In 2012 Obama lost significant vote share relative to his 2008 win, while Romney gained slightly from McCain. The shifts just weren't enough to swing the election to Romney.
One reason why those shifts weren't big enough was because Romney ran a traditional campaign. Lock down the base (he actually did ok there) then try to convince Moderates to vote for you.
Obama took a different approach. Knowing that Moderates had soured on him, his campaign used data-driven methods to increase turnout by Liberals who didn't have that great a record of previous voting. Which turned out to be much more effective and less costly as well.
PJ Media has pretty good article on this. But the key point is summarized on this slide:
It was the unanticipated presence of these new (but Liberal) voters that screwed up all the election night projections, which had Romney winning.
What this would seem to indicate is that the best solution for the Republican candidate is to follow the same pattern and identify and go after Conservatives that don't have a big history of voting. It may not work, but its probably the best option they've got.