The store did $41,000 a month. Two thousand parcels, most of them between $12 and $30, shipped one at a time from a supplier in Yiwu through a consolidator that cleared them into the United States under Section 321 — the $800 de minimis exemption that had, for two decades, been the quiet subsidy underneath the entire direct-from-China e-commerce economy. No duty. No formal entry. No broker. No bond. The owner's spreadsheet had one line for "shipping" and it was accurate. In 2026 the same store, same supplier, same shipping line, faces a bill it cannot find on any invoice: duty on every parcel by HS code, an entry process attached to every parcel, and a November fee that had not been announced when the year's pricing was set. The unit cost of an order rose by more than the entire gross margin on the cheapest third of the catalogue. Nothing about the business broke. The arithmetic underneath it did.
This is the shape of every dropshipping conversation in 2026, and it is distorted in both directions. One camp insists dropshipping is dead, citing the end of de minimis as proof. The other camp is still selling courses built on a 2023 cost model where a $14.99 item shipped from China free and delivered in eleven days. Neither is useful. What actually happened is that the removal of low-value duty relief — for China and Hong Kong on 2 May 2025, for every country of origin on 29 August 2025, made indefinite across all modes including international mail by CBP's interim final rules of 24 June 2026, and legislated into permanent elimination from 1 July 2027 — deleted one specific business model out of six, and forced the other five to re-price.
The model that died was the cheapest one: a low-ticket, single-parcel, direct-from-China order with a thin enough margin that duty and a per-parcel entry cost more than the product. The model that replaced it is a domestic-fulfillment model: pooled stock sitting in a Los Angeles, New Jersey or Dallas warehouse run by a Chinese fulfilment network, or your own inventory in a US 3PL, shipping to the customer in three to five days with no customs event per order. The interesting part — and the part that genuinely upends the standard advice — is that once you price the 2026 rules honestly, the domestic model does not just survive on convenience, it wins on margin at the same retail price.
The Six Things People Mean by "Dropshipping" in 2026
The word covers at least six different supply arrangements with different legal treatment, different cost structures and different survival odds. Every argument about whether dropshipping works is really an argument between two of these models, usually without either side saying which one it means.
| Model | Who holds the stock | Who ships and clears the parcel | 2026 status |
|---|---|---|---|
| 1. Retail arbitrage — buy from Walmart, Target, Home Depot or another Amazon seller and route the order | Another retailer | That retailer, in its own packaging, with its own paperwork | Prohibited. Amazon's dropshipping policy bars buying from another retailer and having it ship to your customer; a third-party packing slip or receipt in the box triggers suspension. Most suspensions in this category come from exactly this, plus late shipment and third-party invoices |
| 2. China-direct supplier fulfilment — AliExpress, CJ, an agent or your factory ships each order from China | The supplier, in China | Supplier or its consolidator, one parcel per order, one customs event per parcel | Strained but alive above roughly $35. Every parcel now attracts duty and its own entry overhead, so the low-ticket end is arithmetically dead. Delivery remains 8–20 days door to door on the standard lines |
| 3. US-warehouse dropship from a China-based fulfilment network — CJ-style platforms, Chinese forwarders with US nodes | The fulfilment network, in the US | The network, domestically, 3–5 business days, no customs event per order | The 2026 mainstream. Duty was paid once, on a consolidated inbound shipment at wholesale value. No MOQ in the classic sense; you pay per order, often at a 10–20% premium over the China-warehouse price |
| 4. Domestic wholesale dropship — a US distributor, or a Chinese factory with a US distribution centre, ships your orders | The distributor | The distributor, domestically, usually with a branded or neutral packing slip | Alive, thinnest margin, least risk. Typically requires a reseller agreement and sometimes a small opening order; margin is 20–35% rather than 60%+ |
| 5. Print-on-demand — the product is manufactured after the order, domestically | Nobody (made to order) | A domestic POD facility (printed apparel, mugs, wall art) | Alive and structurally protected from the de minimis change, because no parcel crosses a border per order. Blank goods are imported in bulk at wholesale value |
| 6. Marketplace semi-managed — Temu, AliExpress Local+ / Choice, Shein programmes | You, in an overseas warehouse the platform approves | The platform or its logistics arm, with a delivery-time promise and a listing badge | Growing fast. Temu's semi-managed model pushes sellers into overseas stock; AliExpress Local+ works through partner warehouses with a seven-day delivery promise; Shein has opened warehouse capacity to sellers. It is dropshipping with the inventory risk handed back to you |
Two of these six — the arbitrage model and the low-ticket direct model — are the ones that dominated YouTube in 2021 and dominate headlines in 2026. The other four are ordinary supply chain businesses that happen to hand fulfilment to somebody else. If you are deciding whether to start, the question is not "is dropshipping viable" but "which of the four viable arrangements fits my product, my price point and my capital".
