Cluster Delivery Addresses into Geographic Cohorts/Polygons

Design Problem: Given a set of delivery addresses (lat/long points), how would you group them into cohorts or polygons (e.g. for delivery zone planning)? Requirements: Produce geographically coherent, reasonably balanced clusters/zones from raw address coordinates. Design: Discuss geospatial clustering approaches (e.g. k-means/DBSCAN on lat-long, or grid/H3-hex-based binning), then deriving zone boundaries via convex hulls or alpha-shapes around each cluster. Cover trade-offs: density-based clustering (DBSCAN) handles irregular delivery density better than k-means; balancing cluster size for operational fairness (rider load) vs pure geographic compactness; and how to handle addresses on cluster boundaries.