The last mile accounts for somewhere between 40 and 60 percent of total logistics cost depending on the category, the delivery density, and whether you're talking urban or regional. It's been that way for a long time. What's changed in the last few years isn't the underlying cost structure — it's the degree to which that cost structure is legible and therefore optimisable in real time. Carriers have always had route data. Retailers have always had order data. The question is whether those two data streams can be reconciled at speed, and whether the intelligence applied to that reconciliation is good enough to change allocation decisions before the van leaves the depot.
The prevailing model in Australian e-commerce delivery, as of early 2023, is still largely rules-based carrier selection. A retailer's order management system applies a decision tree: order weight and dimensions map to a carrier tier, postcode maps to a service level, and the cheapest option within the service level is selected. That model was defensible when carrier options were limited and rate cards were negotiated annually. It stops being defensible when you have five to eight carrier options for a given corridor, real-time capacity data, variable transit performance by day of week, and consumer delivery preference data that you could be using to inform both the carrier choice and the expected experience. The rules-based model treats all of that as noise. It leaves significant cost and experience improvements on the table every single day.
AI-native delivery orchestration — the category that Shippit has been building into since we backed them in 2022 — treats carrier selection as a continuous optimisation problem rather than a rules lookup. The inputs are carrier capacity signals, historical transit performance at postcode×carrier granularity, consumer preference data where available, retailer service level commitments, and cost parameters. The output isn't just carrier selection — it's a delivery promise the retailer can actually stand behind, communicated to the consumer at the point of purchase, not retrofitted post-despatch. That shift from reactive to predictive fulfilment communication is the piece that most retailers haven't operationalised yet, and it's where a significant portion of customer satisfaction variance in e-commerce lives.
The cross-border dimension adds another layer of complexity that pure domestic carriers don't navigate well. An Australian retailer sending to Singapore faces customs classification, landed duty and tax calculation, carrier hand-off at origin, and a last-mile partner in-country who may have no data-sharing relationship with the origin carrier. The orchestration layer for cross-border fulfilment has to handle that entire chain — not just the allocation decision, but the compliance layer, the consumer-facing tracking unification, and exception management when something goes wrong at the border. That problem is structurally harder than domestic orchestration, and it's the one most relevant to the APAC corridor opportunity we think is significant.
We're not suggesting that logistics orchestration software solves the physical infrastructure constraints in last-mile delivery — road quality, address standardisation, and carrier network density in regional Southeast Asia are real constraints that software alone can't fix. What orchestration software does is ensure that the available infrastructure is used as intelligently as possible, that the consumer experience is managed even when the physical delivery doesn't perform perfectly, and that the cost efficiency of the network improves as data accumulates. Those are meaningful, compounding improvements, and in a category where margins are measured in fractions of a percent, they matter enormously.