The Rule That Built the Model — And the Timeline That Killed It
Section 321 of the Tariff Act allowed goods valued at $800 or less per consignment, per importer, per day, to enter the United States free of duty and without a formal customs entry. At its peak CBP processed over a billion such shipments a year, the overwhelming majority originating in China. A decade of cross-border e-commerce economics was built on that single line of law, and it was removed in stages that every importer needs on one page.
| Date | What changed | Why it matters to a dropship programme |
|---|---|---|
| 2 April 2025 / 2 May 2025 | Executive Order 14256 eliminated de minimis treatment for goods of Chinese and Hong Kong origin, with a flat postal fee option ($100 per item, later $200) | The origin of most dropship stock lost duty-free entry first. The immediate workaround — routing through a third country — was closed four months later |
| 4 July 2025 | The One Big Beautiful Bill Act included permanent elimination of the de minimis exemption for all countries from 1 July 2027 | This is the part that should end any plan built on the exemption returning. Congress wrote the deletion into statute, not just into an executive action |
| 29 August 2025 | Suspension extended to low-value shipments from all countries of origin | No third-country rerouting, no Hong Kong or Singapore transit workaround. Every origin now needs an entry |
| February 2026 | The suspension was continued by executive action; the Supreme Court's ruling struck down a separate set of IEEPA-based reciprocal tariffs but did not touch Section 321, which rests on different authority | Read this twice, because the two things are constantly conflated. The tariff ruling did not restore de minimis, and there is no de minimis case pending |
| 28 February 2026 | The simplified flat per-item postal duty option ended. Postal items are now charged on an ad valorem basis at the country-of-origin rate, with carriers billing full duty | Postal-network shipping lost its simple fixed-fee path. Carriers now front duty and bill it back, which shifts working capital onto you and produces surprise door charges for the end customer |
| 24 June 2026 | CBP interim final rules made the suspension of the $800 exemption indefinite across all modes of importation, including international mail | The definitive state of play. All entries of low-value merchandise use standard CBP entry procedures and are subject to applicable duties, taxes and fees |
| 1 July 2026 | EU removed the €150 customs duty exemption under Council Regulation (EU) 2026/382, replacing it with a €3 flat duty per HS6 product category on in-scope B2C consignments, transitional until the Customs Data Hub takes over around 1 July 2028 | The EU closed its own loophole the same year. A parcel with three different product categories pays €9, not €3. 91% of sub-€150 e-commerce shipments into the EU had been coming from China |
| 8 July 2026 | CPSC eFiling became mandatory: certificate of compliance data is transmitted with the customs entry | If you are the importer of record on a bulk inbound shipment, you now own a data obligation as well as a duty obligation. A dropshipped parcel whose certificate names somebody else's brand has no answer to this |
| 24 July 2026 | The Section 122 surcharge expired and the tariff stack was rebuilt, with a reported 12.5% baseline on most origins and 10% on some, layered on the underlying MFN and Section 301 rates | Any landed-cost model built in June is wrong. Classify each SKU and confirm the live duty for its HS code before you price |
| 1 November 2026 | A separate per-parcel handling fee on low-value postal shipments is scheduled to take effect in the United States no later than this date; the EU's proposed Union handling fee (around €2 per consignment) is targeted at the same window | This is the one that lands in the middle of Q4 peak. If any part of your supply chain still moves through postal channels, price it before your holiday buy, not after |
The pattern across that table is consistent: three separate mechanisms — duty on entry, per-parcel entry processing, and postal handling fees — and only the first one is proportional to product value. The other two are fixed costs per parcel, which is why the damage is concentrated at the cheap end of every catalogue.
The Per-Parcel Economics That Decide Whether Your SKU Lives
Start with the fixed overhead. Every low-value consignment now needs its own entry process. When a carrier or consolidator fronts that entry, the practical cost is commonly in the range of $5 to $15 per parcel at the low end (postal and carrier-facilitated clearance) and $25 to $75 per shipment when it is handled as a standalone informal or formal entry with brokerage attached. Set that against the factory cost of the product and the shape of the problem becomes obvious.
| Product factory cost (FOB) | Duty at an illustrative 30% combined stack | Entry and clearance at $5/parcel | Entry and clearance at $15/parcel | Fixed + duty as a share of FOB |
|---|---|---|---|---|
| $3.00 | $0.90 | $5.90 total | $15.90 total | 197% – 530% |
| $8.00 | $2.40 | $7.40 total | $17.40 total | 92% – 217% |
| $15.00 | $4.50 | $9.50 total | $19.50 total | 63% – 130% |
| $30.00 | $9.00 | $14.00 total | $24.00 total | 47% – 80% |
| $60.00 | $18.00 | $23.00 total | $33.00 total | 38% – 55% |
Rates vary by HS code, origin and the state of the tariff stack, and the entry cost varies by carrier, consolidator and mode — so treat the columns as a shape, not a quote, and have your own broker classify your own SKUs. The shape is what matters. On a $3 item the compliance overhead is between two and five times the cost of the goods. On a $60 item it is under half. Nothing about that arithmetic involves the product, the marketing or the supplier's competence: it is a fixed cost per parcel applied to a variable price, and it is the reason the low-ticket end of direct-from-China died while the mid-ticket end did not.
