What practices have folks used for logically grouping federated entities

My team at S&P Global is reviewing ways that we can logically group companies based on a number of criteria.

Our supergraph defines a company as a federated entity (e.g. Instn) and the team is looking to provide a common capability in the institution subgraph to define sectors of companies based on such things as headquarters (geography), industry, and other aspects.

One idea is to define a federated InstnSector entity that would allow other subgraphs to operate against a collection (e.g. [Instn!]!) but how that would work may lead to defining the set of Instn identifiers via encoded string (base64 encoding of json / csv) which seems less than ideal from a design perspective.

Another idea is to avoid defining the InstnSector as a GQL entity and have the subgraphs directly pull from the owning insitution subgraph via an GQL intraservice call (e.g. client->router->subgraph->router->institution) which is less than idea from the context of observability and cost analysis at the router level.

How have others here tackled this kind of use case in the past?