Product Recommendations is in closed beta and available only to selected clients. If you are a VTEX client and want to adopt it, contact Commercial Support. Additional fees may apply. For a product-focused introduction, see the Help Center tutorial Product Recommendations (beta).

Product Recommendations let your store surface relevant products through multiple strategies (for example, similar items, cross-sell, or personalized shelves). They can improve discovery, conversion, and average order value.

This page is the entry point for developers: it summarizes prerequisites to train models, lists strategies, points to implementation guides by storefront model, and includes frequently asked questions.

You need a storefront implemented with one of the following:

• Store Framework
• FastStore
• A headless storefront that can load scripts and call VTEX APIs
• A mobile app capable of making authenticated HTTPS requests

Request activation and the strategies you want enabled through Commercial Support. After approval, VTEX typically prepares your account by creating a recommendations workspace, syncing the catalog with the recommendation service, and issuing API access where needed.

Recommendation quality depends on catalog quality, tracked shopper behavior, and, for some strategies, order volume. Before models are trained and campaigns go live, your integration should:

• All storefronts must have cart and order event capture configured. Store Framework handles click/view events through the vtex.recommendation-shelf app. For other events in Store Framework, or all events in FastStore, headless, and mobile, configure Activity Flow as described in the implementation guides.
• FastStore, headless, and mobile must also call the Start session endpoint to link browsing behavior to completed purchases.
• All channels should send the product view events expected by the recommendation service so models can learn from real traffic.

Some strategies need minimum purchase history and product views so joint purchase patterns can be learned. Until those thresholds are met, VTEX may not train that strategy to full quality. See Recommendation strategies and FAQ.

The exact set of strategies enabled for your account is agreed during onboarding. Some strategies require enough sales or behavioral history before models perform well. Until then, rely on strategies that do not depend on those signals or tune shelves with your VTEX contact.

Strategy	How it works	Data requirements	Typical placement
Best sellers	Highlights top-performing products by sales volume	None	Home, category, search
Recommended for you	Personalizes picks based on individual browsing and purchase history. For shoppers with no item page visit events or completed transactions, recommendations are generated based on the first items clicked in the last 90 days by other first-time visitors (see FAQ)	At least 1,000 unique profiles who visited a product page more than once. At least 10,000 of one of the following: product.view events or transaction.charge events.	Home, category, search, PDP, cart, checkout, post-purchase
Similar products	Matches items by catalog attributes such as category, brand, and specifications	Complete and consistent catalog attributes	PDP
Frequently bought together	Surfaces items commonly purchased alongside the reference product	At least 1,000 transactions with more than one item in the cart	PDP, cart
Recently viewed	Shows the shopper's last-seen items based on view and interaction events	View and interaction event tracking	Home, category, search, PDP, cart, checkout, post-purchase
Visually similar products	Matches items by visual similarity using image recognition	Visual model enabled for the account	PDP
Manual collection	Displays a merchant-defined set of products from a collection or rules	None	Any

For FastStore, headless, and mobile implementations, campaign identifiers and API usage (including BFF campaign types such as rec-persona-v2, rec-cross-v2, and others) are documented in Integrating product recommendations in headless or FastStore web stores and the Recommendations BFF API reference. Store Framework uses the app and block model described in the vtex.recommendation-shelf guide.

Choose the guide that matches how your storefront is built:

Storefront	Developer guide
Store Framework	Product Recommendation shelf (vtex.recommendation-shelf): blocks, props, and Intelligent Search pixel prerequisites for tracking.
FastStore / Headless	Integrating product recommendations in headless or FastStore web stores: session management, fetching recommendations, event tracking, and Activity Flow setup.
Mobile	Integrating product recommendations in mobile stores: session management, fetching recommendations, and manual event tracking via API.

FastStore, headless, and mobile implementations all use the Recommendations BFF API. Store Framework uses the app and block model and does not call the BFF API directly.

Yes. You request the strategies you want during activation. VTEX enables and trains what is agreed for your account.

It depends on the strategy. Strategies that do not rely on model training, such as Best sellers, Recently viewed, and Manual collection, are available as soon as the integration is configured. Strategies that require training become available once the minimum data thresholds are met:

• Recommended for you: requires at least 1,000 unique profiles with repeat product page visits and at least 10,000 product view or purchase events.
• Frequently bought together: requires at least 1,000 transactions with more than one item in the cart.

Beyond those thresholds, time to readiness also depends on catalog size and overall traffic volume. Your onboarding team can give more accurate timelines for your store.

Typically includes navigation and product interaction events used to learn views, affinity, and personalization signals, plus order data when identifiers link a session to a completed purchase. Your integration guide lists the events you should send explicitly.

Strategies that rely on co-purchase patterns, such as Frequently bought together, require a minimum of 1,000 transactions with more than one item in the cart. Until that threshold is reached, the strategy remains inactive. Once the volume is met, VTEX trains the model and activates the strategy. Until then, consider relying on strategies that do not depend on co-purchase data, such as Best sellers or Similar products.

Recommended for you depends on a personalization model trained from real shopper and purchase behavior. When a shopper has no item page visit events or completed transactions, recommendations are generated based on the first items clicked in the last 90 days by other first-time visitors in the store. As the shopper builds a history, recommendations gradually shift to reflect their individual behavior.

Until there is enough account-level history for VTEX to train and enable the model, the campaign does not behave as fully personalized. What shoppers see in the meantime is configured as part of onboarding (for example, emphasizing other strategies such as Best sellers, hiding the shelf, or using a transitional setup). Confirm the exact behavior with your VTEX contact so your UI matches expectations.

Campaign VRNs follow the format vrn:recommendations:{store-name}:{campaignType}:{campaignId}. You can retrieve the full VRN from the store Admin under Storefront > Recommendations: in the shelf list, click the three dots next to the shelf and select Copy ID. You can also request it through VTEX Support. For the list of available campaign types, see Integrating product recommendations in headless or FastStore web stores or Integrating product recommendations in mobile stores.

Open a ticket with VTEX Support with examples (page, strategy, time). Wrong or sparse results are often tied to catalog structure, missing events, or a strategy that is still gathering training data.