Data‐Driven Approaches to Targeting Promotion E‐mails: The Case of Delayed Incentives

Abstract

T his paper empirically investigates using the e-mail channel to target customers with a delayed incentive promotionspecifically, gift card promotion-and derives data-driven e-mail targeting policies. Gift card promotions are popular across retailers because they incentivize customers to spend more than a fixed expenditure level on regularly priced products by rewarding customers with a gift card to be redeemed against a future purchase. The e-mail channel provides retailers with new sources of customer-level data, which enables better prediction of customers' responsiveness to e-mails (e.g., clicking) and the sales promotion that comes with it (e.g., participation in the promotion). We formulate the retailer's promotion e-mail targeting problem by maximizing two objectives-the promotion's profitability (i.e., profit-based targeting) and e-mail click-through rate (i.e., CTR-based targeting). We also take into account the retailer's promotion budget and exclusivity concerns in targeting e-mails. We use a comprehensive dataset from a Fortune 500 luxury fashion retailer's online channel and utilize both parametric and non-parametric methods to predict customers' response to promotion e-mails. Our data-driven targeting policies improve the promotion's profitability by 5.57% and e-mail CTR by 472.57%, on average, compared to our partner retailer's current e-mail policy. We also find that the CTR-based targeting policy lowers the promotion profitability by, on average, 9.09% compared to the profit-based one. However, the CTR-based policy recuperates the short-term losses in the long-term and increases the long-term profitability by 3.94%, on average, compared to the profit-based targeting policy.

Gender mismatch and bias in people‐centric operations: Evidence from a randomized field experiment

“…We will be able to empirically examine clients' purchase likelihood during our experiment, as well as their repurchase behaviors after the experiment to assess the long-term effectiveness of gender disclosure and client-consultant gender (mis) match. With the emerging literature on promotion optimization and targeting promotions in the OM literature (e.g., Kadiyala et al, 2021), we believe that our results will also contribute to the understanding of the relationship between clients' perceived service quality and the effectiveness of targeting promotions based on gender.…”

Model-Free Approximate Bayesian Learning for Large-Scale Conversion Funnel Optimization

“…Though our framework is general, our numerics focus on an email marketing application, an area explored by others as well. For example, Kadiyala et al (2021) propose a data-driven approach to targeting promotional emails while capturing delayed/long-term incentives. However, in contrast to our sequential learning framework, they focus on a onestep estimate-then-optimize approach.…”

“…This bucketing proves helpful in controlling the size of the MDP state space (Section 3), and as we show, it suffices to explain consumer behavior with high accuracy. For the prediction function, we follow the prior literature (Yoganarasimhan et al., 2023; Kadiyala et al., 2021) and give ourselves the flexibility to capture possibly nonlinear interactions between the features by fitting boosted trees via the package (Chen and Guestrin, 2016). We discuss the implementation details in Section EC.1 in the Supplemental Material.…”

“…However, in contrast to our sequential learning framework, they focus on a one-step estimate-then-optimize approach. We refer the reader to Kadiyala et al. (2021) for a more detailed review of related work.…”