The break-even price moved 58% in one cycle
Now run it through a whole order. Take a 400g consumer hardgood with an $8 factory cost, sold to a US customer through your own store, billed at a supplier-quoted price that in 2024 included shipping to the end customer. Assume an illustrative 30% combined duty stack for the 2026 case, $6 of blended per-parcel entry and clearance cost, $5.50 of shipping to the customer, a 15% marketplace or channel fee, 2.9% + $0.30 payment processing, a 6% refund allowance, 12% ad spend at a $30-ish average order value, and $0.90 of per-order support and app cost.
| Line | 2024 China-direct (pre-suspension) | 2026 China-direct | 2026 US-warehouse dropship |
|---|---|---|---|
| Product cost (FOB) | $8.00 | $8.00 | $22.00 landed at the US node, sold to you per order |
| Duty (illustrative 30%) | $0.00 | $2.40 | Inside the $22.00 — paid once on a consolidated inbound shipment at wholesale value |
| Per-parcel entry and clearance | $0.00 | $6.00 | $0.00 per order |
| Shipping to the end customer | $5.50 | $5.50 | Inside the $22.00 (3–5 day domestic) |
| Delivered cost of goods | $13.50 | $21.90 | $22.00 |
| Variable take rate (channel + payment + refunds + ads) | 35.9% of price | 35.9% of price | 33.9% of price — refund rate and ad spend both improve on a 3–5 day delivery promise |
| Break-even selling price | $22.46 | $35.57 | $34.64 |
| Selling price for a 15% net margin | $29.33 | $46.44 | $44.81 |
Three conclusions fall out of that table, and they are the strategic core of this guide.
First, the China-direct break-even price rose from about $22 to about $36, a 58% increase, with no change to the product, the supplier or the marketing. A $19.99 order that used to lose roughly $1.60 per sale now loses roughly $9.99. There is no operational excellence that recovers nine dollars on a twenty-dollar sale; the only answers are to raise the price, change the product, or change the supply model.
Second, a domestic node breaks even at almost exactly the same price ($34.64) and nets more per order above it — $16.75 against $15.65 at $59.99, $29.97 against $28.47 at $79.99. The US-warehouse model pays a 2.75x markup on the product to avoid duty per order, per-parcel entry, and an 8–20 day delivery window, and the trade is worth taking once you account for the lower refund rate and the better conversion that a 3–5 day delivery produces. The comparison is not China versus America on unit cost; it is China's unit cost plus its fixed parcel overhead plus its conversion penalty, against a domestic wholesale price that already has all three solved.
Third, the fixed per-parcel cost does not care how good your store is. It is charged whether the order is a first purchase or a repeat, whether the customer is delighted or furious, whether the ad converted at 1% or 6%. Anything you cannot amortise over a bigger basket — a bundle, a multi-unit order, a subscription — is exposed to it on every single order.
Where the model lives and dies by price band
| Selling price band | China-direct viability (2026) | What actually works here |
|---|---|---|
| Under $15 | Dead — the fixed overhead exceeds the product cost, and the contribution is negative before a single click is bought | Print-on-demand, domestic wholesale dropship, or a bundled multi-unit SKU that pushes the order above the threshold |
| $15 – $35 | Negative to marginal — below the ~$36 break-even, so every order contributes cash out, not cash in | Domestic supplier fulfilment, POD, or a re-engineered product with a higher price and a real reason for it |
| $35 – $60 | Viable but inferior — positive contribution, worse than the same SKU from a US node, and it exposes you to duty reassessment and door-charge refusals | US-warehouse dropship, with China-direct as a long-tail overflow for low-demand variants |
| $60 – $150 | Sweet spot for high-margin, light, non-restricted goods — the fixed overhead shrinks to 10–20% of price and the DDP-inclusive supplier rates absorb the duty cleanly | China-direct for long-tail and exclusive SKUs, US 3PL for anything with a demand curve |
| Above $150 – $200 | Dropshipping's service layer breaks — customers at this price expect stock confirmation, a real returns process, warranty handling and a brand behind the product; a supplier-fulfilled parcel with no US presence produces chargebacks and negative reviews | Hold stock. Import in bulk, land it in a 3PL or FBA, and accept the inventory risk as the cost of the price point |
Most failed dropship programmes in 2026 are not failing for lack of traffic. They are selling a $19.99 product in a cost structure that needs $36, and paying for ads to do it faster.
The Entry-Overhead Trap: Why 300 Parcels Cost More Than One Container
Duty is unavoidable and proportional, so it is easy to model. The entry process is not, and it is where the money actually goes. A parcel shipped individually needs its own clearance event. Three hundred parcels need three hundred clearance events. One consolidated LCL shipment needs one.
