Fulfilling 500 Units From China: FBA, a 3PL, or Direct Parcel

One product, three American routing stacks, and the per unit arithmetic for the fourth quarter of 2026, priced against a 27.8% platform take rate

SeaPRwire – Five hundred units of a ceramic homeware item, boxed, 1.2 kilograms each, retail at USD 24, made in Guangdong and shipped to the United States in the fourth quarter of 2026.

That is the whole brief. There are three ways to get those units from a factory gate to a customer’s door, and the difference between the cheapest and the most expensive is larger than the entire gross margin on the product. Most sellers pick a stack by reading a rate card, and the rate card is the least useful document in the decision, because it prices a unit and the decision is about a ratio.

The claim this piece defends is narrow: at a USD 24 retail price, the platform stack wins on cost and loses on control, the third party logistics stack loses badly on cost at any volume below about 8,000 units a quarter, and the axis that actually decides the answer is not volume, it is price per kilogram. Sort on that and the choice stops being a matter of taste.

A note on method before the numbers, because it changes how much weight they carry. I did not get quotes from any of these vendors. The fulfilment figures below are modelled from published rate structures and from the market aggregates in the reference library, and every assumption is stated so that a reader with a real quote can replace one number and rerun the table.

The Shipment I Priced

Define the task tightly, because a loose task is how rate card comparisons go wrong.

Table 1: The shipment, and what each stack does with it

Source: modelling assumptions for this piece, anchored to the market aggregates cited below.

Two of the three stacks require you to put inventory in the United States before you have sold anything. That is a cash decision, not a logistics decision, and it is the one sellers skip.

Stack One: FBA, and What 27.8% Buys

The Amazon stack is the simplest to price because the platform publishes one number that absorbs almost everything: 27.8%.

That is the aggregate take rate for Amazon in the second quarter of 2026, as calculated by CICC from company disclosures. It bundles the referral fee, the fulfilment fee and, crucially, the advertising spend that has become mandatory to be seen. It is not a list of fees. It is the share of gross merchandise value that the platform captures.

At the USD 24 price used here in 2026, 27.8% is USD 6.67 a unit, or USD 3,335 across the 500 units.

What that buys is not just pick and pack. It buys Prime badging, which is the single largest conversion lever in American ecommerce. It buys returns handling that does not come back to you as a warehouse problem. And it buys the removal of every operational decision in this article, which is worth something even when the arithmetic is unfavourable.

The cost is control and the cost is the exit. Inventory in an Amazon fulfilment centre is inventory you cannot easily move to a second channel, and the fee structure changes on the platform’s schedule, not yours.

Total landed cost per unit on this stack in the fourth quarter of 2026 is USD 12.77, against a gross margin of USD 11.23, or 46.8%.

Stack Two: A 3PL and Your Own Storefront

The second stack is the one sellers graduate to, and at 500 units a quarter in 2026 it is a trap.

Pair a third party logistics provider with your own storefront and the platform take disappears, replaced by a payment processor at roughly 2.9%, or USD 0.70 a unit. What replaces it is everything the platform was doing for you, priced separately.

Budget in 2026 for pick and pack on a 1.2 kilogram parcel at about USD 6.10, storage at USD 0.08 per unit per month, and a fixed onboarding and minimum commitment of about USD 7,500 covering account setup, EDI integration and a three month warehouse minimum.

That fixed cost is the whole story at this volume. Seven thousand five hundred dollars divided by 500 units in the first quarter is USD 15.00 a unit before a single parcel moves. Total cost per unit lands at USD 28.66, above the retail price.

At 8,000 units a quarter in 2026 the same fixed cost is USD 0.94 a unit and the stack lands at about USD 13.86. At 10,000 units it is USD 13.40, which is where it finally crosses the platform stack in the same year.

Table 2: Per unit cost by stack, at two volumes

Source: modelled from published rate structures and market aggregates in the reference library, with assumptions as stated in the text.

The lesson in that table is not that 3PLs are expensive. It is that a fixed cost divided by a small number is a large number, and 500 units is a small number. Sellers who move to a 3PL in their first quarter because the per parcel rate looked better than the platform fee have usually not divided by volume.

Stack Three: Direct Parcel After De Minimis

The third stack is the oldest and the one that changed most recently.

Ship each order from Shenzhen. No US inventory, no warehouse, no fixed cost. Budget about USD 9.50 a unit for an express parcel including duty and the entry processing that now applies to low value shipments, since the de minimis exemption that used to let small parcels enter duty free was suspended in 2025.

Total cost per unit in 2026 is USD 14.40, leaving a gross margin of USD 9.60, or 40%.

It is the second cheapest stack at 500 units in the fourth quarter of 2026, and the only one with no cash tied up in inventory. For a first order, or for a product whose sell through you cannot forecast, that flexibility is worth more than the USD 1.63 per unit gap to the platform stack.

