Delivering By Air: The Arithmetic Of The Last Mile

AFew technologies have been promised as often, or by as many people, as the delivery drone. Since 2013 the trade press has carried photographs of parcels lowered on winches into suburban gardens, together with forecasts that the last mile of the supply chain — the stretch from the local depot to the doorstep, and long the most expensive part of it — would soon be flown rather than driven. Those forecasts borrowed their arithmetic, and with it their assumptions, from an older literature on road freight, in which efficiency is counted per tonne-kilometre and a vehicle's worth rises with the load it can take. The inheritance sits awkwardly on a machine that lifts under three kilograms, travels empty half the time and cannot combine errands. Whether the enthusiasm records a real advance in logistics, or merely the habit of judging a new vehicle by a yardstick built for an old one, is the question that has occupied the field for a decade.

BHalvard Nyberg, a transport economist at the Roros Institute for Freight Systems, has spent much of that decade putting numbers where the photographs had been. Between 2019 and 2023 his group instrumented 8,400 commercial flights in three countries, recording for each one the energy drawn from the battery, the mass carried and the distance flown. The median flight consumed 0.09 kilowatt-hours and delivered a single parcel. A diesel van working the same district drew far more energy in a day, as everyone expected; what interested Nyberg was the figure per parcel, which in the busier districts fell to 0.06 kilowatt-hours, comfortably below anything the aircraft managed, because a van that stops once can hand over eleven items whereas a drone that flies once can hand over one. The vehicle had not grown more efficient, Nyberg observed; the round had grown more crowded, and only one of the two machines was built to profit from that.

CThe account yields a prediction, and Nyberg has tested it. If the aircraft's advantage rests on the scarcity of drops rather than on anything in the machine itself, it should evaporate wherever deliveries cluster, however short the flights become. In the compact centre of Bergen, where a van averaged 9.4 parcels at every stop, substituting drones raised energy use per parcel by a fifth; across the scattered farms of the same county the identical substitution cut it by 61 per cent. Reductions of more than thirty per cent were confined to rounds on which a van collected fewer than 1.4 parcels at each stop. The asymmetry is awkward for the industry, since the customers it most wants are the ones packed into cities. Nyberg presents the result, characteristically, as a boundary around one method of accounting rather than a verdict on the aircraft.

DPriya Ramanathan, who directs the Kealey Centre for Distribution Studies, accepts the arithmetic and rejects what is made of it. Counting the energy of a flight, she argues, measures a vehicle when the interesting object is a network: an operator able to reach a village by air need not keep a depot near it, and the emissions of that depot, of the lorry that restocks it and of the staff who drive to work there appear nowhere in a ledger built from single flights. Her own modelling of a rural health service found that replacing two of five district stores with one hub served by drones cut emissions across the whole system by 34 per cent, even though the flights were the least efficient link in the chain. She concedes two things. Her comparison covered one service carrying blood samples, whose consignments are light and urgent in a way that most parcels are not; and the annoyance caused to residents, which she counts as a cost, was inferred from complaint counts, because no agreed measure of it exists.

EThe argument has consequences outside the journals. Rules adopted in 2021 in several jurisdictions permitted flights beyond the operator's line of sight along designated corridors, and applications for those corridors are now judged partly on the environmental case their sponsors make; councils in three cities have refused permission on grounds of noise. Nyberg points out, uncomfortably for his own position, that nothing in his accounting bears on whether residents will tolerate the aircraft, since a favourable energy balance and an intolerable noise are not alternatives. Ramanathan replies that the two questions cannot be kept apart in practice, because regulators read efficiency figures as evidence about acceptability, and read the absence of such figures as evidence against it. Both regret the same habit: the publication of results from single showcase corridors, chosen and staffed for the purpose, rather than counts of what ordinary operations achieve across a full year.

FA settlement of sorts is taking shape, in which the two accounts are assigned to different tasks rather than ranked against one another. Where a consignment is light, urgent and bound for an address a van reaches slowly, the aircraft wins on any measure either researcher accepts; where parcels are heavy, tolerant of delay and destined for doors within a few metres of one another, the van wins by the same reasoning. What neither can settle is whether the pattern of demand that makes the first case profitable will survive the arrival of the service itself, since households supplied within the hour may order more often and in smaller quantities, and no figures on that yet exist. What both reject is the assumption behind the photographs of 2013: that a new vehicle must be either an improvement on the one before it or a toy.