Are you creating an extract from the csv? If the string columns have many distinct values then they may not compress well. A million values is not that many though so there may be other issues causing the performance you are seeing. How did you determine that the string columns are the cause of the problem?
Ok - thanks. I hadn't determined that the strings were the issue but was more wondering out loud. The performance issues I encountered were more actually due to trying to render a million rows of data in one visualisation rather than anything. Thanks for the response!