If the total spend of customers in New York City is about 9% higher than customers in New Hampshire and we can only increase marketing in one place or the other, we should choose New York to maximize revenue. People in New York spend more than people in New Hampshire. Sounds reasonable, right? It is, until you consider that our “total spend” number includes an 8.875% sales tax in New York City, which is absent in New Hampshire.
You’ve probably hear the term “Correlation is not Causation.” Just because two variables are highly correlated doesn’t necessarily imply that one causes the other. There might be other reasons for the relationship that weren’t considered in the analysis.
The debate over global warming is full of these examples. People have demonstrated a high correlation between global warming and the amount of carbon dioxide from greenhouse gasses. Others have shown a remarkable correlation between global warming and the population of pirates, the change in U.S. postage rates and the price of a share of Oracle stock. The list goes on. Children who live in a home with more appliances tend to do better in school. The best tennis players wear Lacoste. The amount of ice cream we consume correlates strongly with the number of shark attacks and children with more cavities have a larger vocabulary.
Real business examples are much more complicated than toasters and shark attacks. If our data analysis leads to a targeted marketing campaign and sales volume increases after the rollout of the campaign, can we claim victory for the data analysis? Or does the fact that the campaign was promoting a reduction in price have anything to do with the increased sales volume? In reality, it’s probably a little bit of both.
We are increasingly under pressure to demonstrate the value and ROI of data analytics and business intelligence. To do that, it’s necessary to show a cause and effect relationship between BI and some measurable business metric such as sales or profitability. If you can’t create a controlled experiment, it’s worth taking the time to think about what other factors might influence success. Including these factors in your analysis to let people know you considered them will probably help increase the validity of your business case when you present it to others.
On a more serious note, when the NFC wins the SuperBowl, the stock market goes up 90 percent of the time – and – the annual average S&P return is double what it is when the AFC wins. Does that change who you might root for?
Subscribe to our blog to stay informed on various data analytics topics.
Steve McDonnell
Spotfire Blogging Team
