随着中国外卖行业的飞速发展，外卖平台不但为用户提供了获取便利餐饮服务的重要渠道，也为大量骑手提供了就业岗位。作为一个多边平台，外卖平台需要同时兼顾用户和骑手满意度。然而，当前部分平台存在用户和骑手满意度失衡的问题，引发了社会的广泛关注。因此，对于平台而言，准确识别并定量刻画影响用户和骑手满意度的重要因素，并使用户和骑手满意度保持在相对平衡的水平，是关乎平台可持续发展的重要问题。外卖配送中，订单预计送达时间（Estimated Time of Arrival，ETA）是平台的重要决策变量。它不仅是平台给予用户的服务承诺，也是平台对骑手的配送要求。所以，订单ETA会同时影响用户和骑手满意度。目前，平台上存在列表页ETA、点菜页ETA、提单页ETA和订单ETA。本文仅研究提单页ETA和订单ETA对用户和骑手满意度的影响。本文从互动层次和规模层次两个角度研究用户满意度，使用差评率刻画互动层次用户满意度，使用访购率刻画规模层次用户满意度，并使用拒单率刻画骑手满意度。基于平台的观测数据，本文相应构建了三个计量经济学实证模型，以研究外卖配送中订单ETA及相关时间变量与用户和骑手满意度的因果量化关系。由于外卖配送流程复杂，影响用户和骑手满意度的因素众多。在建模时，本文基于对业务流程的梳理和分析，综合考虑了各类因素对用户和骑手满意度的影响，细致地挑选了变量组合，并使用Heckman两阶段法解决互动层次用户满意度模型的内生性问题，以提升模型得出的因果量化关系的准确性。实证分析表明，对于互动层次用户满意度，显式差评率显著高估实际差评率，且相比于订单ETA，用户对延误的敏感程度更大。进入提单页的用户在订餐午高峰时段和非订餐午高峰时段对ETA变化敏感程度不同。不同类型的众包骑手对ETA变化敏感程度不同。平台在决策时需关注用户和骑手的行为差异。此外，在现行订单ETA基础上增加ETA时间后，用户和骑手满意度会出现相反的变化。平台应有平衡意识，避免用户和骑手满意度失衡。本文的研究结论可以帮助平台设计数据驱动的综合考虑用户和骑手满意度的订单ETA决策机制，以促进平台的可持续发展。

With the rapid growth of the food delivery industry, food delivery platforms provide an essential channel for customers to obtain convenient catering services and, at the same time, provide job opportunities to couriers. As a multi-sided platform, food delivery platforms must consider customer and courier satisfaction. However, some platforms‘ imbalance between customer and courier satisfaction recently caught public attention. Therefore, accurately identifying and quantifying critical causal factors of and maintaining customer and courier satisfaction is important to the sustainable development of the platforms.In food delivery services, the estimated time of arrival (ETA) of delivery orders is a crucial decision of the platforms. It is both a service commitment to customers and a delivery requirement for couriers. Therefore, the ETA of delivery orders affects both customer and courier satisfaction. At present, there are ETA at merchant listing pages, ETA at food ordering pages, ETA at order submission pages and ETA of delivery orders on the platforms. This thesis only studies the impact of ETA at order submission pages and ETA of delivery orders on customer and courier satisfaction.This thesis studies customer satisfaction from two perspectives: the interaction level and the scale level. It uses the dissatisfaction rate to describe the interaction level of customer satisfaction, the cart-to-order conversion rate to describe the scale level of customer satisfaction, and the rejection rate of dispatching orders to describe courier satisfaction. The thesis formulates three econometric models to quantify the causal relationship between the ETA of delivery orders and the above customer and courier satisfaction, respectively, using observational data from a food delivery platform. Considering the complexity of food delivery processes and the multitude of factors affecting customer and courier satisfaction, this thesis carefully considers the impact of various factors on customer and courier satisfaction and selects explanatory variables based on the comprehensive review and analysis of the operational processes of food delivery. It also uses the Heckman Two-Stage Method to solve the endogeneity problem in the interaction level customer satisfaction model, to improve the accuracy of the quantitative causal relationship. The results of the empirical analysis show that in the interaction level customer satisfaction model, the explicit dissatisfaction rate (i.e., negative review rate) significantly overestimates the actual dissatisfaction rate, and customers are more sensitive to delays compared to the ETA of delivery orders. Customers who enter the order submission pages react differently to the change in ETA at different times of the day (e.g., lunch peak hours vs. non-lunch hours). Different types of crowdsourced couriers react differently to the change in ETA. Therefore, the platforms should consider the behavioral issues of customers and couriers for better decision-making. Lastly, when increasing the ETA of delivery orders from its current level, customer and courier satisfaction changes in opposite directions, so the platforms should consciously maintain a good balance between the two. The conclusions of this thesis can help platforms design their data-driven ETA of delivery orders decision mechanism that comprehensively considers customer and courier satisfaction, so as to promote the sustainable development of the platforms.