How do you forecast demand for last-mile delivery in smaller cities?

Thin data is the classic problem here — what usually works is leaning on proxy indicators like population density, e-commerce penetration rates, and household income quartiles to build a baseline, then calibrating against comparable smaller cities where you do have data. Tools like Esri's tapestry segmentation or even census ACS data can fill a lot of gaps. Starting with a bottom-up addressable market estimate rather than top-down scaling from metros tends to be more accurate in these cases.