| Cost component | 300 individual parcels | One consolidated LCL formal entry |
|---|---|---|
| Duty | Same rates, assessed and collected 300 times; commonly advanced by the carrier and billed back to you | Same rates, calculated once on the commercial invoice value |
| Entry processing | 300 separate informal or postal entries, each with its own processing charge | One merchandise processing fee: 0.3464% of value, with a floor around $34 and a cap around $652 in FY2026 |
| Brokerage and handling | Commonly $25–75 per parcel when a carrier fronts the entry; $7,500–22,500 across 300 parcels | One broker filing, typically a few hundred dollars all-in including bond amortisation |
| Per-unit overhead | Roughly $25–75 per unit | Roughly $1 per unit, and often less |
| Working capital | Duty and fees fronted by the carrier, billed back on a lag, with monthly duty reporting and bond requirements of their own | Duty paid once at entry, known before the goods move, financed by the same cash you already committed to the PO |
| Failure mode | A door charge the customer did not expect. Reported parcel-refusal rates on duty-unpaid international deliveries run at a meaningful double-digit percentage, and every refusal is a refund plus a return at your cost | Duty is invisible to the customer. The only visible number is the delivery date |
The correct reading of that table is not "start importing containers". It is that the fixed cost of a customs event is only worth paying once, and the way to pay it once is to buy in bulk and fulfil from a domestic position. That is precisely the behaviour the platforms are engineering, and it is why Temu's semi-managed model, AliExpress's Local+ partner warehouses and Shein's seller warehouse programmes all moved in the same direction at the same time: the €150 exemption ended in the EU on 1 July 2026 and the American equivalent had already gone, and the direct-from-China parcel that all three were built on stopped being the cheapest way to reach a customer.
The EU Side: €3 Per Category, Not Per Parcel
Europe's change is smaller in headline number and nastier in structure. From 1 July 2026, under Council Regulation (EU) 2026/382, the €150 customs duty exemption is gone for consignments entering the EU from outside the bloc. In its place is a temporary flat duty of €3 — charged per HS6 product category in the consignment, not per parcel, and not per item count.
The mechanics matter for basket design. A parcel containing one phone case attracts €3. A parcel containing a phone case, a charger and a pair of earphones attracts €9, because those are three tariff sub-headings, even though the whole box may be worth less than €150. The Council's own illustration uses two wool blouses and a silk blouse: €6, because silk and wool fall under different sub-headings even though all three are blouses — while two identical wool blouses in different colours share one sub-heading and attract €3. Low-value consignments still clear on the simplified H7 declaration; C2C consignments between genuine private individuals remain exempt under existing thresholds; shipments over €150 were never inside the exemption and are unchanged.
Stacked on top of the duty is a second layer of national and Union fees that arrived on their own schedules: a French small-parcels tax of €2 per unique HS6 category cleared in France from 1 March 2026, an Italian parcel contribution of around €2, a Romanian logistics tax of roughly 25 lei per non-EU parcel under €150 from January 2026, and a proposed EU handling fee of about €2 per consignment targeted at late 2026. From 1 November 2026, low-value declarations are also expected to require up to three product identifiers per line — a data obligation that presupposes you know the manufacturer and the model of what you are shipping.
For a dropshipper the practical consequence is a floor. A single-item parcel into the EU now carries €3 of duty plus, in several member states, €2 or more of national fees, before any carrier handling. On a €19 item that is a 25%+ cost increase collected at the border; on a €9 item it is fatal. The transitional regime is scheduled to run until the EU Customs Data Hub can calculate duty on e-commerce transactions, around 1 July 2028, at which point parcels revert to normal classification-based rates. The number will change. The direction will not.
Marketplace Policy: The Four Gates That Suspend Accounts
Customs law decides what a parcel costs. Marketplace policy decides whether you are allowed to sell it at all, and it is enforced by scanning rather than by customs officers. Amazon permits supplier fulfilment — it is a legitimate model — but only behind four simultaneous requirements, and violating any one of them is an account-level event rather than a listing-level one.
- You are the seller of record on all documentation. Your business name appears on packing slips, invoices and shipping labels — not the supplier's. Amazon's own framing is that the seller on the record must be the seller of record; in practice this means a written agreement with the supplier that they will ship under your identity, and packages that carry your documentation.
- All third-party branding is removed from the shipment. No supplier logo, no supplier invoice, no QR code, no thank-you card driving the customer to a factory's store, no packaging that identifies anyone other than you. Automated systems scan for it, and a customer complaint about an unfamiliar brand in the box is treated as a policy violation.
- You handle all customer returns and service. Amazon expects you to accept and process returns under your own business name, within its return windows, without routing the customer to a foreign address or to a factory.
- You never use another retailer as your supplier. Buying from Walmart, Target, Home Depot, eBay or another marketplace and having it ship directly to your customer is the prohibited form of dropshipping, and it is the most common cause of suspension in this category.
The operational layer sits underneath those rules and polices the model whether or not anyone reviews your packaging. Late shipment rate, order defect rate, valid tracking rate and cancellation rate are measured continuously, and a China-direct programme with an 8–20 day delivery window fights those metrics permanently. Amazon customers expect a 3–5 day window; a domestic node hits it, a Chinese supplier shipping by consolidator does not. Sellers report that slow delivery drives the negative-experience rate that eventually costs the account, which is a compliance argument for the domestic model that has nothing to do with duty.
