Datalab Seminar
Machine learning and Bayesian modeling for optimization of marketing campaigns
Speaker: Pablo Angulo (Zerel)Date: Thursday, 27 March 2025 - 12:00Place: Aula Gris 1, ICMAT
Abstract:
Media consumption is undergoing rapid changes. Television is losing audience to new media like video online or connected TV, and traditional ways of measuring media consumption do not work any more. Among other problems, media campaign planners are having a hard time trying to reach a certain percentage of their target population, since buying more advertisements in the same media will only reach the same people more times. Our company, The Basement, partners with Fluzo, which provides us with data that can answer such questions. We present a probabilistic Bayesian model based on mixtures of products of Gamma distributions, codename blindshooter, that gives coherent answers to all the planners’ questions, and allows for robust optimization. We will discuss why standard machine learning approaches to this kind of problems are infeasible.
