E-commerce in India is at a nascent stage with players innovating and investing in different capabilities to improve speed, cost, and reliability. To offer a wide selection of customers, players onboard multiple sellers in their marketplace. In a city logistics environment’s seller pickup leg (first mile logistics), the challenge in planning and procurement of fleet for these sellers is accommodating ‘variability’ from the plan. The key complexities in predictive and systemic fleet planning for sellers in this network are dynamicity of touchpoints, high variation in seller load giving patterns and their geographical spread. In addition to these network characteristics, intrinsic orchestration to prevent any capacity bottlenecks in the supply chain adds to the volatility of actual load/ seller nodes from the plan. Multiple fleet planning decisions and optimizations are rule-based and improved by feedbacking. This paper proposes to fill that gap by investigation of seller characteristics and drivers of fleet planning and arrive at a segmentation for the first - mile seller network. Various approaches are discussed (load based segmentation, load and location-based clustering). Each of these planning methods pointed to the insight that catering a few infeasible sellers though alternate mediums could improve the unit economics of seller pickups. We will discuss approaches to alternate mediums of pickups from these sellers and insights from proof of concepts.