Choose a dropshipping agent on customs capability first, not product price. Since the $800 de minimis exemption ended, every US-bound parcel needs a formal entry and duty payment, so an agent without US or EU forward stocking now costs you margin on every single order.
The agent question used to be about price per unit. Since August 2025 it has been about who files the customs entry, and most agents still do not have a clean answer.
In fiscal 2024, over 1.36 billion low-value parcels entered the United States under the $800 de minimis exemption, up from 139 million in 2015. Almost all of them cleared customs on a bill of lading with no duty collected. That door is now closed, and it closed on the exact business model most dropshipping agents were built to serve.
If you are choosing an agent right now, the vetting questions that worked two years ago will not surface the thing most likely to break your margin. Product price, shipping speed and English fluency are still worth checking. They are no longer the deciding factors. The agents worth hiring in 2026 are the ones that solved customs. The ones that did not are still quoting a per-unit price and letting you discover the entry costs on your own invoice.
What follows is a vetting sequence for merchants sourcing from China. If you are doing $10K months and testing your first three products, the calculus is different and I will flag where. If you are past $2M with a stable catalog, you may be past agents entirely, and there is a section on that at the end.
Define your agent requirements from four numbers before you contact anyone: monthly order volume, your top five SKUs by units, your destination country mix, and your current landed cost per unit. Category matters far less than most sourcing advice suggests, because a competent dropshipping agent is a logistics operator first and a product specialist second.
Order volume sets the floor. Below roughly 100 orders per month you have almost no leverage in a negotiation and almost no ability to hold an agent accountable, because your account is not material to them. Between 100 and 1,000 orders per month you are in the range where agents compete for you and where private labeling, custom inserts and negotiated per-unit pricing become genuinely available. Above 1,000 orders per month you should be asking whether an agent is still the right structure at all.
SKU concentration is the number most merchants skip. If 80% of your units come from five SKUs, you can pre-position inventory and your agent selection changes completely. If your volume is spread across forty rotating SKUs, you need an agent built for constant sourcing, and you will pay for that flexibility in per-unit cost and in transit time.
Destination mix determines which regulatory regime you are actually operating under, which is the subject of the next section. A store selling 90% into the United States has a very different agent requirement than one splitting evenly between the US, the EU and Australia. Write these four numbers down before the first call. Agents will scope to whatever you tell them, and vague inputs produce a proposal designed to sound good rather than one designed to work.
Direct-from-China parcels no longer enter the United States or the European Union duty free, which means the cost structure that made agent-fulfilled dropshipping profitable no longer exists in its old form. This is the single largest change to the model in a decade and it is not a temporary policy swing.
In the United States, duty-free de minimis treatment was suspended for China and Hong Kong on 2 May 2025 and for every other country on 29 August 2025. On 24 June 2026, US Customs and Border Protection indefinitely suspended the exemption by regulation for merchandise arriving by any mode other than the international postal network. The practical effect is that a shipment valued at $800 or less can no longer be released from manifest. It needs a proper entry with a 10-digit classification and applicable duties paid. The Entry Type 86 process is suspended, leaving Entry Type 11 as the main informal route. The statutory repeal of the exemption takes effect 1 July 2027, so planning around a reversal is not a strategy.
The EU moved on the same problem. The Council of the European Union abolished its own €150 duty relief threshold with final approval in February 2026, citing 4.6 billion small packages entering the EU in 2024, 91% of them from China. From 1 July 2026 an interim flat €3 per item category applies until July 2028, charged per tariff sub-heading rather than per parcel. A shipment containing a silk blouse and two wool blouses carries €6, not €3.
Ask every candidate agent three questions on this. Who is the importer of record on my orders. Which entry type do you file and can you show me a recent filing. How do you handle a shipment held for classification. An agent that cannot answer all three is not an agent, it is a forwarder with a website.
There are four viable sourcing structures in 2026 and each breaks at a predictable point, so pick the one that matches your current volume rather than the one that matches your ambition. Choosing the model you want to grow into is the most common and most expensive mistake at the $500K to $2M stage.
