correlation-causation

Spurious Correlation: Are You Making This Mistake in Your Market Research?

Marketers’ relationship with data is changing. What was once seen as a savior to campaign targeting and development is now perceived as an unwieldy behemoth, to the point that some brands appear to be giving up trying to separate the signal from the noise.

A recent Experian report found that data management, the top challenge marketing executives reported facing last year, fell to fourth place this year. Just 36 percent of survey respondents prioritized it, despite the fact that many challenges still haven’t been met: Over a third of those surveyed said they still lack the internal resources, technology and accurate data to target consumers the way they would like.

july15

“The future will challenge marketers with an ever-widening array of data and digital channels,” Experian said of its results. “Finding the right strategic mix to best navigate those opportunities while remaining sufficiently conscious of brand messaging and positioning remains crucial to success.”

So what are the best methods for separating the signal from the noise?  The answer to this depends greatly on what question you are trying to answer!  Data is considered noise when it is not strongly tied to the objective at hand, but just as the saying “one man’s trash is another man’s treasure” data that is considered noise in the explanation of one objective may very well be the signal in explaining a separate objective.  This is where the value in qualified market research teams come into play.

In today’s world, data is relatively easy to come by, partially contributing to the data noise problem.  However the value isn’t in the data itself, but the story which the data tells.  Misinterpretation of this story can have significant repercussions on your business model, while proper identification can be your key to increasing your market share.   The proper interpretation of any set of data begins with correlation versus causation.  For example, looking at the divorce rate in Maine compared to the US per capita consumption of margarine might lead someone to think eating less butter leads to happier marriages in Maine!  This is an example of spurious correlation (correlation does not mean causation) – thus that statement would be wrong!

july15-2

While the above example is rather comical, making the same mistake in a b2b market research data mining context can be disastrous for any company.  Market research teams specialize in this area of statistical analysis and can often spot trends or connections in the market which have previously been disregarded as noise.  These teams are able to confidently separate correlations versus causations with sophisticated statistical analysis as opposed to relying on the “gut feeling”.  These statistically significant trends can then be used with confidence in anticipating new market movements, forecasting growth of different segments within the same market, and predicting the ever evolving preferences of your target market, among many others.  When considering the magnitude of potential benefits provided from sound market research compared to the relative low cost of a single study, the value of such services shine through and often lead to a new period of growth for the adopting company.

 

Written by Bluebox Research – A Pharmaceutical Market Research Company