palocssw@jmu.edu} U sing performance data on a common test instrument administered to more than 4,000 business statistics students studying under 16 different instructors over a period of 6.5 years, we identify areas of consistent student weakness. In addition to verifying difficulties found by earlier researchers, we document a fundamental problem in students' ability to reason with cumulative probabilities, a problem with implications for solving problems with both discrete and continuous distributions. Our performance data also suggest that it may be useful to view statistics problems using a taxonomy based more on solution procedure than on traditional statistics topics. We employ a new methodology based on cluster analysis to identify areas where the observed student difficulties may be particularly addressable. Finally, we offer some preliminary findings on how these problems may be addressed with practical teaching suggestions based on the approaches employed by our more successful instructors.