Test & Roll: Profit-maximizing A/B tests

Maxine Fang




Assistant Professor of Marketing

Project Summary

This summer, I was able to work with Professor Berman and Professor Feit to build an R Package and R Shiny web app implementing the test & roll methods for A/B tests. Professor Berman and Professor Feit developed the Test & Roll model which can determine optimal sample sizes for maximizing profit. Under test & roll, an experimentation process for A/B testing, a subset of customers is randomly assigned to a treatment and customer response data is collected in the “test” stage. In the “roll” stage that follows, the better treatment is deployed to all remaining customers based on the test results.

From this project, I learned how to create and document an R package as well as develop a web app using R Shiny. During the first half of the project, I also completed many exercises working with ggplot, efficient matrix calculations, and general R programming. I believe I have emerged from this project with a much deeper understanding of the R language and its capabilities.

In addition to the technical aspects of the project, I also enhanced my communication and design skills. Through the individual weekly meetings, I learned how to plan meeting agendas and discuss project timelines. The PURM project also exposed me to GitHub, a platform for portfolio-building and project collaboration. While designing the R package and web app, I also learned to consider end-user preferences when it came to usability and clarity. This included tasks such as generating help documents, documenting code, and creating effective visual hierarchy in the website design.

Overall, working on Test & Roll was a rewarding experience that has deepened my understanding of programming, app development, and the research process as a whole.


To see my poster, visit Penn Presents: https://presentations.curf.upenn.edu/poster/test-roll-profit-maximizing-...