ICYMI (link in resources), in my previous posts I had highlighted why one can’t RCT MMM.
I will just recap two points from that post that will provide a perfect segue for the topic that we are going to cover today.
◾ Marketing Reality is complex and RCT can’t control for all variables
When we talk about MMM. It is all about understanding marketing effectiveness i.e. what all variables affect the KPI and what all interactions between them also affect the KPI.
One can’t Randomize marketing strategies and neither can one control for all the variables and their interactions.
◾RCTs and Experimentation are uni-variable in nature
Most RCTs and Experimentation test the effect of one variable on the KPI. In reality Marketing is a multivariable problem.
So with these two points recapped, lets understand first the concept of Principle of Marginality.
📌 Principle of Marginality
The principle of Marginality states that it is wrong to interpret/test/estimate only the main effects of independent variables while there are interaction effects between the independent variables.
Similarly, it would be wrong to interpret/test/estimate only the interaction terms of the independent variables while dropping the main effects of the independent variables.
Let’s see this in an example.
Assume the below as a well specified model.
Y = β0 + β1* X1 + β2*X2 + β3* X1X2 + e
Now, if we were to write the equation as Y = β0 + β1* X1 + β2*X2 + e, then we would be violating the principle of marginality since we have totally ignored the interaction effects between X1X2.
Similarly, the equation Y = β0 + β1* X1X2 + e, too would be wrong as we have totally ignored the main effects of the independent variables X1 & X2.
As you can see even in a multi variable model it is so easy to violate principle of marginality.
One can only imagine how gravely we would be violating principle of marginality in case of using RCT to validate Marketing Mix Models (MMM) !!
The best way still to validate MMM is through DID (check link in resources).
Resources:
Use Experimentation to validate your MMM models, not calibrate it.
https://www.linkedin.com/posts/venkat-raman-analytics_marketingmixmodeling-experimentation-statistics-activity-7155901631820177408-tCs4?utm_source=share&utm_medium=member_desktop
Why Difference in Difference (DID) Experimentation is the ideal way to validate your MMM model.
https://www.linkedin.com/posts/venkat-raman-analytics_marketingmixmodeling-marketingattribution-activity-7157623304307056642-cloZ?utm_source=share&utm_medium=member_desktop
