What importing from China really costs, landed in Australia.

Landed cost is what one sellable unit costs by the time it's on your shelf, everything included. Here's the anatomy for any business importing into Australia, line by line, whether it's a new product's opening batch or a routine stock reorder. It's structural, not quoted: your numbers depend on your product, your volumes and current rates, and anything we estimate on a real program is labelled as an estimate.

The lines that make the number

Unit price: the factory quote, at a named incoterm (see our incoterms explainer; quotes on different terms aren't comparable). Tooling: moulds and fixtures, paid up front and honestly amortised across realistic volumes, not optimistic ones. Freight: sea for margin, air for speed; the gap between them is large enough to change a product's viability, and peak-season rates move. Insurance: small, skipped by people who later wish they hadn't.

Duty: many Chinese-origin goods enter Australia at reduced or zero duty under the China–Australia Free Trade Agreement, but only with correct origin documentation, and product categories differ: classify properly and verify current rates with your broker. GST: applied on the landed value (goods plus freight, insurance and any duty), claimable for GST-registered businesses but real for cash flow. Clearance and port charges: customs brokerage, terminal and handling fees. Local delivery: port to your warehouse or 3PL. And beneath all of it, the quality line: inspection costs, and the expected cost of the rework or rejects that inspections exist to prevent. It's the line that never appears in the optimistic spreadsheet.

The mistakes that flatter the spreadsheet

Comparing unit prices across different incoterms. Amortising tooling over the volume you hope for rather than the volume you've ordered. Budgeting sea freight, then air-freighting the launch batch because the schedule slipped. Forgetting GST applies to the freight too. Treating inspection as an optional cost instead of insurance priced at a fraction of the shipment it protects. Taking a tidy generated estimate for a quote. Each one makes the early spreadsheet prettier and the first reorder uglier.

The agent-native lens: an estimate that shows its working

When the buyer sends an assistant. A landed-cost question put to an AI assistant can come back as a confident total built from general rates it can't vouch for on the day you ship. Treat that as a first draft of the model, never a quote. Have it show every line in the stack above, mark each as quoted, estimated or assumed, and date the rates it used; then ask what it left out, since the quality line is the usual casualty. Check that GST sits on the freight, insurance and duty as well as the goods; a quick total can miss that. If you supply other businesses, the same holds for your quotes: dated, itemised and on a stated term, they drop into a buyer's model without guesswork.

When the business runs on agents. The landed cost stops being a spreadsheet built once before tooling and becomes a model checked against every shipment. Agents can post each real charge as it arrives (factory invoice, freight bill, broker's entry, duty and GST) against the line it was estimated on, and report the variance per unit. They can recalculate tooling amortisation on the volume actually ordered and flag the shipment where air quietly replaced sea. As more invoices arrive as structured data rather than PDFs, that reconciliation gets cheaper to run. People set the target margin, decide what a variance means for price or range, and approve the change. Every figure should trace back to the document it came from.

Working it properly

Build the landed cost before committing to tooling, pressure-test it at half your forecast volume, and keep it live: re-run it when anything material changes (freight mode, order size, term) and reconcile it against what each shipment actually cost. A model checked against real invoices gets more accurate with every order; one built once only gets older. On our programs this model is a standing deliverable of the production & freight practice, and the feasibility version appears even earlier, in strategy & research, because the kindest time to learn a product doesn't work commercially is before it exists.