The middle two rows are where the interesting decision sits. An agent shipping direct to the customer keeps your capital free and lets you kill a SKU in a week, which is exactly what you want while you are still finding out what sells. The moment a SKU proves itself across three consecutive months, the per-order entry cost on that SKU becomes pure waste, and moving it into forward stock usually pays for itself inside a quarter. Most merchants make this move too late because they run the decision on the whole catalog at once instead of SKU by SKU.
Verify an agent through their existing clients and a paid sample order, because every agent website in this category makes identical claims about quality, speed and support. Referrals from merchants in your niche remain the highest-signal input, but ask for specifics rather than a recommendation: what went wrong, how long it took to fix, and what the agent did when they were at fault.
Directories are useful for building a longlist and close to useless for shortlisting. The same warning signs that apply to supplier directories and the red flags that show up in them apply to agents. Ongoing monthly service fees separate from per-order costs deserve a hard question. So does an agent selling to consumers at the prices they quote you. So does a minimum order quantity attached to what is supposedly a no-inventory arrangement.
Trade shows still work if your volume justifies the trip. Canton Fair and the Yiwu market circuit put you in front of both agents and the factories behind them in the same week, which is the fastest way to learn whether your agent is sourcing directly or reselling from a middleman. Below roughly $500K in annual revenue the travel cost rarely returns, and a structured video call with a factory walkthrough gets you most of the way.
One research step almost nobody runs: place a small order through the agent’s own storefront or a competitor’s store that uses them, and time everything. You will learn more about processing time, packaging and tracking quality from one $30 order than from four discovery calls.
Pay for a sample run of 10 to 20 units per SKU and inspect them yourself before any agent handles a real customer order, because quality control claims are the easiest thing on an agent’s website to write and the hardest to verify remotely. Budget $200 to $500 and two weeks for this. Merchants who skip it almost always pay more later in refunds and chargebacks.
What you are testing is not whether the product is good. It is whether the agent’s inspection process exists. Ask for the inspection criteria in writing before the sample ships, then deliberately check the items against those criteria yourself. If the agent says they check for scratches, seam integrity and functional defects, count how many of your 20 units have a problem they should have caught. A 5% miss rate is workable. A 20% miss rate at sample stage, when they know you are watching, tells you what production volume will look like.
Photograph everything on arrival, including packaging condition. Packaging is the cheapest tell in the whole evaluation. An agent that ships your sample in a reused box with a scuffed poly bag will do the same to your customers, and that unboxing moment is the only physical contact your brand has with the buyer.
If private labeling matters to you, test it in the same run. Ask for your logo on the packaging and a branded insert on five of the twenty units. Agents will say yes to customization on a call and then quote a setup fee and a 500-unit minimum once you commit. Finding that out during the sample stage costs you nothing.
Ask every agent for processing time and transit time as two separate numbers, because agents quote the second and merchants get burned by the first. Processing time is how long the agent takes to source, inspect, pack and hand the parcel to a carrier. Transit is the carrier’s clock. An agent advertising 7 to 12 day delivery who takes four days to process an order is really selling you 11 to 16 days.
This is the same discipline that applies to your own storefront, where separating processing time from transit time is what stops customers feeling misled by a fast shipping promise. You cannot communicate an honest delivery window to your buyer if your agent will not give you an honest one first.
Get processing time as a range with a stated ceiling, not an average. Averages hide the tail, and the tail is where your support tickets come from. Ask what happens to processing time during Chinese New Year and in the two weeks before Black Friday. An agent who says nothing changes is either lying or has never operated at volume. The honest answer is that processing time roughly doubles for two to three weeks and the good agents tell you the dates in advance so you can pre-position stock.
On tracking, confirm the carrier and the scan quality, not just that a number exists. Agents routinely provide tracking numbers that show one scan at origin and nothing again until delivery. That produces a support ticket on every order over ten days old. Ask which carriers they route through and whether the tracking feeds cleanly into an app like AfterShip or 17TRACK from your Shopify admin.
