In the digital economy era, take-out shops relying on the online-to-offline e-commerce platform have gathered in central cities on a large scale, forming large-scale food processing spaces—"Takeaway Factories"— to serve immediate local catering needs. In addition to the impact of the platform economy, consumer demand, and space costs, riders are an important and dynamic factor that promote the final formation of a "Takeaway Factory". Based on the analysis framework of virtual and real-space interactions using the participatory survey method, the specific roles of crowdsourcing and special delivery riders in the formation of "Takeaway Factory" was elucidated. Riders, as the core intermediaries of virtual agglomeration that leads to physical agglomeration, play a role at different stages and spatial scales. In the initial stage, because dedicated takeout shops tend to prefer low-cost, high-demand urban gap spaces, they face challenges in matching with special delivery riders until they are included in the special delivery distribution station. Thus, crowdsourcing riders with free-movement attributes are their first choice. Crowdsourcing riders have a strong preference for store clusters and select food through the two tools of the regional order heat map and grab order hall, bringing about differences in distribution efficiency and promoting the phenomenon of large regional differentiation of relatively concentrated and dispersed stores within the city scale. In the middle stage, in the face of the uneven distribution efficiency of different areas in the city, the platform divides the size of the stations according to the size of the shops in different areas, and matches the special delivery riders of the corresponding scale, which eliminates the uneven distribution efficiency of different areas to a certain extent, and promotes the distribution efficiency of different sections to reach a basic balance. With the emergence of special delivery systems, the special delivery riders are distributed in the station area through the delivery system, which effectively alleviates the distribution pressure of the relatively dispersed, inferior-location stores and promotes the balance of the distribution efficiency of different sections. In the final stage, the special delivery and crowdsourcing riders in the area jointly promote the agglomeration of shops in the area, and ultimately promote the formation of a more polarized takeaway factory. Crowdsourcing riders continue to promote agglomeration from the initial stage, whereas the emergence of special delivery riders promotes both equilibrium and agglomeration. With the increasing maturity of the special delivery and crowdsourcing system, the time-space behavior of different types of riders will eventually promote the completion of the high-density and large-scale agglomeration of takeaway shops in the area (distribution site) through virtual tools, such as site order thermal maps, order dispatching systems, resident points, and order-grabbing halls.

1 骑手群体分为专送和众包2种类型。专送骑手负责接收特定区域（专送配送站）内的店铺订单，接受系统派单和配送站长的管理，站长可以干预派单。众包骑手负责接收任意区域的店铺订单，并在系统展示的多个订单中自主抢单，无站长干预订单分配。因为工作模式的差异，促使2类骑手在“外卖工厂”集聚的不同时间阶段、空间尺度中发挥着作用，所以加以区分。

2 单量分布和店铺分布并非完全等同，但两者之间有着非常密切的关系。通过笔者在参与式调查中的实地踏勘体验和与骑手的多次深度访谈，发现店铺集聚区往往就是订单集聚区，两者之间有着高度的重叠，并不存在订单较多、店铺较少或订单较少、店铺较多的区域，所以将两者近似等同。

赵奕涵：论文撰写、数据收集与处理、分析框架构建与图表绘制；

罗震东：研究方法和分析框架构建指导、论文审阅与修改；

张 佶：研究思路与方法指导、论文审阅。