Web Content Display Web Content Display

Breadcrumb Breadcrumb

Analytics Blogs

Back

The Correlation / Causation Smackdown!

Francesco Sguanci
Posted 5/21/20

Get the Truth from Your Marketing Mix Analysis

If you are a teenager, and you drink soda, you are much more likely to knife someone. No kidding! It turns out that soda, which is already being blamed for obesity and tooth decay, also turns teens into killers! At least according to some reporting sources (Check out The Washington Post’s blog “Soda Boosts Violence Among Teens” - https://www.washingtonpost.com/blogs/the-checkup/post/soda-boosts-violence-among-teens-study-finds/2011/10/21/gIQADmsY3L_blog.html

This claim was based on a study by a Harvard Economics professor who found a “strong association between soft drinks and violence,” suggesting “there may be a direct cause-and-effect relationship, perhaps due to the sugar or caffeine content of soft drinks.”

Nobody questions the results of the study or whether there is a positive association between teens who commit violent felonies and the number of soft drinks they consume; what is questionable is the assumption of a CAUSAL link between the two, which cannot be established with certainty. In fact, while the professor suggested that “maybe the sugar and the caffeine are the cause of increased violence,” there could be many other reasons that would be very hard to investigate. Perhaps teens in lower income families tend to eat more fast food (including soda), and violence is linked to poverty. Again, this, like many other theories, would be hard to “prove” with an observational study; only a controlled experiment could prove the causal link.

There are many “good examples” of media publications, government agencies, and corporations arriving at the wrong conclusions based on internal (or external) studies; when corporations take actions based on biased findings, they can experience a financial loss as well a loss to their image.

Let me give you a real-life example, from a major online retailer here only referred to as ABC Shopping. ABC Shopping wanted to quantify the number of conversions generated from their marketing mix, specifically TV flights, PPC advertising, and Direct Mail. They performed their internal analysis using historical marketing and sales data from January 2016 to November 2017.  One of the takeaways was that “TV flights are not statistically significant, while every additional dollar spent on Paid Search generated $3.09 in sales.” Using the results of the study, management decided not to run any additional TV flights for the remainder of the year (November 12 to December 31, which included the profitable periods of Thanksgiving and Christmas) and to use the TV budget for more Paid Search.

The decision was made based on the statistical association that was found between higher-than-usual sales and spikes in Paid Search. The assumption here was that higher levels of Paid Search CAUSED higher sales, and that by increasing the spend on Paid Search, the company could positively impact sales.

It turned out for this specific client that when the TV ads were run, various online channels, including Paid Search, always observed a bump in volume, as illustrated in the chart below and that TV ads had an essential role in generating conversions.

Conversions for the remainder of the year did not increase and actually dropped 3.6% from the same period of the previous year despite an increased level of spend on Paid Search. This provides a good example of the Correlation/Causation Smackdown related to your marketing mix analysis!

The takeaway from this should not be that association among variables is not important or that it can never be used to make decisions. Rather, I want to point out the complexities and intricacies of trying to determine cause-effect relationships and applying them to a business problem.

Do you have actions you could take to increase your profit but are reluctant to do so without a full understanding of the cause-effect relationship among your business drivers? Why not leverage Dunn Solutions’ marketing mix expertise?

Dunn Solutions’ predictive analytics team would be glad to give you a no-charge, no-commitment evaluation of your data and discuss whether there’s an opportunity to design the correct approach to reach your business goals!  Contact us today at info@dunnsolutions.com!