Hawkes Processes in Marketing

Loading...
Thumbnail Image
Degree type
PhD
Graduate group
Marketing
Discipline
Marketing
Operations Research, Systems Engineering and Industrial Engineering
Subject
Advertising
Privacy
Stochastic Processes
Funder
Grant number
License
Copyright date
01/01/2025
Distributor
Related resources
Author
Laurino dos Santos, Henrique
Contributor
Abstract

This dissertation offers a detailed investigation of the use of univariate Hawkes processes to modelconsumer behavior in digital marketing applications. Datapoints in these contexts alternate long periods of sparsity and short bursts of intense activity, presenting many challenges for traditional models. Hawkes processes, meanwhile, provide a flexible framework to capture complex dynamics of dormancy and momentum buildup, but require both adaptation and innovation to appropriately describe consumer data. This dissertation elaborates on their properties, connects them to models in the extant literature, develops new parametric forms, and benchmarks their use in both simulation studies and empirical applications. Special attention is given to empirical estimation under strict privacy laws, a novel and pressing use-case. The model is then extended to the derivation of optimal advertising policies under various market conditions. Ad delivery scheduling, dynamic bidding, and constrained bidding problems are studied analytically, in simulation, and empirically. Once more, special attention is given to the positive and negative impacts of privacy laws for both consumers and advertisers. This work highlights the empirical and theoretical advantages of Hawkes processes as models of dynamic human behavior, as well as both the costs and benefits of emerging privacy policies.

Advisor
Fader, Peter, S.
Van den Bulte, Christophe
Date of degree
2025
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation