I’ve read a couple blurbs about conjoint analysis and I am not yet totally clear on how it works, but the general idea is to provide different products (or in this case candidates) and ask respondents to rank them. So one might have compared a veteran former Obama staffer who wants to get rid of the tariffs vs a farmer that created private sector jobs who wants to make food more affordable (etc), and the scores come out of a bunch of rankings.

I speculated up thread that the poll respondents might be ranking the candidates more on what they believe the general public wants or who will most likely win an election regardless of personal preferences (which feels like a hazard in accurate interpretation and it depends on the exact question wording), or they might’ve just had a sample that skewed older (as Chana observed).