Commerce Cloud Einstein
Commerce Cloud Einstein is the AI feature set inside Salesforce Commerce Cloud that drives product recommendations, predictive search, personalized sort, content recommendations, and commerce insights for B2C and B2B storefronts.
Definition
Commerce Cloud Einstein is the AI feature set inside Salesforce Commerce Cloud that drives product recommendations, predictive search, personalized sort, content recommendations, and commerce insights for B2C and B2B storefronts. The features run on shopper behavior data (clicks, views, carts, purchases) plus catalog metadata, surface inside the standard storefront templates without custom code, and feed analytics back into Commerce Cloud reports. The point is to raise conversion and average order value through personalization the merchandiser would not have time to build by hand.
Commerce Cloud Einstein is not one product but a bundle that ships with Commerce Cloud at most editions. The most-used features are Einstein Product Recommendations (cross-sell, also-bought, recently-viewed), Einstein Predictive Sort (reorders category listings by per-shopper relevance), and Einstein Search Recommendations (suggests searches and ranks results). The newer Commerce Agentforce builds on the same data foundation to add conversational shopping. Together they form the personalization layer most B2C storefronts run today.
Why Commerce Cloud Einstein is a bundle, not a feature
The feature bundle and what each component does
The bundle has six main features. Einstein Product Recommendations powers the "Customers who bought this also bought" and "Recently viewed" carousels across the site. Einstein Predictive Sort reorders category listing pages so the products most likely to interest the current shopper rise to the top. Einstein Search Recommendations completes search queries as the shopper types and ranks results by personal relevance plus business rules. Einstein Content Recommendations surfaces blog posts, guides, and editorial content tied to the shopper interest. Einstein Commerce Insights surfaces business intelligence on shopper behavior and merchandising effectiveness. Commerce Agentforce wraps the data foundation in a conversational shopping agent.
How the underlying data foundation works
All Einstein features run on shopper behavior streams plus catalog metadata. Behavior streams include page views, clicks, add-to-cart events, purchase events, search queries, and account interactions. Catalog metadata includes product attributes, category trees, prices, inventory, and merchandising tags. The data is ingested in near real time, anonymized for cross-shopper modeling, and tied back to known shopper IDs for personal modeling. The model training cadence varies by feature; some features retrain hourly on fresh behavior data, others daily. A new storefront needs a minimum of 30 days of behavior data before Einstein features start producing useful recommendations.
Einstein Product Recommendations in depth
Product Recommendations is the workhorse feature. It serves multiple recommender strategies: Recently Viewed, Customers Also Viewed, Customers Also Bought, Complete the Set, Top Sellers, and Recommended for You. Each strategy uses a different algorithm under the hood (collaborative filtering, content-based, sequence-aware). Merchandisers pick which strategy fits which placement in Business Manager. The same product detail page can host three recommenders in three carousels (Also Bought, Recently Viewed, Top in Category) without code changes. Business rules override the algorithm for compliance reasons; out-of-stock items are excluded by default, and merchandiser-pinned items take priority over algorithmic picks.
Predictive Sort and the merchandising trade-off
Predictive Sort reorders category listings on a per-shopper basis. The shopper who looked at running shoes last week sees running shoes near the top of the Footwear category; another shopper sees boots. The merchandiser controls how aggressively the algorithm overrides their pinned ordering. Three modes: Off (manual order only), Hybrid (top three slots pinned, rest algorithmic), Full (algorithmic with merchandiser overrides for compliance only). Most merchandising teams settle on Hybrid because it preserves brand storytelling at the top of the page while giving personalization the long tail. Full mode often produces uplift in raw conversion at the cost of brand voice control.
Search Recommendations and synonym tuning
Search Recommendations works on three layers. Type-ahead suggestions complete queries as the shopper types. Result ranking applies a personalized score on top of the standard text relevance score. Synonym tuning learns from queries that returned no results and suggests synonyms the merchandiser can review and approve. The synonym suggestions are the most underused feature; pulling the weekly Einstein Search Insights report and approving the top 10 synonym suggestions takes 20 minutes and meaningfully improves zero-result rates within a week. Search Recommendations does not replace the underlying search engine; it augments ranking and discovery.
Commerce Insights and proving merchandising hypotheses
Commerce Insights is the analytics surface for the Einstein bundle. Default dashboards cover bounce rates by landing page, conversion funnel by traffic source, top abandoned products in carts, and recommendation acceptance rate per placement. The dashboards expose the data needed to A/B test merchandising decisions: which homepage carousel converts better, which category sort mode produces higher AOV, which recommender strategy belongs in which placement. The data is most useful when paired with a test framework that controls for traffic source and time of day; raw conversion numbers can mislead when not stratified.
