EDD Recommendations stores cross-sell and upsell mappings in product postmeta on the download post type: an array of related product ids per source product, plus optional weight or display order. The default product-edit screen lets you manage that array one product at a time, but there's no cross-product view of "which products have empty recommendations" or "which recommended pair actually converts at checkout".

SleekView reads the download post type with its recommendation postmeta exploded into a per-pair table: source product, recommended product, weight. Joining edd_order_items for both the source and the recommended slot turns the same view into impact analytics — which recommended pair shows up in the most orders together, which mappings never convert, which products are recommended back to themselves by accident.

Inline edits to the mapping array route through the EDD download CRUD layer (save_post_download), so cache invalidation, related-products hooks, and any custom recommendation logic listening for the meta update fire as expected. Bulk-add or remove a product from a cohort of source products takes one filter and one write, not a tour of the product list.

Cross-sell and upsell mappings have outsized leverage on average order value, but tuning them requires data the default plugin admin doesn't expose. Which pair actually converts? Which products have empty recommendations after launch? Which mappings are self-references that slipped through a bulk import? The mapping data lives in postmeta on each source product, the conversion data lives in edd_order_items for each order. Joining them is straightforward in SQL but inaccessible in the WordPress admin without a custom screen.

SleekView reads the same postmeta EDD Recommendations writes to and joins it to edd_order_items for per-pair conversion counts. Marketing sees which pairs work and which to cut. Product strategy spots cluster gaps and reciprocal-mapping failures.

Operators catch missing recommendations on new products before launch instead of after. Same database, same hooks, the recommendation flywheel finally gets the feedback loop the data already supports.