Read a quote by token

How much of the eligible catalogue the candidate query looked at, from the same snapshot as the lines. Discovery and sourcing counts, taken before any semantic policy runs: a semantic annotation changes neither eligible nor these counts, because it removes nothing. eligible above counts collapsed domain+placement slots; this counts candidate OFFERS before the collapse, so it is the field that shows whether the 4,000-offer retrieval cap bound the result. eligible_offers equal to retrieved_offers means every matching offer was ranked; a larger eligible_offers with truncated true means the basket was chosen from the first 4,000 in the brief's requested priority (demand, leaders, then fastest/safest/cheapest/recommended, with raw price and offer id as the deterministic tiebreak). Raw-price retrieval is an approximation of the final buyer-credit sort, and budget fill can end a basket before the cap matters, so truncated false is not a promise that the basket is globally optimal. sourced_offers and the fields beside it describe the second, bound read: the sourcing expansion re-reads every eligible seller of each chosen placement, so the winner per placement is chosen from the whole supplier set rather than from the retrieval cut, and a placement is either sourced whole or not at all. Aggregate counts only.

Aggregate semantic-annotation coverage for this basket. Present only when the request carried customer context, at least one placement was affordable, and a semantic service was reachable — absent otherwise, which is not the same as 'judged and found nothing'. mode is the service's mode (off | shadow | rank) or unavailable when it was configured for the deployment but could not be built or failed; status is the service's own coverage status, or unavailable for that same failure. rank is the only mode that can change a recommended basket, and only in that strategy: a rank-usable annotation may reorder it within the explicit demand/leader priorities, and no judgment — below threshold, unjudged, missing, invalid or failed — removes anything. shadow annotates only, and cheapest, fastest and safest are unchanged. The service judges a sampled publisher page resolved by site id, not the offered placement, so no judgment is a verdict on whether to buy a slot. Topical fit is a semantic judgment on the service's own scale, NOT a probability of SEO impact, and it never affects price. Aggregate counts only — no customer ids, provider state or evidence text.

Only when exclude_domains_linking_to was given: the target's profile bands beside the basket's.