Marketing Mix Modeling (MMM)

Marketing Mix Modeling (MMM)

Problem Statement:

Marketing Mix Model for ProSiebenSat.1 Media SE is a European mass media company, based in Germany.


This framework uses spends or impressions values of different media channels like SEA Impressions, Affiliate Impression, Paid/Owned Social Media Impressions, Display Spends and TV Spends to determine which marketing channels are currently providing the best returns and to create a strategy for reallocating budget euros in order to increase revenue.

Our 4-step approach to provide optimized spends across marketing channels is:

  1. EDA and required variable transformations
  2. Pre-Modeling: Adstock transformations are performed to account for short-term and medium-term
    memory impact of the TV advertisements on revenue. Interaction-models are created to capture the TV driven part of the other channel spends. CCF (Cross Correlation Function) analysis to identify lags (between revenue and TV expenditures) at which the impact is highest (echoes) and the lags till which the impact persists (extend of memory).
  3. Modeling: GAM (Generalized Additive Model) is used to fit a piece-wise polynomial curve basis relationship between the dependent and the independent variables.
  4. Post-Modeling: Redistribution of contributions of different channels to estimate accurate ROIs.

Techniques Used:GAM (Generalized Additive Model), Regression Analysis, DLM, R, MS Excel