Marketing Mix Modeling (MMM) is just linear regression ? or is it ?
There is more to MMM than meets the eye
Marketing Mix Modeling is just linear regression ? or is it ?
MMM is something which looks deceptively simple at first but as you keep peeling the onion layer, you realize how complex it really is.
All the things that your statistics professor would have warned you that could go wrong with Linear Regression does go wrong with MMM. 😅
And funnily what your stat professor or statistics book forgot to mention could go wrong also goes wrong with MMM. 😂
When you build a MMM model, you are bound to face the following problems:
1) MulticollinearityÂ
2) Endogeneity
3) Autocorrelation
4) Omitted Variable Bias
5) Suppression Effect
6) Regression Dilution
7) P>N problem
8) Negative R squared Value
9) Inflated Zero problem
10) Interaction effects and confounding.
And not to mention, you need to also causally prove effect in MMM.
To get MMM right you need two things:
1) Deep statistical knowledge and statistical rigor
2) Deep understanding of Marketing
To get MMM right you need two things:
1) Deep statistical knowledge and statistical rigor
2) Deep understanding of Marketing
Luckily at Aryma Labs, we were lucky to have both of them.
We also continuously perform numerous R&D to enhance our MMM models.
We deeply emphasize on statistical rigor in our MMM models to clients because MMM model is the nucleus in the MMM project.
Saturation curves, ROI estimates, Budget Optimization and scenario planning are all artifacts of the MMM model. If you don't get the model right, none of the downstream outputs are going to accurate.
Link to mine and Ridhima's posts on Linear Regressions and MMM are in resources section.
You can also subscribe to our substack to know everything about MMM or visit the blogs section in our website that has 100+ blogs on MMM.
Resources:
Marketing mix modeling 101 - https://towardsdatascience.com/market-mix-modeling-mmm-101-3d094df976f9
What Marketing Mix Modeling domain can learn from Biostatistics
https://www.linkedin.com/posts/venkat-raman-analytics_marketingmixmodeling-marketingattribution-activity-7165576080299376640-Q76m?utm_source=share&utm_medium=member_desktop
Why linear regression is not about prediction -https://www.linkedin.com/posts/venkat-raman-analytics_statistics-datascience-linearregression-activity-7081862582940168193-FfQu?utm_source=share&utm_medium=member_desktop
Conditional distribution -Â https://www.linkedin.com/posts/venkat-raman-analytics_linearregression-statistics-datascience-activity-7076796588462915584-Z29V?utm_source=share&utm_medium=member_desktop
T value and p value in regression:Â https://www.linkedin.com/posts/venkat-raman-analytics_statistics-datascience-linearregression-activity-7066682786274824193-llNx?utm_source=share&utm_medium=member_desktop
Most Important Assumption Checks of Linear Regression -Â https://bit.ly/3eu47ky
 Linear Regression: the most written topic in Data Science - http://bit.ly/3WLJV1y
How useful is F test in Linear Regression? https://www.linkedin.com/posts/venkat-raman-analytics_statistics-datascience-datascientists-activity-7053981090830594048-COe0?utm_source=share&utm_medium=member_desktop
 Is there really a need to learn Linear Algebra when it comes to Linear Regression? - http://bit.ly/3Q2nuDl
Linear Regression is Just Projection. Always has been -Â http://bit.ly/3WTJdzI
Are least squares and Linear Regression same? -Â https://bit.ly/3CceoOK
Linear regression does not model the raw values of Dependent Variable -Â https://bit.ly/3mGmHKN
Suppression effect:Â https://www.linkedin.com/posts/ridhima-kumar7_marketingmixmodeling-marketingattribution-activity-7043481405292498944-HywE?utm_source=share&utm_medium=member_desktop
Thanks for reading.
For consulting and help with MMM implementation, Click here
Stay tuned for more articles on MMM.