Causal Inference in Marketing: A Practical Toolkit for Panel Data

The global advertising market is roughly US$1.1 trillion, with digital channels accounting for most of that activity. Marketing measurement therefore increasingly depends on complex data environments: high-dimensional covariates, machine-learning systems, continuous treatments, platform reporting constraints, and organisational pressure to turn evidence into decisions. These settings create opportunities for richer causal analysis, but they also raise difficult questions about validity, uncertainty, diagnostics, reproducibility, and whether an estimated effect is useful for the decision at hand.

Volume 2 of Causal Inference in Marketing: A Practical Toolkit for Panel Data carries the framework of Volume 1 into the advanced and operational half of the book. It extends the core panel toolkit into machine learning, high-dimensional adjustment, continuous and nonlinear treatment settings, threats to validity, inference, diagnostics, applied marketing workflows, data and measurement systems, reproducibility, and open problems. The emphasis throughout is on applying causal principles under the constraints of real marketing data and real organisational settings.

Key Features:

• Develops machine-learning and high-dimensional methods for panel data, including orthogonalisation, cross-fitting under panel dependence, heterogeneous treatment effects, policy learning, regularisation, and double selection.
• Provides a diagnostics and inference playbook covering pre-trends, placebos, sensitivity analysis, bootstrap and randomisation inference, multiplicity, weak instruments, and uncertainty communication.
• Connects advanced causal methods to marketing applications, including media mix models, geo-experiments, platform data, pricing, promotions, customer lifetime value, retention, measurement systems, and reproducible evidence production.

Written for data scientists, marketing analysts, econometricians, and applied researchers, this volume is intended for readers who are comfortable with regression and applied statistics and who want to extend causal design into robust implementation, diagnosis, and reporting. Volume 1 develops the foundations, including potential outcomes, design-based thinking, difference-in-differences, event studies, synthetic control, factor and matrix methods, dynamics, heterogeneity, interference, and spillovers.

Charles Shaw is a Data Science Director at WPP Media, where he leads econometric measurement and optimisation for global brands. His work focuses on causal inference, econometric measurement, Bayesian modelling, machine learning, and marketing effectiveness. He develops applied frameworks for privacy-constrained attribution, media incrementality, platform effects, dynamic pricing, and scalable causal workflows in commercial settings.