Einstein Product Recommendations powers cross-sell / upsell on Salesforce Commerce Cloud storefronts — "Customers who viewed this also bought," "Recommended for you," "Frequently bought together." Auto-trained on shopper behavior; merchandisers configure WHERE recommendations appear, the model decides WHICH products.
- Confirm Commerce Cloud + Einstein for Commerce licensing
Setup → Einstein Setup. Product Recommendations is part of Commerce Cloud Einstein.
- Open Commerce Cloud Business Manager → Einstein → Configuration
Commerce Cloud has its own admin (Business Manager). Different from CRM Setup.
- Tick Enable Einstein Recommendations
Activates the model training pipeline.
- Configure recommendation types per page placement
Product Detail Page: "Customers Who Bought." Cart Page: "Frequently Bought Together." Category Page: "Top Trending."
- Configure exclusion rules
Categories / brands / product attributes to exclude from recommendations (e.g., don't recommend competitor brands, don't recommend out-of-stock).
- Wait for model training
Initial model uses category fallbacks until enough behavior data accumulates.
- Embed via Commerce Cloud page templates
Storefront templates have placeholders for recommendation tiles. Confirm placement and styling on the page templates.
Customers Who Bought / Frequently Bought Together / Trending / Personalized.
PDP / Cart / Category / Order Confirmation.
Category / brand / attribute filters.
Fallback for new shoppers without history.
- Recommendations need shopper behavior data to be useful. New storefronts use category-based fallbacks; meaningful personalization requires weeks-to-months of clickstream / purchase data accumulation.
- Privacy regulations (GDPR, CCPA) affect recommendation eligibility. Cookie consent + opt-out signals should gate which shoppers get personalized recommendations vs generic.
- Recommendations can be wrong / awkward (recommending a product the customer just returned). Pair with exclusion rules and A/B testing to refine.