The vocabulary of agent shipping, defined once

Six definitions plus the four terms most often used interchangeably. That figure is the reason this reading exists, and everything below traces back to the same 96-entry sample collected on 2026-10-05.

Most confusion in this subject comes from four terms used interchangeably: billed weight, scale weight, rehearsal and consolidation. They describe different things and cost different amounts.

Landed cost

Item plus domestic delivery plus service fee plus freight plus duty plus insurance. Six segments, not one.

Billed weight

The greater of scale weight and volumetric weight. What the tariff applies to.

Volumetric weight

Volume in cubic centimetres divided by a divisor, usually 5000.

Rehearsal

Packing the parcel before the international payment so freight can be quoted on packed dimensions.

Consolidation

Combining parcels from several sellers into one shipment. Not the same as packing.

QC

Photographs taken at the warehouse before shipping; a separate step from rehearsal.

Each misuse has a cost attached. Treating billed weight as scale weight understates freight on every bulky parcel; treating consolidation as packing overstates the saving available.

• Billed weight ≠ scale weight.
• Consolidation ≠ packing.
• Rehearsal ≠ QC.
• Landed cost ≠ item price + freight.

Everything in this reading comes from one of three sources, and each figure says which: measured from our sample, community-reported, or read from a published term. Where none applies we write “Not measured” instead of filling the gap.

Measured

Stage timestamps and weights from the 96-entry sample, collected on 2026-10-05.

Community-reported

Patterns that recur in buyer discussions and that we could not reproduce ourselves.

Published

Read from carrier tariffs, agent fee pages and destination thresholds.

Not measured

Anything we could not source. Written as such rather than estimated.

Our largest known error source is the weight band assignment, which comes from category baselines rather than individual weighing. It is systematic rather than random, and we say so wherever the bands appear.

Sample sizes are given for every range on this site. A range without a sample size is not a measurement, it is an opinion.

Most estimates fail because they treat the item price as the cost. Across the 96 entries we hold, the non-item segments on a jackets order run from $7.50 to $74.40 — a range wide enough to change the decision about whether to order at all.

The practical consequence is that optimising the item price is a different activity from optimising the order. A 10% cheaper item saves 10% of one segment; a packing change can move two bands of another.

Transit is not one wait, it is eight. On the US economy lane the median from payment to doorstep is 19 days, and the p90 is 31. The gap between those two numbers is where every complaint lives.

The stages people plan around are not the stages that cost time. Clearance is fast in the median case; the seller-to-warehouse stage and the last mile together account for more than half the total.

Holiday closures

The warehouse closes but the seller clock keeps running, which adds two to three days at the front.

Battery or liquid declaration

A different export route, adding two to four days.

Address failure

A failed delivery attempt adds two days and sometimes a redelivery fee.

Peak season priority loss

The economy lane loses queue position from late October.

The declared figure sets two things at once: the duty the parcel attracts and the ceiling on any compensation if it is lost. They cannot be optimised separately, which is why the choice is a decision rather than an entry field.

Estimate error concentrates near thresholds. That is a property of the rules, not of the model: a step function produces large effects from small changes exactly at the step.

Declare accurately and keep the order record. A plausible figure with evidence survives a claim; a precise figure without evidence does not.

Four variables carry almost all the sensitivity: billed weight, declared value, lane and item value. Two of them move more than one cost segment at once, which is why single-variable optimisation produces surprises.

• Billed weight moves freight, and only freight — but by a full band.
• Declared value moves duty and the compensation ceiling together.
• Item value moves the service fee, and only under a percentage model.
• Lane moves freight and transit; it does not move duty at all.

On the reference lane a heavy (> 2 kg) parcel quotes $14.80–$24.30. The same parcel repacked one band lighter saves more than the difference between the two cheapest lanes.

Six cost calculators exist in this niche, including two of ours. The useful comparison is not what they show but what they omit, because an omission is invisible in the output.

Test any calculator with a large, light parcel. If the freight does not move, the volumetric step is missing — and if it moves without explanation, it is being applied silently.

1. Enter a parcel with high volume and low weight.
2. If freight does not change, the volumetric step is absent.
3. Then change only the declared value.
4. If the total does not change, duty is not modelled.

Density decides whether volume or mass drives the bill, and density is a property of the category rather than the parcel. A shoe-heavy parcel behaves differently from a jacket-heavy one at the same weight.

The boundary is around 40% and 70% by weight of the bulky category. Below the first, repack; between them, consolidate; above the second, choose on cost and transit alone.

• Below 40% bulky by weight: compress before considering a lane change.
• 40–70%: consolidate, and watch the handling tier boundary.
• Above 70%: choose the lane on cost and transit; packing has little effect.

Everything here traces to one of three sources and carries a collection date. Item prices are published figures read on 2026-10-05; freight and duty are modelled from published terms; transit figures are measured from order timestamps.

Where a figure was unavailable we record “Not measured”. That is a deliberate choice: an empty cell is more useful than a plausible one, because it tells you where the uncertainty actually is.

Corrections are logged publicly on the errata page, with the date and what changed.