So, if I'm following this, the main advantage then is performance when only selecting certain information? As in, you have an index on 5 columns, but you only require the data from 2 or 3, so it only has to read the data for those 3 columns instead of all 5? Seems like it could really boost performance on large data sets, though it won't help us when we're doing day to day work as this is definitely targeted at BI/OLAP.
At this point of time, Column Store Indexes are for read only analytics. You cannot use DML statements. It is specifically for BI/OLAP.
Please check the first link in the Suggested Readings section at the bottom of my post.