It also carries the worst delivery experience. Two to three weeks against two days is a conversion tax that does not appear anywhere in the table, and for a USD 24 ceramic item the breakage rate in postal networks is materially higher than in a domestic parcel flow.

Where the Three Diverge

Three differences matter, and none of them are in the rate cards.

The first is who owns the demand. On the platform stack you rent access to buyers who are already there. On the other two you buy it, and the advertising cost that the platform bundles into its 27.8% becomes a separate line that most sellers underestimate by half in their first year.

The second is what happens to a return. On the platform stack a returned ceramic item is the platform’s problem for a fee. On the other two it is your problem, and the cost is not the refund, it is the unit that comes back unsellable.

The third is the shape of the fee. A percentage take scales with price. A fulfilment fee does not. That single difference decides everything at the top of the price range, and it is what the next section is about.

One claim here is falsifiable and worth stating as such. If a seller moving 500 units a quarter at a USD 24 retail price can obtain a third party all in fulfilment quote below USD 6.70 a unit, including onboarding and storage, then this piece is wrong about the trap in Stack Two and the crossover moves. Ask for that number. If it comes back under 6.70, ignore the rest of this article and sign.

The Axis You Should Sort On: Price Per Kilogram

Here is the calculation that makes the decision, and it is not the one most sellers run.

Hold weight constant at 1.2 kilograms and move the retail price. At USD 24 in 2026, the platform takes USD 6.67 while the 3PL charges USD 6.10 to move the box. They are almost identical, which is why the two stacks look interchangeable at this price.

At USD 60 in the same year, the platform takes USD 16.68 and the 3PL still charges USD 6.10. Now the 3PL is USD 10.58 cheaper per unit before you count the payment fee.

At USD 120 in 2026, the platform takes USD 33.36 and the 3PL charges USD 6.10. The gap in 2026 is USD 27.26 a unit, and no amount of platform convenience closes it.

Table 3: What each stack costs at three retail prices, holding weight at 1.2 kg

Source: arithmetic from the take rates and fulfilment assumptions stated above.

The crossover in 2026 is at about USD 26 retail for this weight. Below it, pay the percentage. Above it, pay the parcel.

Now run the same logic on weight and the axis rotates again. A 200 gram item at USD 24 in 2026 pays the platform 6.67 and a 3PL about 4.20, so the platform is worse, but a direct parcel costs 3.10 and wins outright. A 4 kilogram item at USD 24 in 2026 pays the platform 6.67 and a 3PL 9.80, so the platform wins on cost and wins big.

Sort your catalogue on price divided by weight and the stacks separate cleanly. High value, low weight goes direct or to a 3PL. Low value, high weight goes to the platform. The middle band, which is where most consumer goods sit, is decided by volume, and volume is decided by whether you have demand you did not rent.

Which One, by Situation

Concrete, so it can be acted on.

• Under 2,000 units a quarter, unknown sell through: direct parcel. In 2026 that means paying the 1.63 dollar premium and keeping the cash.
• Under 2,000 units a quarter, proven sell through, retail under USD 30: platform. The bundled advertising and returns handling are worth more than the fee difference.
• Over 8,000 units a quarter, retail above USD 40: 3PL with your own storefront. This is the only configuration where the fixed cost amortises and the percentage fee stops taxing your price.
• Fragile, heavy, or high return rate: platform, and accept the take rate as an insurance premium.
• Subscription or replenishment product: 3PL. Repeat orders make the fixed onboarding cost amortise fast and the platform’s percentage take becomes a permanent tax on loyalty.

The Cost Nobody Puts in the Model: Breakage

Every stack in this article is priced as if the unit arrives intact. Ceramic does not always arrive intact, and the channel determines how much of the loss you eat.

Three different exposures, and they are not comparable.

On the platform stack in 2026, a damaged unit is a return, and the return is processed inside a system you did not build. You pay for the return in the fee structure, you lose the sale, and in many cases you lose the unit as well because a broken ceramic item is not resellable. What you do not lose is the customer relationship, because the platform owns it.

On the 3PL stack, a damaged unit is a customer service ticket, a refund you fund yourself, and a unit that comes back to a warehouse that will charge you to receive it. Budget for the inbound return leg, the inspection, and the disposal. None of those appear on a rate card.

On the direct parcel stack, the exposure is highest. Two to three weeks in postal networks, multiple handling points, and a box that has crossed an ocean in a container before it even starts. For a 1.2 kilogram ceramic item, a damage rate in the high single digits is realistic on this route and low single digits domestically.

Put a number on it and the ranking changes. At a 7% damage rate on a USD 24 item, you lose USD 1.68 per unit sold in unrecovered cost, on top of the refund. That is more than the entire 1.63 dollar gap between the platform stack and the direct parcel stack. At a 2% rate you lose USD 0.48, which is noise.