The other marketplaces have converged on the same principle from different angles. Walmart's seller standards place the same seller-of-record and customer-service obligations on the seller, with an added emphasis on item-level traceability and product safety documentation. TikTok Shop permits supplier fulfilment but holds the seller responsible for shipping speed promises, requires accurate handling times, and will request product documentation where a category is flagged — and its shop-performance metrics punish a foreign fulfilment node faster than a domestic one. eBay is structurally more permissive and correspondingly slower: its money-back guarantee and seller-performance standards still make a 20-day delivery a defect risk, and buyer-protection disputes are resolved against the seller on timing. Etsy sits at the far end, with the lightest formal requirements and the same underlying liability law; a made-to-order model (POD) fits its expectations far better than a parcel from Yiwu does.
| Channel | Supplier fulfilment permitted? | The rule that actually bites | Best-fit fulfilment model |
|---|---|---|---|
| Amazon | Yes, under the dropshipping policy | Seller of record, brand-free packaging, returns handled by you, no retail arbitrage; plus late-shipment and order-defect metrics | US-warehouse dropship or your own stock in a 3PL; an 8–20 day China-direct window is a metric problem even when it is compliant |
| TikTok Shop | Yes, with handling-time accuracy | Delivery-speed promises and shop-performance metrics; documentation requests in flagged categories | Domestic node; content-led demand and slow fulfilment produce refund requests and reviews in public |
| Walmart Marketplace | Yes, with tighter traceability expectations | Item-level traceability and product safety documentation; performance standards on on-time delivery | US-warehouse dropship; the traceability requirement makes undocumented overseas parcels a poor fit |
| eBay | Yes, historically the most permissive | Seller-performance standards and buyer-protection disputes resolved on delivery timing | Either, if the delivery estimate is honest — but a late parcel is nearly always a lost dispute |
| Etsy | Permitted, but the culture and the made-to-order expectation point elsewhere | Buyer expectations around handmade, made-to-order and shipping origin | Print-on-demand; a resold mass-market SKU from a Chinese warehouse contradicts the platform's premise |
| Your own store (Shopify, WooCommerce) | Yes, with no policy gate — and no policy protection | Payment-processor risk, chargeback ratios and consumer-protection law, which is where the gatekeeping moved | All six models; the constraint is the processor's tolerance for dispute rates and delivery complaints |
Note the asymmetry in the last row. On a marketplace, policy protects you from some disputes and disciplines you with others. On your own store, no policy gate exists — which means the payment processor becomes the enforcer, and a chargeback ratio above roughly 1% of transactions is an existential problem for a small merchant. That risk is entirely a function of delivery experience and product honesty, which is another way of saying it is a function of who ships your parcels.
Sourcing Without MOQ: What You Give Up When You Buy One Unit at a Time
The appeal of dropshipping is the absence of an inventory commitment. The cost of that absence is a supply chain you do not control, priced at a markup, with no quality agreement and no ability to differentiate the product. Being clear-eyed about the rungs of the ladder prevents the most common failure: spending a year and an ad budget building demand for a commodity SKU that four hundred other sellers can buy from the same platform.
| MOQ | What you get | Unit cost vs factory FOB | What it unlocks | What it does not |
|---|---|---|---|---|
| 1 unit (marketplace resale, AliExpress, agent, fulfilment network) | A tested SKU you can list tomorrow, with no capital at risk | 2.5x–5x | Demand validation, speed to market, catalogue breadth, testing creative and price points | No differentiation, no packaging control, no price protection, no private label, no QC, and a permanent risk that a bigger seller undercuts you on the same listing |
| 10–50 units (a sample order plus a first paid batch) | Physical inspection of production units, not samples, and a first honest read on batch variance | 1.8x–2.5x | A custom sticker, an insert card, a bundle, a packaging overlay, and the ability to ship a replacement when one breaks | Still an off-the-shelf product; the factory has no incentive to hold your spec |
| 100–500 units | A real purchase order with a spec, a QC checkpoint and a delivery date | 1.5x–1.8x | Custom colour or print, an OEM box, an owner's manual with your brand, and a landed cost you can plan against | Private tooling; expect the same public mold as everyone else in the category |
| 1,000–5,000 units | Factory attention, negotiated terms, deposit and balance structure, inspection access | 1.2x–1.5x | Small modifications to an existing mold, custom tooling inserts, a proprietary accessory, certification in your own name | Exclusivity, unless it is written and priced |
| Private mold | A product nobody else can list | 1.0x plus tooling | Actual IP position, a defensible listing, a price you set rather than inherit | Nothing is guaranteed — the mold is only a moat while the tooling is exclusive and the design is protectable |
The strategic point is that dropshipping is a demand-testing instrument, and demand testing has a natural end. Its job is to find out which SKU converts, at what price, with what return rate — and then you buy inventory on the winner, because the winner is the only place where owning stock pays for itself. Sellers who stay on the one-unit rung for years are paying a 2.5x–5x product markup forever to avoid a decision they have already made. Sellers who climb prematurely buy containers of a product nobody wanted. The sequencing matters more than the rung.
