Visibility is one stage. Selection is another.

Ask which decision is happening

A person researching software through an assistant still has a different buying path from an autonomous agent selecting an endpoint within an authorised budget. The same visibility metric cannot describe both.

Map eligibility, retrieval, evaluation, selection, authorisation and execution. Then name the evidence available at each stage.

Find the actual bottleneck

If the endpoint is absent from a relevant catalog, offer testing alone cannot solve access. If it is discoverable but loses on capability or budget, more citations may not solve selection. If it wins but fails, product work comes first.

Measure a connected journey

Keep stage-level metrics and connect them when identifiers and access permit. That gives the team a useful answer to what to change next, without pretending that an isolated discovery gain is a purchase.