MMM

A marketing mix model has two end goals:

​

1. Optimal budgeting

2. Channel mix optimization

​

Let's start the discussion with addressing the Optimal Budgeting piece. First, we break down the data by regions to isolate for impact that customer behavior and market conditions bring in a certain locality.

 

Idea is to create a table that has monthly data on budgets, MQLs, SQLs, WON $, # of WON deals, and the influenced metrics for these same pipeline metrics. 

​

An example of that is the following:

​

 

 

 

​

 

 

​

 

​

​

​

​

 

 

​

Other predictive metrics that can help with a better model are... 

• factoring for the seasonality (...high conversion months VS. low conversion months)

• building an adstock effect to show diminishing effect of high budget months

• other factors (i.e. macroeconomic metrics, competition in the market, disruption etc.) 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

​

​

Next step is creating a correlation of adstock budget (Budget transformed on the visual above) vs. Pipeline (Sql $). This correlation will return a marketing response curves for each of the regions (visual below). 

​

​

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Depending on the tool you're using for running this multiple regression, there will be different steps involved to get the equation of the trend. It will return a log-based equation... i.e.

 

 

 

 

 

 

 

In order to get the Budget where we get the max $ Return level, we have to get the point where the total Return (SQL minus Budget) number maximizes. Plotting the Return curve, you will get the graph like the one below...
 

​

​
 

​

​

​

​

​

​

​

​

 

The plot itself will show you the maxima of the graph.... (or you can be more precise by doing a differential of SQL minus budget!)
 

​

 

 

 

​

​

​

​

​

​

​

This budget will be the point at which the the pipeline return maximizes. However,  this may not all turn into billings. So you have to repeat this same process for Budget vs. Won $ to find the sweet spot where the billings maximize. The decision from there on is a blend of quantitative data and understanding of your business. 

​

If for example, the optimal budget recommendation on Budget-pipeline correlation is $80k, and the optimal budget recommendation on Budget-billings correlation is $60k, you need to begin testing spending close to $50k and change spend every quarter to raise it or lower based on the response.

 

To visualize this...

 

 

 

 

 

 

 

 

 

 

 

 

 

​

​

​

​

 

​

Considerations on Budgeting:

​​

• As you start spending close to the Optimal spend mark based on your marketing response curve, do increase and decrease your spend (especially on Paid and other demand capture channels) based on seasonality. This variation could stay in the range of +- 10-20%.

• With other mktg efficiency and process improvement projects in flight (channel mix optimization, lead scoring revamp, campaign efficiency deepdives), do note that these curves can change quite significantly. That impact must be measured at regular intervals and this curve, adjusted.

​

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

​

​

The next application for the MMM is Channel mix optimization. 

This will require you to lay out the budgets for all your marketing channels historically against the marketing-generated SQL $ or WON $ amounts.

​

To visualize this,

​

​

​

​

To be continued...

​

​