This is why the cheapest stack on the cost table is often not the cheapest stack in the quarter. If you sell anything fragile, price the damage rate before you price the parcel.

Two Numbers in This Industry Nobody Can Agree On

One warning before you build a model on market data, because the market data does not agree with itself.

The United States third party logistics market is USD 225.69 billion in 2026 according to Mordor Intelligence. It is USD 367.95 billion in 2026 according to Market Data Forecast. The 2026 gap is 63%, or about USD 142 billion, and both are published by firms that sell research.

The difference is definitional. One counts contract logistics revenue for outsourced distribution, the other sweeps in a wider set of transportation and value added services. Neither is wrong, and a seller who quotes the larger number in a business plan is quoting a market that includes businesses they will never compete with or buy from.

The same caution applies to the structure figures. For 2026, domestic transportation is 47.55% of the 3PL market, manufacturing 36.85%, and asset light operators 43.35%. Value added warehousing is forecast to grow at 7.55% a year through 2031, while the broader freight and logistics market grows at 3.8%. Those are useful for understanding where capacity is being added. They are useless for pricing your 500 units.

What I could not do is check a single one of these rate cards against a live quote. The published structures are not comparable across providers, the de minimis suspension is recorded here without an effective date because I could not confirm one, and the duty treatment of a specific product classification can move the direct parcel stack by several dollars. That is the weakest part of this piece and a reader with an invoice should trust the invoice.

For context on the flow you are joining: trucking moved more than 72% of domestic freight tonnage in 2024, on USD 906 billion of industry revenue, and logistics costs run 8% to 10% of American GDP. The August 2026 goods and services deficit was USD 105.6 billion, up 13.7% on the month, with the year to date deficit down 19.9%.

Start With the Deposit, Not the Rate Card

The step people skip is not the comparison. It is the cash calendar.

Before you choose a stack, write down three dates and three amounts: when the factory wants the balance, when the inventory lands in the United States, and when the first marketplace payout clears. On the platform stack in 2026, inventory is committed roughly 45 days before the first sale and the payout arrives about two weeks after it. On the direct parcel stack there is no inventory commitment at all, and the parcel is paid for after the order.

That timing difference is worth more in 2026 than the USD 1.63 per unit separating the two cheapest stacks. A seller with USD 20,000 of working capital in 2026 cannot run the platform stack at 500 units without financing the gap, and the financing cost will exceed the fee difference.

So do this today, in this order. Price your product on price divided by weight and write the number down. Take the USD 26 crossover and see which side you are on. Then take your quarterly unit forecast, divide 7,500 by it, and see whether the 3PL stack is even arithmetically available to you. If the answer is under 8,000 units, stop shopping for warehouses and start with the parcel flow, because the stack that requires the least cash is the one that survives a bad first quarter.

Data basis

Data as of October 2026. Platform take rates are CICC’s calculation from company disclosures for the second quarter of 2026, an aggregate of referral, fulfilment and advertising revenue as a share of gross merchandise value, and are used as a single blended figure rather than a fee schedule. Fulfilment, storage, onboarding and per parcel figures are modelled for this piece from published rate structures and are labelled as assumptions in Table 1; they are not quotes and should be replaced with a seller’s own. Inbound cost of USD 1.90 per unit assumes sea LCL from South China including duty and port charges, and is the most sensitive input in the model. The 3PL market size range, segment shares and growth rates are as published for 2026, with the definitional conflict between the two market size estimates stated in the text rather than resolved. Trucking revenue and tonnage share are 2024 figures. The crossover price of USD 26 solves for the retail price at which a 27.8% take equals a 6.10 fulfilment fee plus a 2.9% payment fee.

Limitations

No vendor quoted for this piece, so every fulfilment figure is a model rather than a price, and a real quote could move the crossover by several dollars. The de minimis suspension is treated as a 2025 policy change without a specific effective date, and the duty treatment of a given product classification can change the direct parcel stack materially. The two market size estimates are presented, not adjudicated, because neither publisher makes its scope definition fully available.

Sources

• CICC via TTM Financial , platform take rate comparison for the second quarter of 2026, October 2026 (site home)
• Mordor Intelligence , United States 3PL market size and segment shares, 2026
• Market Data Forecast , United States third party logistics market, 2026
• American Trucking Associations , American Trucking Trends, 2025 edition
• US Census Bureau , FT900 goods and services trade, August 2026

This article was prepared by the expert team at Global Infinity Labs, drawing on insights and data provided through its MCP data services.

About Global Infinity Labs

Global Infinity Labs is an AI innovation company specializing in MCP data services.

Email: support@globalinfinitylabs.com