Commerce Agentforce and the conversational shopping layer
Commerce Agentforce is the newer addition to the bundle. It is a conversational shopping agent grounded in the storefront catalog and the shopper conversation history. Shoppers ask questions ("I need a dress for a beach wedding under $200") and the agent recommends products with reasoning, can refine based on feedback, and can complete the purchase if the storefront has agent checkout enabled. Commerce Agentforce uses the Einstein data foundation for grounding and the Atlas Reasoning Engine for the conversation. It is licensed per conversation through the Agentforce SKU rather than bundled with Commerce Cloud. Most retailers evaluating it pilot on a specific category before broad rollout.
How to get value out of Commerce Cloud Einstein on a real storefront
Most teams turn on Einstein features in week one and forget about them. The teams that get sustained value pick specific KPIs per feature, measure baselines, run controlled tests of placements and modes, and revisit weekly. The features pay back; the discipline of measuring them is what separates real lift from theoretical.
- Confirm prerequisites and licensing
Einstein is bundled with most Commerce Cloud editions. Confirm your edition in Business Manager. Commerce Agentforce is a separate Agentforce SKU. Without 30 days of shopper behavior data, Einstein features run on global defaults that are weaker than personalized models.
- Enable Einstein in Business Manager
Business Manager, Administration, Site Development, Einstein Configuration. Enable the bundle. Pick which features to activate first; most teams start with Product Recommendations and Predictive Sort, add Search Recommendations after a month, and evaluate Commerce Agentforce once the core features are tuned.
- Place recommender carousels in your storefront templates
The storefront templates ship with recommender carousel components. Add them to product detail, cart, homepage, and category pages. Pick strategy per placement (Also Bought on PDP, Recently Viewed on cart, Top in Category on category landing).
- Pick a Predictive Sort mode per category
Hybrid is the safest default. Test Full on long-tail categories where personalization has the most room to lift. Keep Off on categories with strong brand storytelling reasons for manual order.
- Wire Search Recommendations and approve the first synonym batch
Turn on Search Recommendations. Open Einstein Search Insights, sort by zero-result query frequency, and approve synonyms for the top 10 to 20. Repeat weekly.
- Build a weekly Commerce Insights review
Schedule 30 minutes weekly with the merchandising lead. Review recommendation acceptance rate, search zero-result rate, predictive sort lift by category. The review drives the next week's tuning.
- Test Commerce Agentforce on one category before broad rollout
Pick a high-consideration category (electronics, gifting, fashion) where conversational guidance helps. Pilot for four weeks. Compare conversion against a control before broad rollout.
Which Einstein features are enabled (Product Recommendations, Predictive Sort, Search Recommendations, Content Recommendations, Commerce Insights, Commerce Agentforce). Most teams enable in waves.
Which algorithm (Also Bought, Recently Viewed, Top Sellers, etc.) drives each carousel. Configurable in Business Manager without code.
Off, Hybrid, or Full. Drives how aggressively the algorithm overrides merchandiser pinned order.
Merchandiser-curated synonyms based on Einstein suggestions from zero-result queries. Highest-ROI underused feature.
Compliance and merchandising rules that override the algorithm (exclude out-of-stock, pin sale items, exclude restricted products). Configurable per feature.
- Einstein features need a minimum of 30 days of shopper behavior data to produce useful personalization. New storefronts run on global defaults that often underperform manual merchandising.
- Predictive Sort Full mode can boost raw conversion at the cost of brand storytelling control. Most teams settle on Hybrid for a reason.
- Search synonym suggestions accumulate quickly if no one reviews them. The weekly approval cadence is the discipline that turns the feature on; turning it on without the cadence wastes the data.
- Recommendation acceptance rate per placement varies widely. A carousel that converts on PDP can fail on cart. Test each placement individually rather than assuming the recommender works equally well everywhere.
- Commerce Agentforce is licensed separately from the Einstein bundle. Confirm SKU before assuming it ships with your Commerce Cloud edition.
Trust & references
Cross-checked against the following references.
- Salesforce Commerce CloudSalesforce
- Einstein for Commerce Cloud referenceSalesforce
Straight from the source - Salesforce's reference material on Commerce Cloud Einstein.
- Einstein for Commerce CloudSalesforce Help
- Einstein Product RecommendationsSalesforce Help
- Einstein Predictive SortSalesforce Help
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What does Commerce Cloud Einstein do?
Q2. What does Einstein use to learn shopper preferences?
Q3. What is Open-Time Email Personalization?
Discussion
Loading discussion…