The dropship-friendly product screen
Before any supplier conversation, run the candidate against the screen. Products that fail it do not fail on merit — they fail on the logistics and compliance stack that surrounds them, and each failure has a specific cost attached.
| Criterion | Pass | Why a failure costs money |
|---|---|---|
| Weight and size | Under 500g, under 40cm in the longest dimension, not dimensional-weight heavy | Freight is charged on volumetric weight; a bulky light item can double the shipping line and destroy the per-order margin, and it also pushes you into higher retail price bands where customers expect better service |
| Battery | No lithium cell at all | Lithium batteries bring UN 38.3 transport rules, restricted carrier lines, marketplace battery verification, an insurance premium uplift, and a fire-risk liability profile. Most dropship consolidators refuse or surcharge them |
| Chemistry and biology | No liquid, gel, aerosol, cream, food contact, ingestible or cosmetic | These categories require FDA facility registration or cosmetics documentation, CPSR safety reports in the EU, ingredient labelling and heavy-duty packaging. A dropshipped stock SKU from a closed platform rarely has a file that stands up |
| Children's products | Not marketed for under-12s | Children's products require a Children's Product Certificate based on third-party accredited-lab testing, permanent tracking labels naming the manufacturer and date, and CPSC eFiling data at entry. A stock SKU cannot satisfy a tracking-label requirement it was never made for |
| Sizing and fit | No apparel, footwear, or anything the customer must try on | Fit-driven return rates of 20–30% are unmanageable when each return is a refund with no recoverable unit. The unit economics of sizing on a dropped order are close to the worst in commerce |
| Fragility | No glass, ceramic, mirror or fragile composite | Every damaged arrival is a full refund plus reship, and the damage rate on a single-parcel line is materially worse than on a palletised domestic lane. Damage also produces the reviews that kill conversion |
| Price band | $25–$80 selling price for a domestic-node model; $60+ for a China-direct model | Below the band, fixed per-order costs consume the contribution. Above it, customers expect a level of service a dropship arrangement cannot credibly deliver |
| Competitive density | Not a top-50 bestseller with dozens of sellers on the identical SKU | A commodity SKU that any seller can buy from the same platform on the same terms ends in a price war with no differentiation and no defensible margin |
| Warranty and spare parts | Replacement units or parts are purchasable | You will need 3–8% of volume for warranty replacement. If the supplier will not sell you five extra units, your only remedy for a defect is a refund, which converts a $12 problem into a $60 problem |
| Regulatory category | No medical claim, no radio transmitter beyond a compliant module, no laser, no mains voltage | Each of these imports its own certification regime, and each one makes the buyer — you — the party who must produce a document that a stock SKU does not have |
A product that passes all ten is rare, and that is the point: the screen is doing the work that an inventory commitment would otherwise do. The second-order effect is that the screened product is usually also a better product — heavier, more specific, less commoditised, and easier to explain in an ad — because you have deleted everything that only works at scale on a container.
Compliance Follows the Seller, Not the Warehouse
The most expensive misunderstanding in dropshipping is the belief that a supplier shipping directly means the supplier is the responsible party. Customs and product-safety law do not work that way, and the 2026 rule changes made the gap wider and more visible.
US product safety duties land on importers, distributors and retailers. The Consumer Product Safety Act's reporting obligations apply to manufacturers — a term the regulation defines to include importers — and separately to distributors and retailers. If your name is on the listing and you sold the product to a consumer, you are inside that perimeter regardless of who put the parcel on a plane. A substantial product hazard you learn about is reportable under section 15(b) within a 24-hour window, and the fact that the factory in Guangdong has not answered your email is not a defence.
CPSC eFiling made the compliance file a customs file. Since 8 July 2026, certificate data elements for products subject to a consumer product safety rule are transmitted with the entry. If your goods enter as individual parcels, you often have no visibility into what was declared — while still being the seller whose listing carries the claim. If your goods enter in bulk to a 3PL and you are the importer of record, you own the obligation, which at least means you can control it. The version of this that goes wrong is a supplier shipping DDP on a materially under-declared value: the duty saved is real, the customs exposure is yours if the arrangement names you, and the penalty is not proportional to the saving.
Europe requires a responsible person, and it must be yours. Under the General Product Safety Regulation in force since December 2024, a product placed on the EU market needs an economic operator established in the EU who is accountable for it — manufacturer, importer, authorised representative or fulfilment service provider — and whose name and address appear on the product or its packaging and on the listing. Some Chinese dropship suppliers will offer an EU responsible person as a service. Unless that responsibility is contractually assigned to you or to an entity acting for you, it does not cover your listing: a responsible-person declaration naming another seller's brand is the wrong document for your product, and marketplace compliance checks will treat it that way.
CE marking and declarations of conformity are product-specific and brand-specific. A CE mark on a stock SKU evidences that somebody declared conformity for that model. The declaration carries a name and an address. Using a stock SKU means the declaration is in the manufacturer's name, or a brand you have no relationship with, while you market the product and carry the listing. Where a market surveillance authority asks for the file, "the supplier has it" is not a file.
Marking obligations are physical, not administrative. US country-of-origin marking requires the article itself to be marked with its origin; GPSR requires manufacturer details, the EU responsible person, warnings and batch identification. You cannot retrofit any of these onto a parcel that is already in a mail bag. If the supplier's packaging does not carry them, either the supplier changes the packaging or the product is not sellable in the territory you are targeting.
Liability exposure is unchanged and follows the chain. Strict liability in the United States reaches every commercial seller in the distribution chain, including sellers who never touched the goods. If the product injures someone, the party with a bank account in the plaintiff's jurisdiction is the party that funds the settlement. A dropship programme does not reduce that exposure; it hides it until the claim arrives, and it also reduces the quality of the documentation you will need in order to defend or insure the product. A commercial general liability policy with products and completed operations cover is the floor, and it is priced on the file you can produce.
