Could margarine consumption really be the leading cause of divorce in Maine? Probably not, but as business leaders, understanding the difference between correlation and causation can save you from making bad decisions, like taking margarine off the shelves in Maine (is margarine still a thing?).
Correlation is described as a relationship between two variables where if one increases the other increases. For example, as the temperature goes up, ice cream sales goes up…pretty basic.
Causality, on the other hand, is the quality of the relationship between cause and effect. It digs deeper into exploring the root cause of why things have an impact on one another….not so basic.
I recently came across some pretty amusing analysis by a Harvard law student that showed the relationships between 2 completely random events. Not exactly light summer reading, but it made me laugh and opened my eyes to how marketers can make bad decisions looking at data that does not get to the real cause.
The Harvard study compared unrelated and in some cases morbidly random topics. Here are a couple of my favorites: there was a 99% correlation between margarine consumption in Maine and the divorce rate.
How about this one: there was 97% correlation between people who died becoming tangled in their own bed sheets to revenue generated by ski resorts. Seriously….who knew?
It would be absurd to argue (such as in the above examples) that one random topic was the cause of the other. The reality is that marketers have been known to make similar mistakes when working with large sets of unrelated data, trying to draw meaningful insights and strategies.
If you rely solely on correlation when looking at consumer marketing data and don’t get to the root causes of events, then you are putting together a puzzle without all of the pieces.
At SpotRight, we analyze billions of pieces of consumer data based on social graph connections and content to understand the relationships between consumers, brands and interests. We use data and insights to inform strategies for Fortune 500 global brands and agencies. We have worked with Super Bowl brands to help them uncover the people who were engaging with them through social media during a multichannel, mass media campaign, getting down to demographic data. This was a first for them and challenged some of their previous thinking on who was engaging with their television ads.
In one specific case, we were working with a leading CPG brand who didn’t think that their consumers were on Twitter and did not have a cohesive strategy to market to them. When we analyzed their consumer data we saw that not only were they on Twitter, but they were heavily engaged with their top competitors and represented their most valuable consumer segment. Data can be a powerful tool for marketers to infer correlations, but it can’t stop there. Here are some additional ways to get to the root cause of events:
- Ask the right questions up front and use the data to find the right answers. Data is a science but creating actionable strategies requires digging into the why. Who are my customers currently? Why have they engaged with my brand? How do they engage with my competitors? How do I engage them with relevant messaging and content?
- Integrate all available data at the consumer level into your CRM. As a colleague of mine says, trying to make good decisions with partial data is like bowling with a sheet covering the pins. Go beyond your own data to 3rd party data and look to gain insights from emerging channels, such as social, that provide a completely different pool of insights. If you don’t have your own data, tap into valuable insights from social around your target or competitive audience.
Next time you see a correlation in your marketing data that doesn’t seem right, stop and challenge your thinking behind how you are drawing conclusions. Are you asking the right questions? Do you have access to all the data possible? And does it make sense? If you cannot answer these questions chances are you may be making a strategic decision that may hurt not help your business. For a good laugh at some funny correlated events, have a look at the original analysis.