ROI: Shift Your Marketing Into High Gear

Advertising ROI

ROI: Shift Your Marketing Into High Gear

A saying attributed to John Wanamaker generally produces nervous smiles from many marketers “Half the money I spend on advertising is wasted; the trouble is I don’t know which half.” I don’t know about the 50% part but the sentiment is usually true. If you have just produced a world-beating radio spot, an ad that puts others to shame, and an outdoor campaign accused of causing several car accidents, and revenues go up by 20%, how do you know what was the main driver?

The short answer is that it is not easy. However, it can be done by performing marketing mix modeling (MMM) using multivariate regression analysis.

The reasons for developing an MMM are simple:

  • You can determine how much each marketing tactic is contributing to your sales
  • Provide you a roadmap for what activities to stop, boost, or continue at current levels
  • What is the ROI of each activity and how budget should be allocated
  • How long the impact of each activity last
  • Predict how allocation of certain dollars to individual marketing activities will affect sales

To get started, you would first need to collect performance data for your dependent and independent variables. Generally, sales are your dependent variable because its value depends on the independent variables, which in this case would be frequency, reach, or dollars allocated to radio, ads, billboards. You would need at least 2 years of historical data to develop a robust model. The saying, ”garbage in-garbage out” strictly applies here.

The good news is that you will know out how good the model is in predicting sales based on the inputs that were put into the model. The reliability of the model is determined by goodness of fit or R2 value, which basically tells you how well the sales values are explained by the model. It is the percentage of the response variable variation that is explained by a linear model. R2 is between 0 and 1. The 0 indicates that the model does not explain variability of response data around its mean and 1 means that it explains all variation. In general, higher the R2 value, better the model.

If the R2 value is low then you may be missing an important variable that impacts sales, e.g., a new service line, expansion of sales force, competitive forces etc. or you just need to get more data points. The model works best when volumes are high, e.g., consumer packaged goods or number of patients and data scarcity is not an issue. This may be a concern for capital equipment.

The coefficient of the model will provide you the information you are looking for. It will tell you the relative effectiveness of radio, ads, or outdoor campaign on sales. That will help you make informed marketing decisions as you develop your next marketing plan.