I don’t know about you, but one of the perennial issues I experience when communicating analytical findings to clients, or fellow business leaders, is to help them avoid the pitfall of assuming that correlation equates to causation.
Once a relationship can be shown between some customer characteristics and the objective of interest, say likelihood to purchase, people love to rush to hypotheses as to why this makes sense – even when it is extremely unlikely and causation has not been proven.
Now there are plenty of studies showing examples of spurious correlations, like the proportion of blue-eyed customers coming into a store in Moscow and the murder rate in Los Angeles. So, an extreme example can normally be thought up to illustrate this danger. However, too few people actually understand causality and how it can be proven statistically. This is also important because of the unconscious bias that we all have to seek to simplify problems and attribute causation as soon as possible; thus it can feel like ‘swimming up stream’ to suspend judgement and seek robust evidence.
So, I’m pleased to share this guest content, by Vincent Granville, recommending a classic text to help with this very challenge:
Have you read this? How do you help others understand whether to not they have proven causality?