The honest summary: a dropship SKU should be restricted to categories where the compliance stack is light (non-children's, non-battery, non-ingestible, non-electrical hardgoods), and where the supplier will contractually provide the file in your name or acknowledge in writing which party carries which obligation. Anything else is a product you should be importing in bulk with a proper compliance file, not reselling one parcel at a time.
Returns, Refunds and the Margin Line Nobody Models
Returns are where dropship businesses discover the difference between a contribution number in a spreadsheet and cash in a bank account. On a dropped parcel, the unit is generally not coming back. The customer has it, and the cost of retrieving a $15 item from another continent exceeds its value, so the standard remedy is a refund without a return — revenue reversed, goods gone, ad spend already spent, and the order counted as a loss rather than a break-even.
| Category type | Typical return rate | China-direct handling | US-warehouse handling |
|---|---|---|---|
| Apparel, footwear, anything fitted | 20–30% | Refund-only in practice; the item is not worth recovering. Every return is a full loss | Returnable and resellable at a resale rate of maybe half; still the worst category economics in commerce |
| Consumer electronics and accessories | 8–12% | Refund-only, high defect share, spare-parts reserve essential | Return inspected, most units restocked, defective units claimed against the supplier |
| Home goods, kitchen, tools | 5–8% | Refund-only on most, damage-driven rather than preference-driven | Returnable, often restocked, freight damage recoverable from the carrier in some cases |
| Beauty, consumables, personal care | 3–6% | Refund-only and non-resellable by law in most cases; also carries the heaviest compliance load | Same, but the compliance file is at least producible |
| Heavy, oversize, assembled furniture | 5–10% | Not a dropship product at all — the return freight exceeds the margin | Domestic return freight is expensive but survivable; damage rate is the real cost driver |
Model the refund allowance at the measured rate for your category and add three further costs that a spreadsheet usually misses. The refunded ad spend: a refunded order was acquired with a click you already paid for, so a 6% refund rate on paid traffic costs 6% of your customer acquisition cost as well as 6% of revenue. The replacement order: if you reship rather than refund, you pay the product and shipping cost twice on one sale. The platform consequence: on a marketplace, a refund is also a defect metric, and defect metrics reduce the traffic the platform gives you next month.
Two operational decisions contain the damage. First, a US returns address — either your 3PL, a dedicated returns partner or a domestic supplier that accepts returns on your behalf — because marketplace policy and consumer expectation both assume one, and because a domestic return is the only version of a return that can be resold. Second, a spare-parts reserve: buy 5% extra units (or a batch of the components that fail — cables, seals, brackets, mounting hardware) and hold them at the returns address. Shipping a $3 part that fixes a $60 order beats refunding it, and it converts a defect into a review that mentions your support rather than your supplier.
Two Programmes, Side by Side: The One That Works and the One That Does Not
Everything above collapses into two example programmes with the same gross revenue ambition and radically different outcomes.
| Line | Programme A — the 2024 model sold in 2026 | Programme B — the 2026 model |
|---|---|---|
| Product and price | $19.99 consumer gadget, 300g, marketplace resale from a closed platform, listed on Amazon by FBM | $59.99 home and outdoor hardgood, 700g, sourced from the same factory but stocked at a US fulfilment node |
| Fulfilment | Supplier ships each order from China, 9–16 days, DDP line | US node picks, packs and ships in 3–5 days; inbound shipped consolidated, duty paid once at wholesale value |
| Delivered cost per order | $21.90 including duty and per-parcel clearance | $22.00 including the network's per-order fee; no per-order customs event |
| Contribution at list price | −$9.99 per order (−50% of revenue) | +$16.75 per order (+27.9% of revenue) |
| Break-even price | $35.57 — the product cannot be sold profitably where the market prices it | $34.64 — comfortably below list |
| Ad model | Break-even CPA is negative, so ad spend is a pure loss funded by optimism | Break-even CPA of roughly $16.75 supports a $7–9 target CPA with headroom, which is achievable in a category with a real problem to solve |
| Returns | Refund-only, ~8% of orders, plus a delivery-related defect trail on the account | Domestic returns accepted, inspected, a share restocked, and a cross-border-free returns process on the listing |
| Compliance position | A stock SKU with somebody else's declaration, no tracking label, no responsible person, and a supplier who will not sign a quality agreement | A named model with a file you can produce, a QC checkpoint at inbound, and a supplier who sells you spare parts because the volume justifies it |
| What happens in month nine | The account takes a performance hit, the margin never appears, and the owner concludes that dropshipping stopped working — when in fact the cost structure and the price band were mismatched from day one | The SKU earns the right to be bought in bulk; the next order goes straight from the factory to the 3PL with a better unit price and a lower landed cost |
Programme A is not a bad operator. It is a good operator running a 2024 cost model against 2026 rules, selling a $20 product in a structure that needs $36. Programme B is the same person one insight later: the price band has to carry the fixed cost, and the fixed cost disappears when the parcel is domestic.
The Ten-Test Verification Battery for a Fulfilment Supplier
No platform's badge, rating or review count is evidence that a fulfilment supplier will behave. These ten tests are cheap, and each one is designed to surface a failure before it costs you an account.
