If you run several stores in one workspace, the same product usually lives in more than one of them — same supplier, same images, different market. Cross-store linking makes that relationship explicit, so what one store learns about a product doesn't have to be relearned from scratch everywhere else.
How auto-linking works
Linking is deterministic image-key matching: two products whose images resolve to the same keys are the same physical product, whatever their titles say in whatever language. No fuzzy guessing on names or descriptions — a link either has the image evidence or it isn't proposed. And it's dry-run first: the matcher produces proposed links for you to review before any link is written. You confirm the matches once; the links persist.
Propagating winners
Once products are linked, proven wins travel: a product that earned winner status in one store can have its operations propagated to its linked siblings — and the same goes for the unglamorous direction, propagating an archival so a product you've killed in one store doesn't keep limping along in three others. Every propagation is a logged decision on the receiving store, revertible like any other, not a silent sync.
Fleet priors are advisory only
The Marketing Brain also sees fleet-wide track records as priors — if a lever or product pattern has a record across your other stores, that informs the reasoning. But priors are advisory only: no action fires on any store because of another store's data. Each store's decisions gate on that store's own evidence, thresholds, and autonomy settings. The fleet makes each brain better-informed, never remotely-controlled.
Link on evidence, propagate on approval, and let fleet data advise — never drive.