blog header - graphic - holding up line chart

Get the Most Out of your Marketing Budget with Marketing Mix Modeling

American merchant and marketing pioneer John Wanamaker is famously attributed with the quote: “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” If you’re a business leader, it’s likely that you’ve felt Wanamaker’s aphorism a little too close to home. In uncertain times like these, many companies are forced to cut budgets where they can, and because the effects of marketing are often ambiguous and left unmeasured, those budgets are frequently among the first on the chopping block.

The obvious problem with that approach is that you may be cutting into the half of your marketing budget that isn’t wasted, and the cash constraints that your company is facing may be compounded with lower sales as your company reaches fewer customers. But maintaining or increasing your marketing spend is often quite costly as well, and if the marketing tactics you invest in aren’t effective, that would also be quite deleterious to your business.

Fortunately, there are relatively simple and computationally easy statistical methods available to help us understand the effectiveness of the various marketing tactics a company may deploy. One approach is to use hypothesis testing where you can apply a specific marketing tactic to some consumers and not to others and compare the result. But for business leaders who need to understand their customers immediately, actually designing and implementing such experiments is far too impractical. A better method is Marketing Mix Modeling.

Marketing Mix Modeling (or “MMM”) is the process of using statistical methods, usually regression, to derive a mathematical representation of the effects of the different marketing activities of a company on their sales. In other words, MMM is a way to statistically model your company’s sales to determine how effective each of your marketing tactics is. 

linear regression graph

Most commonly, MMM uses multiple linear regression to predict sales (or profits, market share, or another business objective) as a dependent variable by analyzing the company’s marketing tactics and other relevant factors. Mathematically, this methodology determines the coefficients for each independent variable and intercept that minimizes the sum of all squared errors of the predicted dependent variable. These coefficients can be then interpreted in a meaningful business context, as an “X” increase in a particular marketing tactic can be used to predict a “Y” increase in sales, for example. 

MMM does not only provide valuable insights to the effectiveness of a company’s marketing activities, but also can: 

  • Elucidate the degree to which competitive actions have affected a company’s sales
  • Calculate how seasonal a product is
  • Determine how much of a company’s sales would occur in the absence of any marketing activity
  • Estimate elasticity and demonstrate how changes in price will affect sales

But beyond simply learning about how effective a company’s various marketing tactics are at driving sales, the most useful deployment of this model is how it can be used to guide your decisions as a business leader. Knowing how effective your paid search spend is compared to your banner ad impressions can allow you to know which digital approach should be leveraged. Knowing how effective trade show activities have been for you can inform your decision on how many salespeople to send to the one next quarter.  MMM can tell you which marketing activities should be invested in more, and which should be done away with, and this information can allow you to allocate your budget in a way that maximizes ROI.  

This approach is clearly quite powerful, and fortunately modern computing allows for these models to be run very quickly and efficiently and many free and open source software programs have this capability. That said, modeling is as much an art as it is a science, and the skills and knowledge to understand how to mitigate common problems in regression such as nonlinear relationships, heteroscedasticity, autocorrelation, and multicollinearity are often hard to find. And simply running regression models without understanding the nature of the variables you’re modeling will often lead to impractical conclusions, so understanding the business context as well as the statistical nuances is imperative for using these tools effectively. 

Caveats aside, MMM is a powerful tool that can help your business both save money and grow sales. If John Wanamaker had simply understood multivariate regression, then perhaps far less than half of his marketing budget would have gone to waste. 

Share this:

Share on facebook
Share on twitter
Share on linkedin