- Seller-of-record and brand-free packaging, in writing. Get a written confirmation that all orders ship without supplier branding, invoices, cards or QR codes, and under your documentation. A supplier who will not put that in writing will not do it in practice.
- Stock-depth reality test on your top SKUs. Order one unit per week for three weeks from widely separated destinations. Watch specifically for stock-outs, silent substitutions (a different colour, a newer variant, a different accessory) and rolling changes to the "same" SKU.
- Dispatch-time reality test. Place five orders to five different ZIP codes or postcodes on the same day. Measure two separate times: when the tracking number was created, and when the carrier first scanned the parcel. The gap is the tell — label-created is not shipped.
- Shipping-line and duty test. Establish exactly which line is used, whether duty is prepaid (DDP) or payable by the customer (DDU), and what the door charge is. A DDU parcel that arrives with an unexpected bill produces refusals, refunds and one-star reviews, and the supplier will not mention it.
- Declaration-honesty check. Ask what value and HS code the supplier declares on your behalf. An under-declared parcel is cheap today and a penalty later, and if the arrangement names you, it is your penalty.
- Branding and materials audit on the delivered box. Photograph everything in the parcel: the outer label, the inner slip, the insert, the warranty card, any QR code. Anything that identifies another company is a marketplace violation waiting to be reported by a customer.
- Return path test. Return one order and measure the whole cycle: where the parcel went, who paid the postage, how long the refund took, and what you learned about the unit. A supplier with no return path has no product feedback loop either.
- Spare-parts and warranty test. Ask to buy five replacement units or the two components that fail most often. A supplier who cannot sell you parts cannot support a warranty, and you will fund every defect as a full refund.
- Compliance-file test. Ask, for the exact model you are listing: the declaration of conformity or test report, the manufacturer's name and address, the EU responsible person where relevant, and the age-grading statement. The answer should be documents, not assurances.
- Escalation test. Ask who answers when a stock-out hits during a peak weekend, in writing, with a response-time commitment. Fulfilment failures are always discovered at the worst possible hour, and the difference between a stock-out and a crisis is whether someone picks up.
The Eight-Step SOP
- Choose the model from the price band and the category, not from the supplier's pitch. Under $15: POD or domestic wholesale. $15–35: domestic node only. $35–60: domestic node preferred, China-direct viable. $60–150: China-direct workable for light compliant goods, domestic node for anything with real demand. Above $150: hold stock. Write the decision down before sourcing begins, because it determines which supplier you are even looking for.
- Screen the product against the dropship-friendly list. Weight, battery, chemistry, children's-product status, sizing risk, fragility, price band, competitive density, spare-parts availability and regulatory category. Any failure is a deliberate cost you are accepting, priced in advance.
- Sample three suppliers, and test them like a customer, not like a buyer. Three orders each, to three addresses, and score dispatch time, packaging branding, product condition, documentation and the parcel's whole contents. The supplier who answers fastest is not the same as the supplier who ships fastest.
- Fix the paperwork before the first order. A written supply agreement naming you as seller of record, requiring brand-free packaging, defining dispatch SLAs and stock-accuracy commitments, allocating the return path, and setting out the compliance documents the supplier must provide. This is the only moment where you have leverage.
- Decide the customs posture deliberately. DDP (higher per-order cost, better customer experience, no refusals) or DDU (cheaper headline, door charges, refusals at your expense). For bulk inbound shipments where you are the importer of record: appoint a customs broker, post a bond if required, confirm MPF treatment, and make sure certificate data can be filed under eFiling.
- Stand up a domestic returns address and a spare-parts reserve. A 3PL partner, a returns specialist, or a domestic supplier who accepts returns for you — and 5% of the projected volume held as replacements and consumable spares.
- Price to the 2026 cost floor, including fees that have not started yet. Build the model with duty, per-parcel entry overhead and the scheduled 1 November 2026 postal handling fee included, then stress it: what price makes this work if the effective duty rate rises five points, and what happens if the refund rate doubles.
- Reinvest the first $10,000 of contribution into stock, not into more ads. The unlock in this model is not a better creative or a cheaper click. It is owning the inventory of a SKU that has proven it converts, which cuts the unit cost by 30–60%, removes the supplier's per-order fee, and moves the duty event from every parcel to one consolidated entry.
Dropshipping from China in 2026 is a legitimate, workable business in five of its six forms — as long as the price band carries the fixed cost per order and the compliance file travels with the product. What ended in 2025 and 2026 was not the model. It was a pricing assumption: that a parcel could cross the border for nothing and that the cost of buying one unit at a time would stay small. The sellers who internalised that and re-priced found the model got better, because the discipline that surviving the new arithmetic imposes — real products, real price points, domestic nodes, documented compliance — is exactly the discipline that separates a durable catalogue from a spreadsheet that only worked while the tariff line read zero.
This guide is informational and reflects published 2026 market parameters and regulatory timelines; it is not legal, tax or customs advice. Duty rates, entry procedures and fee schedules change on published dates and vary by HS code, value, mode and origin, and marketplace requirements change without notice — verify your own classification, entry treatment and channel rules with a licensed customs broker and a lawyer before relying on any figure in this article.