Kisilevich

How to optimize the media mix using saturation curves and statistical modelsPhoto by Joel Filipe on UnsplashMarketing Mix Modeling (MMM) is a data-driven approach that is used to identify and analyze the key drivers of the business outcome such as sales or revenue by examining the impact of various factors that may influence the response. The goal of MMM is to provide insights into how marketing activities, including advertising, pricing, and promotions, can be optimized to improve the business performance. Among all the…

How to evaluate uncertainty with Validity, Sharpness, Negative Log-Likelihood, and Continuous Ranked Probability Score (CRPS) metricsPhoto by Santiago Lacarta on UnsplashMany real-world problems require uncertainty estimation for future outcomes for better decision-making. However, most state-of-the-art machine learning algorithms are capable of estimating only a single-valued prediction which is usually a mean or a median of the conditional distribution which suppose to match the real outcome well. What the single-valued…

Capturing non-linear advertising saturation and diminishing returns without explicitly transforming media variablesPhoto by Pawel Czerwinski on UnsplashThe established approach among marketers for modeling marketing mix is to apply linear regression models which assume the relationship between marketing activities such as advertisement spend and the response variable (sales, revenue) is linear. Prior to modeling, media spend variables should undergo two necessary transformations to properly capture the carryover effect…

How to fit a SciPy Linear Regression and call R Ridge Regression from Python using RPy2 InterfacePhoto by Will Francis on UnsplashThe most common approach in Marketing Mix Modeling(MMM) is to use Multiple Linear Regression, which finds a linear relationship between a dependent variable such as sales or revenue, and independent variables including media advertisement channels like TV, Print, and additional variables like trend, seasonality, holidays. One of the questions marketers might have is what effect each media…

Building MMM models using tree-based ensembles and explaining media channel performance using SHAP (Shapley Additive Explanations)Photo by Adrien Converse on UnsplashThere are many ways one can build a marketing mix model (MMM) but usually, it boils down to using linear regression for its simple interpretability. Interpretability of more complex non-linear models is the topic of research in the last 5–6 years since such concepts as LIME or SHAP were proposed in the machine learning community to explain the output of a…