Profit uplift modeling for direct marketing campaigns: approaches and applications for online shops

In order to select “best” customers for a direct marketing campaign, response models are widespread: a sample of customers receives an ad, a catalog, a sample pack, or a discount offer on a test basis. Then, their responses (e.g., website visits, conversions, or revenues) are used to build a predictive model. Finally, this model is applied to all customers in order to select “best” ones for the campaign. However, up to now, only models that reflect website visits, conversions, or revenues have been proposed. In this paper, we discuss the shortcomings of these traditional approaches and propose profit uplift modeling appoaches based on one-stage ordinary regression and random forests as well as two-stage Heckman sample selection and zero-inflated negative binomial regression for parameter estimation. The new approaches demonstrate superiority to the traditional ones when applied to real-world datasets. One dataset reflects recent discount offers of a large online fashion retailer. The other is the well-known Hillstrom dataset that describes two Email campaigns.

Uplift modeling VS conventional predictive model: A reliable machine learning model to solve employee turnover

Employee turnover is the loss of talent in the workforce that can be costly for a company. Uplift modeling is one of the prescriptive methods in machine learning models that not only predict an outcome but also prescribe a solution. Recent studies are focusing on the conventional predictive models to predict employee turnover rather than uplift modeling. In this research, we analyze whether the uplifting model has better performance than the conventional predictive model in solving employee turnover. Performance comparison between the two methods was carried out by experimentation using two synthetic datasets and one real dataset. The results show that despite the conventional predictive model yields an average prediction accuracy of 84%; it only yields a success rate of 50% to target the right employee with a retention program on the three datasets. By contrast, the uplift model only yields an average accuracy of 67% but yields a consistent success rate of 100% in targeting the right employee with a retention program.