After a disaster, the demand for relief is unknown. Roads may be blocked, populations displaced, damage reports delayed or absent. The planner must dispatch trucks with supplies to locations where the need might be zero or might be critical. The information arrives as the operation proceeds — the act of delivering reveals the demand.
The paper introduces a drone-impact metric: the ratio of performance with drones to the best achievable performance without them. This metric, separate from the competitive ratio against a full-information optimal, reveals something non-obvious. Drones help even when information is incomplete — not primarily because they deliver supplies, but because they scout. A drone that flies ahead of the truck convoy and discovers that a village is intact saves the convoy from a wasted trip. A drone that discovers a destroyed bridge reroutes the convoy before it arrives.
The reconnaissance function of drones provides value that is independent of the information quality of the initial damage report. Even with no initial information about demand, the drone's surveillance reduces the operational cost by turning the unknown-demand problem into a progressively-known-demand problem. The convoy makes better decisions because it has a moving information frontier rather than a fixed one.
The two metrics — competitive ratio and drone-impact — measure different things. The competitive ratio measures how much you lose from not knowing everything upfront. The drone-impact measures how much you gain from learning as you go. The first is about the cost of ignorance. The second is about the value of scouting.