Application of uplift modeling to enhance the effectiveness of targeted marketing campaigns in retail

The growing competition in the retail and e-commerce markets requires companies to adopt more precise approaches to planning marketing campaigns and personalizing customer communications. Existing response prediction models fail to isolate the effect of marketing interventions from natural purchasing behavior, leading to irrational budget spending and complicating the objective evaluation of campaigns. Uplift modeling emerges as a solution – an approach that enables causal effect assessment at the individual consumer level and identifies audience segments most responsive to communications. This article presents a comparative analysis of uplift modeling methods to select the most effective one for evaluating marketing impact. The study examines five methods (S-Learner, T-Learner, Class Transformation, X-Learner, and R-Learner) using the open Lenta Uplift Modeling Dataset provided by the Lenta retail chain during the BigTarget hackathon in collaboration with Microsoft. Model performance was evaluated using specialized metrics (Uplift@k, Qini AUC, Uplift AUC, Weighted Average Uplift, Average Squared Deviation). The analysis reveals the strengths and weaknesses of each approach and identifies the top-performing method for this dataset.

1. Popov A., Iakovleva D. Adaptive Look-Alike Targeting in Social Networks Advertising. Procedia Computer Science. 2018;136:255–264. https://doi.org/10.1016/j.procs.2018.08.264

2. Lo V.S.Y. The true lift model: A novel data mining approach to response modeling in database marketing. ACM SIGKDD Explorations Newsletter. 2002;4(2):78–86. https://doi.org/10.1145/772862.772872

3. Hansotia B., Rukstales B. Incremental value modeling. Journal of Interactive Marketing. 2002;16(3):35–46. https://doi.org/10.1002/dir.10035

4. Rzepakowski P., Jaroszewicz S. Decision trees for uplift modeling with single and multiple treatments. Knowledge and Information Systems. 2012;32(2):303–327. https://doi.org/10.1007/s10115-011-0434-0

5. Künzel S.R., Sekhon J.S., Bickel P.J., et al. Metalearners for estimating heterogeneous treatment effects using machine learning. Proceedings of the National Academy of Sciences of the United States of America. 2019;116(10):4156–4165. https://doi.org/10.1073/pnas.1804597116

6. Devriendt F., Moldovan D., Verbeke W. A literature survey and experimental evaluation of the state-of-the-art in uplift modeling: A stepping stone toward the development of prescriptive analytics. Big Data. 2018;6(1):13–41. https://doi.org/10.1089/big.2017.0104

7. Kane K., Lo V.S.Y., Zheng J. Mining for the truly responsive customers and prospects using true-lift modeling: Comparison of new and existing methods. Journal of Marketing Analytics. 2014;2(4):218–238. https://doi.org/10.1057/jma.2014.18

8. Jaskowski M., Jaroszewicz S. Uplift modeling for clinical trial data. In: Proceedings of the 29th International Conference on Machine Learning (ICML 2012), 26 June – 01 July 2012, Edinburgh, Scotland, UK. 2012. URL: https://home.ipipan.waw.pl/sj/pdf/logsemi.pdf

9. Nie X., Wager S. Quasi-oracle estimation of heterogeneous treatment effects. Biometrika. 2021;108(2):299–319. https://doi.org/10.1093/biomet/asaa076

10. Gubela R.M., Lessmann S., Jaroszewicz S. Response transformation and profit decomposition for revenue uplift modeling. European Journal of Operational Research. 2020;283(2):647–661. https://doi.org/10.1016/j.ejor.2019.11.030