Why is Correlation Not Causation..? Part II

In this second look at correlation we review a few cases where misleading commercial relationships result in an apparent paradox, which are resolved by looking at the causal reality at play.

Cavemen and Marketing Directors are, amongst other things, relentless pattern spotters.  For both, noticing and exploiting regularities is a matter of survival, whether that’s where the gazelles go for water or which products over-index on a bank holiday.  The problem is that these informally observed relationships can often be misleading.

In our previous blog we discussed the spurious relationship between chocolate consumption and Nobel Prize winning, the observed correlation being produced by both outcomes having a positive relationship with GDP.  The table below now shows three real-world commercial examples we’ve witnessed in the grocery sector and the strategic blunders they precipitated.

Muddled Correlations in the Grocery Sector

In each case the causal chain generating the ‘commercial insight’ is more convoluted than was immediately apparent.  As a result, the actual relationship was either muted or reversed.  Essentially, the world is a complex dynamic system where everything is connected to everything else via an intricate web of causal links.

As a consequence, just because two features are seen moving in tandem it doesn’t follow that you can automatically use one to control the other.  The grocery examples show how the complex interweaving of events can yield unanticipated consequences which can derail such an approach.

The only reliable solution, and the subject of our next blog, is a randomised controlled trial (RCT).  Basically, you have to do a trial run and manipulate one variable to measure its actual impact on the other variable.  That is why RCTs remain the gold standard for measuring cause and effect.