pol.is has done the super-deluxe version of this, to great effect in Taiwan, and elsewhere.
As I understand it, participants agree/disagree/pass on multiple statements (essentially an opinion survey on a topic area, which they can extend with statements of their own, like “Playing football in high school is reasonably safe”),then pol.is identifies groups of people who more or less agree on sets of statements.
You can then either attempt to craft statements that everyone agrees on, or drill down on one particular cluster to see what differentiates those participants from other clusters, or from one another.
Write-up of the VTaiwan project where this is working here: http://civichall.org/civicist/vtaiwan-democracy-frontier/
Screen demo of the clustering in action:
(skip forward to about 41:00)
Cool beans.