In the Coveo Merchandising Hub (CMH), navigate to Recommendations and create a new recommendation slot, selecting a strategy (e.g., 'Frequently Viewed Together', 'Top Sellers', or a custom ML model).
Configure the slot's global settings: set the number of products to return per call and which product fields should be included in the API response.
If business rules are needed, add boost, bury, pin, or blacklist actions within the slot configuration; these override the ML-driven ranking for specific products.
Integrate the slot into your storefront by referencing the slot ID in a Commerce API recommendation request; the response returns a ranked list of product objects populated with the fields specified in the slot config.
Monitor slot performance in the CMH analytics view; use A/B testing in the Coveo Experimentation Hub to compare different recommendation strategies before rolling them out fully.
Known gotchas
Coveo Commerce models (used by recommendation slots) are trained on event data collected via the Commerce API; a new deployment needs sufficient event volume before the model produces meaningful personalized results.
Pinned products in a slot will always appear at the specified position regardless of the user or context — use this sparingly to avoid degrading personalization quality.
Recommendation slots are tied to specific catalog configurations; if you restructure your product catalog schema, review slot configurations to ensure the correct fields are still being returned.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp