The debate about Correlation vs. Causation is over

December 19th, 2012

Topics: Optimization, Trends


After reading this article in Slate , which went viral recently, I was both surprised and pleased. Surprised to see a technical statistical topic being discussed in the mainstream media, and pleased because the article underscores the importance of distinguishing between these two concepts, a point I always emphasize when I talk about Webtrends Optimize.

My argument follows the following line: Every decision must be based on a cause-effect relationship. Correlation is not sufficient – it just provides additional information that prompts a future investigation. The only way a cause-effect relationship can be discovered is by performing a designed experiment (see any standard statistical text; for example, Michael Sullivan, III, Statistics: Informed Decisions Using Data, 2nd edition, p.15). Then the conclusion is that you have to run designed experiments if you are involved in decision-making.

The following chart provides more detail:


Webtrends Optimize is a platform for running designed experiments in the Digital Marketing environment. The controlled variables are pieces of the content presented to the customers, and the response variable is the conversion rate. The sought-after cause-effect relationship is between the content and conversion rate. As I mentioned above, the only way to reliably identify this relationship is to run designed experiments. In my next blog, I will describe the steps involved in the process of setting up a designed experiment.


Leave a Reply

  • (will not be published)

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>