How to Misuse Analytics

Catherine Tucker at MIT Sloan and Anja Lambrecht of London Business School have a great HBR piece on how to make big mistakes with analytics.  Here are what they see as the biggest challenges:

1. Size isn't everything.  Usually managers have to merge data sets from different sources that were designed for different purposes.  If the data sets cannot be cross-referenced in a meaningful way, their usefulness can be limited.  
2. Our ability to analyze structured data is well ahead of our ability to analyze unstructured data.  The authors observe that firms have had more success with unstructured data when analyzed in conjunction with structured data.  
3. Data processing skills are more critical to reaching meaningful conclusions than bigger data sets.  Companies need to invest in both.
4. Correlation is not causation.  Any well-trained PhD in economics has had this drummed into their head all the way through graduate school.  But now lots of employees are looking at computer-generated results, see a correlation and think they have something.  Field experiments on relatively small amounts of data are likely to lead to greater insight.