Nowadays I am reading a lot of market research reports on technology and business process in various industries as part of my job. As I read these, I am reminded of the book Prof. Laha recommended in statistics course “The Tiger that Isn’t…” (I strongly recommend this book for anyone who is reading reports regularly either for politics, stock trading, general management or even just general newspaper reading).

The general belief that “numbers don’t lie” was shattered as we read that book and discussed in class. It gave my already suspicious mind another reason why business reports and newspaper columns need to be taken with a pinch of salt!!! The causes vary from simple negligence to manipulation. More than anything the book helped me on how to read/critically analyze below statements:

  1. Two drinks daily increases risk of breast cancer by 12% on women!
  2. 1 in 4 brit teens steal!!
  3. Students of single-sex schools fare better; so single-sex schools are better for girls!!!

Averages are merely average:  Example is Indian per capita… do you decide whether you want to live in India based on Indian per capita? Then why choose MBA programs based on averages!!! I remember vaguely reading of white rainbow: Remember the average of all seven colors in a rainbow is white – which doesn’t even exist in a rainbow.

Sampling issues: A survey is only as good as the sample and the sampling mechanism used.

Detractors in market survey questions: if u wanted to get a skewed opinion to one of ur choices you can easily clone other choices to increases chances of favorite.  Same as candidate to split votes of opponents in elections;

Clustering: Whenever there is a cluster our mind likes to demand reason. Accept that clusters can happen randomly too (a quality batsman goes out of form) or it is bcoz of extra attention: recent trends – IPL bids, policy issues in managing airways; honor killings. All these damning things existed all these years! Read again when you just read “Honor killings on the increase!”

The point of this article is not to dismiss numbers or reports. Ask questions and get closer to the numbers. For example, on the above statements, ask 12% of what? What is the baseline risk? What is the absolute risk? What was the sample? What are you measuring? Is the correlation right? Or just ask what do you mean?

Ref: The tiger that isn’t. By Michael Blastland and Andrew Dilnot; creator and presenter in BBC Radio