The rapid proliferation of e-commerce has driven a surge in last-mile deliveries within urban areas, exacerbating greenhouse gas and local emissions. Crowdshipping has emerged as a sustainable alternative to traditional truck-based logistics, offering the potential to mitigate environmental impacts. To expand crowdshipping services, policymakers need an understanding of the key drivers of crowdshipping. However, little is known about the characteristics of crowdshipping, particularly its trip production factors, due to the novelty of the services and limited real-world data. This study utilizes data from crowdshipping trips in Atlanta between 2015 and 2018, coupled with socio-demographic and built environment data, to shed light on the factors influencing crowdshipping trip production. Through data analysis, we revealed crowdshipping trip patterns, preferred delivery times, and package sizes in Atlanta. Moreover, we employed Machine Learning techniques, namely Decision Tree, Random Forest, and Gradient Boosting, to elicit the significant factors influencing crowdshipping trip production. The feature importance analysis indicates that areas with a higher concentration of food and home appliance hubs are associated with more crowdshipping trips. Furthermore, disadvantaged areas with limited access to public transit and shipping services exhibit higher crowdshipping activities. Additionally, areas with a higher proportion of low-income, more educated, or younger individuals incur more crowdshipping trips. The study provides valuable insight for policymakers aiming at promoting crowdshipping as an alternative for last-mile delivery. In this regard, we recommend three policies, upon the insights from the analyses, including improving crowdshipping in disadvantaged areas, awareness/advertisement, and partnership of crowdshipping services with businesses.