Forecasting and influencing product returns and fraud rates in a Covid-19 world

新冠肺炎显著加剧了产品退货率高的问题，过去几年来，退货率一直在上升。这对零售商和社会是一个重大挑战，造成经济、社会和生态危害。退货增加了复杂的加工、运输和资源浪费，因为许多产品无法转售，有被填埋的风险。在线购物在封锁期间蓬勃发展，许多零售商延长了退货期。当非必需品零售商重新开业时，产品退货激增。在正常时期代价高昂的问题(例如，洗衣、欺诈性退款、连续退货)在这一大流行时期变得更加严重。由于最近的研究显示，许多顾客会保留他们新的网上购物习惯，这些问题也将继续存在。我们将进行一项消费者调查，访问25家零售商，并与5家零售商密切合作(附上支持信)，以提高我们对消费者行为的理解。然后，我们将使用可解释人工智能(AI)技术在微观和宏观层面对此进行建模，以预测在由Covid-19条件主导的世界中的回报和欺诈率。为了缓解这种情况，我们将制定一套措施，说明预期的效果和对环境的影响。最终目标是帮助零售商在这个充满挑战的时期高效运营和蓬勃发展，避免裁员，从而增加福利方面的财政负担。这个项目将独一无二地将行为研究与开发可解释的人工智能相结合，零售商可以使用它来减轻产品退货的经济和生态影响。遵循我们之前关于产品退货过程的成本和脆弱性的建议(Frei等人，2020b)，许多零售商已经开始通过执行现有的退货政策来改进和控制退货率、欺诈和相关的退货成本。然而，新冠肺炎抵消了这一成功的大部分。零售商最近报告说，产品退货率上升，相关欺诈增加，以及新的欺诈类型。这消耗了很大比例的资源，而此时零售商仍在努力从长期关闭的商店中恢复过来，在新的限制下运营(例如，试衣间关闭和退回的产品需要隔离)，并面临当地封锁的威胁。此外，限制个人行动，再加上恐惧和不喜欢戴口罩，导致在线购物急剧增加，这通常会导致更多退货和更多欺诈。此外，经济和社会危机导致很大一部分人口的财务和心理斗争加剧，这反过来又增加了欺诈的可能性。需要评估这种情况对环境和经济的影响，并确定减轻这些影响的方法。这一挑战解决了零售商的需求，在最近与ECR社区的讨论中表达了这一需求，并与Frei等人(2020a，b)以前确定的研究差距保持一致。

Covid-19 has significantly aggravated the problem of high product returns rates, which havebeen increasing over the last few years. This is a significant challenge for retailers andsociety, causing economic, social and ecological harm. Returns lead to added complexprocessing, transportation and wasted resources, as many products cannot be resold andrisk going to landfill.Online shopping thrived during the lockdown, and many retailers extended their returnsperiods. A surge of product returns arrived when non-essential retailers reopened. Problemsthat are costly in normal periods (e.g. wardrobing, fraudulent refunds, serial returners) havebecome worse in this pandemic period. With recent research showing that many customerswill retain their new online shopping habits, the problems will stay, too.We will conduct a consumer survey, interview 25 retailers and work closely with 5 retailers(letter of support attached) to improve our understanding of consumer behaviours in apandemic. We will then model this at micro and macro levels using explainable artificialintelligence (AI) techniques to forecast returns and fraud rates in a world dominated byCovid-19 conditions. To mitigate this, we will develop a set of measures, indicating theexpected effectiveness and environmental impact.The ultimate goals are to help retailers operate efficiently and thrive in this challenging time,avoiding the need to cut jobs and thereby increasing the financial burden on welfare. Thisproject will uniquely combine behavioural research with the development of explainable AIthat retailers can use to mitigate the economic and ecological effects of product returns.Following the recommendations resulting from our previous work on the costs andvulnerabilities of product returns processes (Frei et al., 2020b), many retailers had startedto make improvements and get a handle on returns rates, fraud and associated costs ofreturns through enforcing their existing returns policies. However, Covid-19 has negatedmost of this success. Retailers have recently reported on increased product returns ratesand increased related fraud as well as new fraud types. This consumes a large proportionof resources, just when retailers are still struggling to recover from a long period of shopclosures, operating under new constraints (e.g. fitting rooms closed and returned productsneeding quarantine) and with the threat of local lockdowns.Furthermore, the restriction of individual mobility in combination with fear and a dislike ofwearing masks have led to a sharp increase in online shopping, which typically leads tomore returns and more fraud. Furthermore, the economic and social crisis is leading toincreased financial and psychological struggles in a large percentage of the population,which in turn increases the likelihood of fraud.The environmental and economic effects of this situation need to be assessed, and waysto mitigate them need to be defined. This challenge addresses the needs of retailers,expressed in recent discussion sessions with the ECR Community that the principalinvestigator attended, and aligns with the research gaps previously identified in Frei et al.(2020a,b).