Prospective marketing meta-analysis and a novel web-based media-mix modeling experiment

Abstract

Prospective meta-analysis, pioneered in the biomedical field, is the meta-analysis of multiple studies conducted using similar protocols and under similar conditions. To eliminate bias, the inclusion of individual studies in the meta-analysis is agnostic of the findings of the individual experiment. In this thesis, I adapt prospective meta-analysis for use in the field of marketing science. Specifically, I design and create a database for prospective marketing meta-analysis that encourages and facilitates international collaboration and scale-up of marketing science studies and use this platform as the basis for a novel web-based media-mix modeling experiment that aims to model the relative effects of a variety of media. I detail the design and implementation of this web-based media-mix modeling experiment, which introduces the use of a browser extension to modify the media experience for test subjects based on their responses to pre-survey questions. I present preliminary results from a 50-user trial run of the system and analyze improvements and next steps, both for the current experiment and scale-up for future studies to include in the meta-analysis.