Recommendation Containers

What it solves

Recommendation models produce ranked product lists — but a model doesn't know which page it's serving. A "Similar Items" model could serve a product detail page, a cart page, or a post-purchase page. A "Best Sellers" model could appear on the homepage or a category browse page. Recommendation Containers are the mapping layer that connects models to placements.

Without containers, models can't serve recommendations. A container is required for each page type where recommendations should appear.

When to use it

• Adding recommendations to a new page type — create a container that maps the page to the appropriate model
• Swapping the model behind a placement — update the container to point to a different model without changing storefront code
• Disabling recommendations for a placement — remove or deactivate the container

Key concepts

Container — a named mapping between a page type (homepage, product detail page, cart, etc.) and a recommendation model. The placement in your storefront calls the MXP Recommendations API with a container ID, and the API returns ranked products from whichever model is assigned to that container.

Page type — the context in which the recommendation is shown. Common page types include homepage, product detail page (PDP), cart, and category browse.

Model assignment — each container points to one recommendation model. To change which model serves a placement, update the container rather than touching the storefront API call.

How it works

Open Recommendations → Recommendation Containers from the left sidebar. Each row represents a container, showing its name, the page type it serves, and which model is currently assigned.

To create a new container, click + New Container and select the page type and model. The container becomes active immediately.

To reassign a model, edit the container and select a new model from the dropdown. The change takes effect on the next inference request.

Quick example

A team wants to show "Similar Items" on the product detail page. They create a container named pdp-similar, assign it the Similar Items model, and point the storefront's PDP recommendation component at the pdp-similar container ID. Later, they swap the model to a tuned version by editing the container — no storefront change needed.

Related pages

• Recommendation Models — the models you assign to containers
• Recommendation Rules — business rules that filter or override what models return
• Recommendations overview — how the full recommendations system fits together