A Primer on Out-of-the-Box AI Marketing Mix Models
本文比较了传统计量经济学模型与AI驱动的开源营销组合模型Robyn，从技术、业务和实践角度分析各自的优缺点，为学者和从业者选择预算分配方法提供建议。
Marketing mix modeling (MMM) optimizes budget allocation and determines the return on advertising investment through market response analysis. MMM are vital tools to help marketers define their marketing strategies according to the firm's business and marketing objectives while reducing uncertainty in the decision-making process. As AI and automated MMM out-of-the-box packages gain popularity among marketers, it has become evident there is a theoretical and empirical gap in the understanding of the benefits and inconveniences of these new methods over traditional econometric models. To shed light on these questions, two different models using the same database from a telecommunications firm have been developed and tested using a traditional econometric model and Robyn, an AI-powered open-sourced MMM package from meta marketing science. The research compares both methods’ development processes and subsequent outputs from different perspectives: technical, business, and practical. It shows the advantages and shortcomings of each, providing insightful recommendations for academics and practitioners to navigate through the process of adoption of econometric and AI models for budget allocation decision-making. Econometric models are easy to explain and replicate, while AI complexity from the combination of several methods, their parametrization, and the random initialization of iterations during training, hinders its explainability.