How Many Data Points Are Needed for Stable and Accurate Coefficient Signs in Marketing Mix Models (MMM)? A Bootstrap Approach

Marketing Mix Modeling (MMM) is widely used for budget allocation and decision-making, yet one of its persistent challenges is the issue of sample size. Since MMM is typically constrained by the availability of historical data—often limited to a few years—understanding the impact of sample size on model estimation is crucial. In this study, we examine how the number of observations influences the stability of coefficient signs in an already-delivered client model. We employ a semiparametric block bootstrap approach to evaluate the proportion of models with correct coefficient signs across different sample sizes, ranging from 20 to 36 months.