Outside special name-matching branches, source starts score seventy-five, adds four per topic, then six if interview angle exceeds thirty characters or two otherwise. Hypothetical three-topic plans yield ninety-three with long angle and eighty-nine with short. Bars share six pixels per score point. This rubric does not establish a suitable real guest pairing. Hypothetical six topics with long angle produce uncapped one hundred five, but source caps score ninety-nine and floors sixty-five. Bars compare raw and capped results at five pixels per point. An upper-bound label is not measured quality or evidence that a guest agreed to an interview. General source audience mapping adds one point two million to score divided one hundred times one point five million, then displays one decimal. Assigned eighty gives two point four million; ninety gives unrounded two point five five. Bars share two hundred pixels per million. Special named fixtures bypass this mapping, and none of these labels is verified audience reach. Complete original raw pitch JSON/reset/native resize/first-footer checks preserve existing planning.