Multiobjective Personalization of Marketing Interventions

Marketing interventions usually affect multiple outcomes of interest. However, finding an intervention that improves all desired outcomes is often rare, creating a trade-off for managers and decision makers. In this paper, we develop a multiobjective personalization framework that identifies personalized policies to balance multiple objectives at the individual level. We apply our framework to a canonical example of multiobjective conflict between sponsored and organic content consumption outcomes. Partnering with vdo.ai, we conduct a field experiment and randomly assign users to the skippable/long and nonskippable/short versions of the same ad. We document substantial substitution between sponsored and organic content consumption; the version that increases sponsored consumption reduces organic consumption. We find that multiobjective personalized policies can significantly improve both sponsored and organic consumption outcomes over single-objective policies. We show that compared with a single-objective policy optimized for organic consumption, there exists a multiobjective policy that increases sponsored consumption by 61% at the expense of only a 4% decrease in organic consumption. Similarly, compared with the single-objective policy optimized for sponsored consumption, there is a multiobjective policy that increases organic consumption by 53% while decreasing sponsored consumption by just 15%.

History: Olivier Toubia served as the senior editor.

Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mksc.2023.0122.