🛒 Channel & Fulfillment Layer Deep Dive · Updated Sept 2026

Amazon FBA vs Direct Sourcing: Where Your Margin Actually Goes

The channel guide settled where to buy (factory vs market vs agent); this page settles where the goods sit afterward — Amazon's warehouse or yours. FBA trades 30-40% of gross for demand and logistics muscle; direct keeps the margin but rents you the traffic problem instead. Here is the unit math, the inventory drag, and how real buyers split one order across both.

Reading time: 9 min · Updated Sept 2026

Sellers frame this as Amazon versus Shopify. It isn't. Both channels draw from the same factory PO — the decision is about who holds the inventory after customs clears, which fee stack applies per sale, and whose replenishment clock governs your cash. Run the decision on contribution margin per unit per week of inventory risk, and most "FBA or not" arguments dissolve.

Who You Pay on Each Path After the Container Lands

Same factory invoice on both paths. The divergence starts at the port:

Cost lineFBA pathDirect path (own store/3PL)Who really pays
Referral fee~15% of gross (8-15% by category)Payment processing ~2.9% + 30¢FBA path hands Amazon the single biggest line
Fulfillment fee$4-9 per small parcel unit by size tier3PL pick-pack $2-4 + actual postageComparable at ground rates; FBA wins on multi-unit orders
StorageMonthly per cubic foot; long-term surcharge on 365+ days3PL pallet storage ~$15-25/pallet/monthFBA punishes slow movers harder
ReturnsAmazon auto-accepts most categoriesYou write the return policyDirect = control; FBA = refund velocity you don't govern
Demand sidePrime traffic included in the feeYou buy ads; CAC is your real lineThe referral cut is buying traffic — judge it as media spend

The discipline: stop arguing "FBA fees are high" in the abstract — the referral fee is a media budget with logistics attached. Judge it against what you'd otherwise spend on Meta/Google ads to buy the same conversion, and FBA's economics look different for review-rich SKUs than for unproven ones.

The Unit Math: One SKU Through Both Stacks

Take the $4,200 teardown order's mid-tier SKU — a # stainless steel water bottle, factory-ex at $3.10, ocean freight amortized to $0.55, duties and entry costs $0.47 → landed unit $4.12. Now run both channels at a $24.99 shelf price:

LineFBADirect (Shopify + 3PL)
Shelf price$24.99$24.99
Referral / processing−$3.75 (15%)−$1.02 (2.9% + 30¢)
Fulfillment−$5.40 (size-tier standard)−$2.60 pick-pack + $4.80 ground parcel
Storage/burden per unit−$0.35 (with LTSF risk flagged)−$0.40 pallet share
Return allowance−$0.75 (category norm 4%)−$0.45 (stricter policy, fewer impulse buys)
Contribution/unit−$0.31 + demand you didn't pay for$14.90 − ads at market CAC

Same SKU, and FBA shows near-zero contribution until you price the demand it brings. If organic+Prime converts at 12% against your store's 2.5% cold traffic, the media-adjusted comparison is the honest one — and it flips SKU by SKU, not once for the whole store. Two inputs do the heavy lifting here: the landed unit under both stacks comes from the duty & HS code math, and the shelf price you defend comes from what the market, not your cost sheet, will bear. This is why the decision is a spreadsheet, not an identity.

Inventory Drag: The Replenishment Clock Nobody Budgets

How Real Buyers Actually Split: The Hybrid Playbook

Ask a working seller and the binary dissolves. The common 2026 pattern for one 5-SKU PO from one factory:

  1. Proven SKUs to FBA — the two styles with 100+ reviews ride Prime demand; margin per unit is lower but the volume is not yours to manufacture.
  2. Variant depth stays direct — colorways, sizes, and new releases ship to your 3PL and sell on Shopify at full margin, absorbing ads as the real cost line.
  3. One carton split at the forwarder — the consolidation math: freight forwarder receives the full PO, deconsolidates, and trucks two destinations. The container math guide covers the consolidation logic; add roughly $0.10-0.25/unit for the split-shipment handling.
  4. Barcode discipline per destination — FNSKU-labeled cartons for FBA, UPC for retail-ready direct stock, never mixed in one master carton without labeling on arrival. This is the documents layer extended into warehouse tickets: same discipline, different form.
  5. Rebalance quarterly by contribution per cubic centimeter — the honest KPI mixing margin, velocity and inventory risk. SKUs that lose the KPI migrate channels; sentiment doesn't get a vote.

