Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Advantages of a Bayesian approach to Media Mix Modeling

Ander, Jesper LU and Bergman, Gabriel (2025) In Master's Theses in Mathematical Sciences FMSM01 20251
Mathematical Statistics
Abstract
This study investigates the application of Bayesian methods in Media Mix Modeling (MMM) and compares their performance against more traditional regression models. We develop multiple MMM's to quantify the impact of media and non-media variables on a company's performance, including: (1) a Bayesian regression model incorporating prior knowledge and posterior inference, and (2) simpler regression models with uncertainty quantification. All models account for standard marketing dynamics such as carryover and saturation effects. Using data from a single company, we evaluate model performance in terms of predictive accuracy, interpretability, and parameter uncertainty. Results show that the Bayesian model provides richer uncertainty... (More)
This study investigates the application of Bayesian methods in Media Mix Modeling (MMM) and compares their performance against more traditional regression models. We develop multiple MMM's to quantify the impact of media and non-media variables on a company's performance, including: (1) a Bayesian regression model incorporating prior knowledge and posterior inference, and (2) simpler regression models with uncertainty quantification. All models account for standard marketing dynamics such as carryover and saturation effects. Using data from a single company, we evaluate model performance in terms of predictive accuracy, interpretability, and parameter uncertainty. Results show that the Bayesian model provides richer uncertainty quantification through posterior distributions, capturing deeper parameter uncertainty. Additionally, empirical confidence intervals derived from repeated data resampling are narrower for the Bayesian model compared to traditional models, indicating greater robustness and stability in estimation. The findings highlight trade-offs between model complexity, interpretability, and robustness, providing practical guidance for analysts and decision-makers deploying MMM in real-world scenarios. (Less)
Please use this url to cite or link to this publication:
author
Ander, Jesper LU and Bergman, Gabriel
supervisor
organization
course
FMSM01 20251
year
type
H2 - Master's Degree (Two Years)
subject
keywords
MMM, Bayesian statistics, Media Mix Modeling, Hamiltonian dynamics, MCMC, Markov Chain Monte Carlo, Meridian, linear regression
publication/series
Master's Theses in Mathematical Sciences
report number
LUTFMS-3545-2025
ISSN
1404-6342
other publication id
2025:E102
language
English
id
9213808
date added to LUP
2025-10-16 12:56:33
date last changed
2025-10-16 12:56:33
@misc{9213808,
  abstract     = {{This study investigates the application of Bayesian methods in Media Mix Modeling (MMM) and compares their performance against more traditional regression models. We develop multiple MMM's to quantify the impact of media and non-media variables on a company's performance, including: (1) a Bayesian regression model incorporating prior knowledge and posterior inference, and (2) simpler regression models with uncertainty quantification. All models account for standard marketing dynamics such as carryover and saturation effects. Using data from a single company, we evaluate model performance in terms of predictive accuracy, interpretability, and parameter uncertainty. Results show that the Bayesian model provides richer uncertainty quantification through posterior distributions, capturing deeper parameter uncertainty. Additionally, empirical confidence intervals derived from repeated data resampling are narrower for the Bayesian model compared to traditional models, indicating greater robustness and stability in estimation. The findings highlight trade-offs between model complexity, interpretability, and robustness, providing practical guidance for analysts and decision-makers deploying MMM in real-world scenarios.}},
  author       = {{Ander, Jesper and Bergman, Gabriel}},
  issn         = {{1404-6342}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master's Theses in Mathematical Sciences}},
  title        = {{Advantages of a Bayesian approach to Media Mix Modeling}},
  year         = {{2025}},
}