During an on-site trial at a convenience store, Zhiyuan Robotics accidentally mistook customers’ uneaten cooked food for garbage—a minor misstep that nonetheless underscored the core purpose of the test: to demonstrate how robots can be integrated into commercial settings. As the embodied intelligence industry shifts toward commercialization in response to listing regulatory requirements, the lack of sufficient real-world scenario data has emerged as a key bottleneck to large-scale deployment. To address this, Zhiyuan launched a crowdsourced data collection platform called Mifengpai, drawing on the Index model used by similar U.S. projects. The company aims to build shared data infrastructure for the entire embodied intelligence sector and to alleviate heavy-asset pressures through a city-partner franchise and contract manufacturing model. However, this approach faces several challenges: difficulty guaranteeing income for data collectors, limited access to scarce professional scenarios, conflicts of interest between the platform and its users, a small and fragmented downstream buyer base, and pressure on data pricing. Currently, Zhiyuan has paused its IPO process while continuing to advance robot deployment in real commercial scenarios such as convenience stores and theme parks. At the same time, it is seeking to break through data bottlenecks through Mifengpai, attempting to dilute costs through scale and explore a viable path to commercialization.