Segment Discovery: Enhancing E-commerce Targeting

Popular promotions include discounts, bundled offers, free services, etc. By offering these promotions, companies aim to increase revenue and customer base, while also improving customer experience. However, such promotions usually incur a cost and can become unsustainable without any guardrails in place. A popular approach is to target customers with high or low propensity for desired behavior. For example, a retail company is likely to target customers who are at risk of leaving if they want to retain its customers by offering certain incentives. However, previous studies show that this strategy is ineffective and could be detrimental towards the company objectives [2, 6, 7]. Moreover, additional analysis needs to be done for the choice of propensity score threshold for targeting (e.g., target anyone whose propensity to leave is higher than 0.8), because the wrong threshold may lead to sub-optimal outcomes [2]. Each customer responds differently to the same promotion.