How Statistics Drive Smarter Marketing

Matt O'KeefeWeb AnalyticsLeave a Comment

Businessman and problems

 

Online marketing is a complex, ever evolving combination of elements including timing, segmentation, and brand recognition.  Paid search marketing teams work diligently to find the perfect mix of these components in order to allocate resources properly, and generate pertinent client recommendations.

While the correct use of the aforementioned elements are ‘tried-and-true,’ what about statistics?  Most of us have horror stories about college statistics courses, and may shy away from advanced mathematics.  We should embrace the power of statistics.  Complex math and marketing are often seen as separate entities, but in reality, they can be perfectly complimentary.

Statistics can be used to predict any outcome variable.  Consider linear regression modeling.  Using this type of analysis, it is possible to predict revenue based on number of clicks, average position, cost, etc.  The analysis will yield a formula in which you are able to ‘plug in’ numbers to predict your outcome variable (revenue in this example).

Let’s take a look at a sample equation, which can be output using a number of statistical software packages such as R or SAS:

-Revenue=3.25+1.95(number of clicks)-17.19(average position)+1.72(monthly cost).  R2=0.75

Using this formula, if a page gets 3,000 clicks with an average position of 3.4 and a monthly budget of $1,000, we would expect to see over $7,500 in revenue.  This is just one example of a linear regression equation.  Models can be constructed to predict any continuous metric, such as cost, profit, or number of clicks.

While some people may consider this to be gimmicky, the analysis is also able to compute multiple measures of precision, including 95% confidence intervals and R2 values.  These values allow you to determine the amount of explained variation within each constructed model.  For the sake of our model example, an R2 value of 0.75 tells us that the number of clicks, average position, and cost are responsible for 75% of the value calculated for revenue.  The remaining 25% of the variation is considered unexplained (yearly trends, poorly constructed webpages, etc.).  The closer this value is to 1, the better.

Statistics is a powerful tool for any industry.  The use of predictive modeling can increase profits, business opportunities, and marketing success.  Statistical analysis can be easily integrated with Web Analytics, to improve paid and organic search strategies.  Marketing agencies can leverage linear regression analysis and other statistical tests to learn where money can be spent or re-allocated.

Leave a Reply