The Four FBA-Specific Failure Modes (and the Direct-Channel Mirrors)

Every channel has its signature way of eating a first order:

Failure modeFBA versionDirect version
Compliance rejectsInbound receiving rejects (box spec, suffocation warning, FNSKU mismatch) → receiving limbo + disposal feesAudit failure at destination (CE/UKCA/FCC for your own market) → whole-line rework; the certifications guide covers both regulatory sides
Fee-tier cliffs2cm dimension change jumps fulfillment tier; repricing silently erases marginDimensional weight pricing at the 3PL does the same to your postage line
Stockout/perishabilityRank decay after stockout; re-entry costs double ad spend for weeksEmail list goes quiet; catalog rank slip you do control but must notice
Account/relationship riskListing suspension on review-cluster flags — your demand engine freezesPayment-processor reserves hold your cashflow instead; the payment guide logic applies to processor risk too

The IP and registry layer from the NNN playbook cuts across both channels — a hijacked listing and a copied listing hurt differently, but both trace to the same weakest link.

Frequently Asked Questions

What percentage does Amazon FBA take per sale?

Plan on a referral fee (about 15% in most consumer categories, 8-15% range) plus the FBA fulfillment fee by size tier (roughly $4-9 for typical small-parcel items), plus monthly storage and optional prep/labeling. Together these commonly consume 30-40% of gross revenue on a $25 product — before product cost and inbound freight. The referral cut is best judged as media spend: it buys Prime-demand access you would otherwise purchase with ads.

Is FBA still worth it for new sellers in 2026?

For stable, review-defensible products, yes — Prime conversion and Amazon's built-in traffic frequently repay the fee stack. For trend-driven, fast-rotating or oversized items, the fee stack plus inbound receiving lag often eats the advantage, and a direct Shopify channel with your own 3PL wins on contribution margin. Judge SKU by SKU, not once for the store.

How much inventory does FBA require upfront?

FBA adds an inbound receiving window (commonly 1-2 weeks after the container lands, longer in Q4) on top of ocean transit, plus restock caps on new listings. Practical effect: buyers carry 6-10 weeks of cover per cycle versus 3-5 weeks when shipping direct from a 3PL. At a first $4,200 order scale, that difference is roughly one extra purchase order of working capital per quarter.

Can I mix FBA and my own Shopify store with the same supplier?

Yes — and most experienced buyers do. Common split: FBA carries the two or three proven SKUs for Prime-weighted demand; the Shopify store carries colorway depth and new releases. Suppliers ship once to your freight forwarder, and a 3PL splits parcels to FBA plus direct orders. Keep barcodes and carton labels correct per destination (FNSKU for FBA, UPC for retail-ready), and never mix label types in one master carton.

What FBA-specific mistakes do China buyers make?

The recurring four: cartons that fail inbound checks (wrong box spec, missing polybag suffocation warnings), FNSKU/UPC mismatches that strand stock in receiving limbo, missing CE/CPC paperwork the listing later needs, and ignoring size-tier cliffs where a 2cm dimension change jumps the fulfillment fee band. The packing-spec guide carries the carton-level checklist that prevents the first two.

Does sourcing direct beat retail-channel buying once you subtract FBA fees?

Different axes: the sourcing-channel spread (18-35% across factory, market and agent) is about purchase price; the FBA-vs-direct spread is about where inventory sits and which fee stack applies per sale. They stack. Direct sourcing lowers the input cost either way — then this page's unit math decides whether the margin goes to Amazon as demand-spend or stays with you while you buy ads directly.

Deciding your channel split for the next PO?

Send us the SKU list and destination. We'll reply within 24-48h with a two-stack contribution comparison and the carton-labeling plan your forwarder needs — one PO, both channels, no receiving surprises.

Get Your Channel Split Review →

Related Guides

Written & verified by the China Market Guide team

Four on-the-ground researchers in Shenzhen, Guangzhou and Yiwu. Every price anchor, MOQ norm and QC checkpoint on this page is cross-checked against live market visits and buyer-side inspection reports — see our editorial process and independence policy.