Best way to segment last-mile delivery customers?

For last-mile specifically, I'd actually combine both — start with needs-based to identify the core segments (usually 4-5 works well: time-sensitive, cost-conscious, eco-friendly, convenience-driven, etc.), then layer behavioral data like delivery frequency and return rates on top to validate. We ran a similar project using k-means clustering in Python with behavioral variables first, and it kept collapsing into price vs. speed, which wasn't actionable enough on its own. Needs-based gave us the 'why' that made the segments actually useful for the client.