Kevin Drum kindly links to my review of Bob Shrum's memoir. The way he framed the piece makes me want to clarify that I didn't mean the upshot of the review to be that I think Shrum made some particular tactical error in thinking it would be a bad idea for Al Gore to deliver a big speech on global warming during the 2000 campaign (though I do think this was an error). The point is just that Shrum's methods, as reflected in that episode, involve what I guess you'd call a kind of public opinion literalism. First, you identify the three (or five) issues that voters say they care most about. Second, adopt positions that poll well on those three (or five) issues. If there's a topic where your candidate can't in good conscience adopt a popular position, you try not to talk about it.
This is Shrum's method, and it seems to be the method of most of his colleagues in the Democratic consulting game. The only problem with it is that it doesn't work. Indeed, it such a bad method that I don't think anyone would explicitly defend it. Nobody denies that preference intensity matters, that there are threshold criteria, or that elections are primarily decided by low-information voters who tend to vote on character questions and that issues primarily matter because they affect perceptions of character. Nobody denies that stuff, but they keep doing the same thing anyway.
The issue, I think, is that polling is a useful way for the consultants themselves to evade accountability. If Gore took my advice and lost, all you could say about it was that my political judgment had proven to be poor. When candidates take Shrum's advice and lose, by contrast, Shrum gets to point to the polling data as hard proof that his strategy was correct and blame defeat on other factors -- the inevitable gaffes, unfair attacks from the opposition, etc. Admitting that political strategy is more art than science would threaten the position of the strategists, so the strategists rely on the most quantitative tool available, even though on some level (note that political pollsters' findings tend to vary quite a bit according to ideology) they don't really believe in it.