Every prioritization meeting has the same ghost at the table: "the market." Features get argued for because "the market expects it" and against because "nobody cares about that" — claims about an external reality nobody in the room has measured. The loudest ghost-summoner wins, and the roadmap inherits their guess.

The market's actual opinion is on the record: across your category's review corpus, some capabilities get consistently celebrated (table stakes and delight factors), some get consistently roasted (the pain the category hasn't solved), and some get ignored (checkbox features nobody mentions). That distribution — praise, pan, silence — is the market's feature-level verdict, and it's queryable.

deep_search assembles the verdict corpus: the category's substantive reviews and comparisons, quality-filtered so the analysis runs over credible assessments rather than spec recitals.

deep_query extracts the distribution: "across reviews of <category>, which capabilities draw consistent praise, which draw criticism, and which go unmentioned?" — answered with citations to the reviews, so the prioritization meeting sees the market's verdict with receipts. Suddenly "the market expects it" is checkable: the feature either recurs in praise, recurs in pans, or lives in the silence column — and the silence column is where checkbox roadmap items go to be exposed.