There are a number of improvements that can be made to the UX - mabe these should be splitted up but here they are.
Ability to work on a variant with relations and publishing without the UX forcing you to view the whole baseproduct and all other variants - the current UX makes it easy to make misstakes.
Be able to create and add baseproducts and variants direct into selected product category - to make it easier to work with insted of first creation of the product and then connecting it to a category.
When you search for a variant article number the UX give you the baseproduct as a result - let the UX display both variant articlenumber that was found and the "related" base product.
When working on a variant never switch to baseproduct in the UX - let this be a choise for the user - if the user want to work with the baseproduct - just select it in the current dropdown under first tab.
Export of a filtered list - only include the matching variants into the export file - current UX export include all variants under the matching variant baseproduct(s). The front end / website do it correctly.
Enable configuration of how the UX category left menu to display connected base products or even the variants in the left menu tree / Category tree - this would make the UX faster for compines/brand who work with variants as core products.
I agree too with Jennifer and I belive in this case we would have more use of option one.
I fully agree to Jennifer. to use boosetd relation sorting could work in a more generic product context, but for us where each individual product is important the presentation of the relationship order is important and can be critical.
So if we could keep the "mixed" relationship field as it is now - but allow us to better controle the front end how it should be presented and sorted it would be best.
Categorization (sorted) + Manual sorting (based on the related product category)
Since it was not possible to comment on the merged idea PL-I-149 I will respond to Ludvigs comment here: For our use cases, manual sorting per relation would be the most useful option.
The main reason is precision and editorial control. When working with related products, accessories, cross-sells, spare parts, or curated product combinations, it is often important that products appear in an exact order chosen by the editor or merchandiser.
A global boost/ranking score can absolutely be useful in some scenarios, but it is also something that is comparatively easier to implement with custom logic, scheduled jobs, or automated ranking rules outside of the core relation handling itself.
Manual relation sorting, on the other hand, is much harder to recreate cleanly without custom extensions to the UI and relation model.
Because of that, we believe manual sort order on relations would provide the greatest long-term value and flexibility as a standard platform feature.