launch steps you must take to implement Product Recommendations and Predictive Sort.
  
• List the post-launch steps a developer must take to implement Product Recommendations.
  

Introduction

To customize Einstein to your business needs, you implement specific features in Commerce Cloud Einstein. Here’s how feature choices can impact implementation.

Commerce Insights requires just data feeds and a login to Configurator.

These features need a bit more configuration to implement but are well worth the extra effort.

• Search Dictionaries
  
• Search Recommendations
  
• Predictive Sort
  
• Product Recommendations
  

Search Dictionaries

Einstein generates search dictionaries based on these factors.

• The existing synonym groups for a site. If there are no synonym groups, then there are no Einstein Search Dictionaries synonym suggestions.
  
• The amount of traffic going to the site and reaching the no search results page.
  

You have the option to accept the Einstein Data Privacy Agreement and anonymously share and receive synonym suggestions based on B2C Commerce network search and synonym group data. This data is used to provide recommendations to other merchants who have also opted into the shared database. Salesforce ensures that the data will not be disclosed in an identifiable form to other merchants. Additionally, Salesforce may access the contributed data to train and improve the Einstein Search Dictionary and related features or services. If you do not wish to accept the agreement, Einstein uses site data only.

You can add email addresses to the Business Manager search preferences (on the staging instance only) to notify your merchandising team of search terms to approve or decline.

Search Recommendations

Although it’s tempting to want to implement all the Search Recommendations features at once, consider implementing them in increments. Start with the low-effort features that don’t require storefront customizations, and then add the other feature.

Here’s a typical implementation plan.

1. Enable Einstein Search Recommendations. The machine learning algorithm starts to consume search queries and identify search phrases to recommend. Einstein begins to show search suggestions based on actual shopper searches.
   
2. With help from a developer, decide in which locations you want to put Einstein Search Recommendations. Extend the type-ahead search flyout (expanding search box) to render multiple search suggestions (versus the default single suggestion). This requires a customization because it isn’t available out of the box.
   
3. Once the design is ready, you can add add these Einstein Search recommendation options:
   

• Implement recent search phrases by modifying the type-ahead search flyout menu to render a personalized list of search phrases entered by shoppers.
  
• Add popular search phrases to the recent search phrases customization.
  
• Add popular storefront searches.
  
• Add recent personal searches.
  

4. Be sure to replicate search preferences to production when this is ready to go live on your site
   

When developing the flyout, decide what appears on the flyout and the maximum number of results that display for each element, such as:

• Business view
  
• Search for/Did You Mean
  
• Popular searches
  
• Brands
  
• Category
  
• Content
  
• Recently viewed
  

Predictive Sort

Predictive Sort personalizes the order in which products are displayed to customers. To implement Predictive Sort, your developer validates that caching correctly invalidates when they apply Predictive Sort as a sorting rule. This is a critical prerequisite to ensure personalization works correctly. Once they confirm that, your merchandiser can use Business Manager to copy an existing sorting rule and blend Predictive Sort into a dynamic or static sorting rule that meets your business needs. A/B Testing offers the best way to measure predictive sorting.

Product Recommendations

The Product Recommendations feature also requires a team effort.

To implement this feature, the team plans to take the following steps.

• Add the Product Recommendations feature to the product detail page via a content slot associated with a recommender.
  
• Create the recommenders in Configurator.
  
• Configure the content slots in Business Manager to select the specific recommender created to display the product recommendations.
  

Start with the default product detail page (PDP) recommender, which is automatically available in a Business Manager content slot configuration after Einstein deployment. This recommender lets them personalize the product details page through machine learning algorithms that include:

• View to view correlations
  
• Product affinities
  
• Natural language processing
  

Use preview and the validator tool to make sure there are no issues.

Basic Schedule

Here is how you implement Product Recommendations.

Merchants can go live with Search Recommendations in parallel with their site go-live because of Einstein's use of natural language processing.

Multi-Product Implementation

To implement Product Recommendations and Predictive Sort at the same time.

Next Steps

In this unit, You learned about the steps required to implement specific Commerce Cloud Einstein features. In the next unit, you enable the privacy agreement, install the Chrome Extension, and run the deployment in Business Manager.

Resources

• Salesforce: B2C Commerce Einstein API
• Salesforce: Salesforce Trailblazer Community