Different Bottlenecks, Different Automation: Southeast Asia and China's Parcel Networks

Southeast Asia's parcel delivery industry and China's have both invested heavily in automation over the past several years. The two industries have not built the same thing.

Jason K Hanani

9/16/20265 min read

In China, ZTO Express runs 761 automated sorting lines across 95 hubs covering more than 99% of the country's counties. SF Express operates a single hub, Ezhou Huahu, built specifically to sort 280,000 parcels an hour inside a facility spanning 750,000 square meters, positioned within a 1.5-hour flight of regions generating 90% of the country's economic output. Cainiao's Wuxi facility, once described by the company's own vice president as the largest automated warehouse in China, runs 700 automated guided vehicles moving parcels between stations.

In Southeast Asia, J&T Express operates 127 sorting centers with 75 automated sorting machines between them, alongside a fast-growing fleet of unmanned delivery vehicles. Ninja Van has spent $50 million retrofitting nine of its regional sorting hubs with what the company itself describes as semi-automated conveyor systems, not the fully automated lines built into Ezhou or Cainiao's facilities.

Set side by side, the smaller numbers look like a company or an industry running behind. That reading assumes both industries are solving the same problem. They are not.

What a connected network can build

China's parcel industry can concentrate volume into a small number of very large facilities, because the country's road, rail, and air infrastructure lets that volume be routed to them. Ezhou Huahu exists because a single hub can reach almost the entire country's commercial activity within a short flight. ZTO's 95 hubs cover more than 99% of China's counties without needing a proportionally larger number of facilities, because a connected landmass lets a hub serve a wide radius. Cainiao's automated warehouses, similarly, work because there is enough predictable volume moving through a small number of locations to justify the fixed cost of building them.

The automation follows the geometry. When volume can be concentrated, the thing worth engineering is throughput at the point of concentration: how many parcels can move through one sorting line, one warehouse, one hub, per hour.

What a fragmented network has to solve instead

Southeast Asia's geography does not offer that option. An archipelago cannot be served by a small number of mega-hubs reaching a wide radius, because there is no version of the map that lets one facility stand in for the region the way Ezhou stands in for China. J&T's 127 sorting centers are not a smaller attempt at the same design; they are what a network has to look like when volume cannot be aggregated the way it can on a connected landmass.

The same logic explains why Ninja Van's $50 million automation investment went toward retrofitting nine hubs with semi-automated systems rather than building a single facility at Ezhou's scale. Nine hubs, unevenly spread across markets with different infrastructure and different volumes, is the shape a fragmented network's automation investment takes when the underlying problem is coordinating across many smaller nodes rather than maximizing throughput at one large one.

Why the comparison breaks down as a scoreboard

Counting sorting lines or automated guided vehicles and calling the smaller number less advanced treats both figures as answers to the same question. They are not.

A single mega-hub is the right answer only when a network can route enough volume through one location to justify its fixed cost. That option exists in China because the geography allows consolidation. It does not exist across an archipelago, not because no company has tried, but because the physical routing that makes an Ezhou-scale hub viable does not exist between separated islands. Judging Southeast Asia's many-smaller-hub structure against China's few-large-hub structure means comparing the outputs of two different problems as if they were two attempts at the same one.

The economics point the same direction. Fully automated sorting lines and large-scale robotic warehouses carry high fixed costs that only pay off when a facility handles enormous, predictable volume. Spreading that same capital across many smaller, lower-throughput hubs would not produce a more advanced network. It would produce a worse investment, because the volume needed to justify the expense does not exist at that scale, in those locations. A lower automated-line count in a fragmented network is not necessarily evidence of underinvestment. It can just as easily be evidence of correctly not spending capital where the return would not exist.

What this means for how automation actually gets built

None of this means the two industries are automated to an equal degree in some sense a single number could capture. It means the number that would actually indicate a problem is not "how many automated sorting lines does this network have," but "is this network's automation aimed at its own bottleneck." China's parcel industry automates throughput at concentrated nodes because concentration is available to it. Southeast Asia's automates coordination and last-mile handling across a larger number of smaller nodes because that is the actual shape of its problem.

That distinction matters beyond how these two networks should be judged. It shapes how automation gets built in the first place. J&T operates in both markets, and its own description of its strategy treats this as a deliberate choice rather than a workaround: the company describes actively exploring flexible last-mile fulfillment models tailored to the needs of different markets, rather than exporting one design and expecting it to fit wherever it lands. The technology and expertise move between regions. The shape it takes on arrival does not.

The broader principle extends past logistics. Any operation expanding into a structurally different environment will face the same choice: treat automation, process, or infrastructure as a portable template to be installed unchanged, or treat it as a set of capabilities to be reshaped around whatever bottleneck actually exists on the ground. A lower number, in that second case, is not a sign the template failed to transfer. It is a sign the operation actually looked at the problem in front of it before deciding what to build.

Sources:

Jason Kester Hanani

Operations & Product Professional

contact@jasonkhanani.com