Evaluate an agent on how they handle a problem they caused, because response time on a routine question tells you almost nothing about the relationship you are actually buying. Every agent answers a pre-sales email within an hour. The question is what happens on the day a shipment of 60 units goes missing in transit and your customers are emailing you.
Ask candidates directly for a recent example of an order they got wrong, what it cost, and who absorbed it. The answer separates operators from salespeople faster than any other question in the process. An agent who claims they have never had a failure at volume is telling you they have never operated at volume, and an agent who describes a failure honestly and names what they changed afterwards is showing you the thing you are really hiring.
Get the replacement and refund policy in writing before you send a single real order. Specifically: who pays for a replacement unit on a defect, who pays the shipping on that replacement, what the evidence threshold is, and how many days it takes. Agents that handle this well will have a documented process. Agents that handle it badly will tell you it depends on the situation, which means it depends on how hard you push.
Set a communication cadence rather than relying on ad hoc messages. A weekly summary covering orders shipped, orders held, and any SKU running low on stock at the agent’s warehouse takes them fifteen minutes and prevents most of the surprises. If an agent resists that, they are managing more accounts than they can service.
An agent stops being the cheaper option somewhere between 500 and 1,000 orders per month on a stable catalog, and past that point the per-order customs and handling costs typically exceed what bulk importing plus domestic fulfillment would cost. This is not a failure of the agent. It is the model working exactly as designed and then reaching its ceiling.
The signal to watch is not revenue, it is SKU stability. If your top ten SKUs have held their position for two consecutive quarters, you have enough forecasting confidence to import in bulk and pay duty once per container rather than once per parcel. The breakeven math on outsourced fulfillment generally turns favourable around 100 orders per month for domestic stock, and the gap widens considerably once per-order entry costs enter the comparison.
If you do make that move, the advice from a merchant success lead at Shopify Fulfillment Network still holds and applies just as well to agents: choose a partner for where the business will be in one to three years, not where it is today, because switching fulfillment partners is disruptive even when the transition is well planned. The merchants who handle this badly are the ones who switch during Q4 because their agent buckled under holiday volume. Plan the transition for a low-volume window and run parallel for two to four weeks if your volume allows it.
A hybrid structure is often the right answer rather than a clean switch. Keep your proven SKUs in domestic stock and keep the agent for new product testing, where their ability to source a small run quickly is genuinely valuable and where per-order duty on 30 test units is an acceptable cost of learning.
The agent decision is not permanent and it should not be treated as one. Review it every two quarters against your actual order profile, and the review will usually take an afternoon.
Yes, if you are still validating products, and probably not if your catalog has stabilised above roughly 500 orders per month. The end of de minimis raised the per-order cost of direct-from-China fulfillment but it did not change the reason agents exist, which is sourcing speed and the ability to test a SKU without buying inventory. That flexibility is still worth paying for while you are learning what sells. Once a SKU proves itself across three consecutive months, the per-order entry cost on that SKU becomes waste, and moving it into forward stock or a domestic 3PL usually pays back within a quarter.
A dropshipping agent sources products for you and ships them, while a 3PL only stores and ships inventory you already own. The agent handles supplier negotiation, purchasing, quality inspection and fulfillment as one bundled service, which is why they suit merchants without the capital or forecasting confidence to buy stock upfront. A 3PL assumes you have already made the buying decision and holds your goods in a warehouse near your customers. The practical difference in 2026 is customs: an agent shipping parcel by parcel triggers an entry on every order, whereas a 3PL model pays duty once on the inbound container.
Place a paid sample order of 10 to 20 units per SKU, ask for the inspection criteria in writing beforehand, then check the units against those criteria yourself. Budget $200 to $500 and about two weeks. You are testing whether the quality control process exists, not whether the product is good. Count how many units have a defect the agent should have caught: a 5% miss rate when they know you are watching is workable, and 20% tells you what production volume will look like. Photograph the packaging condition on arrival, and test any private labeling in the same run.
Ask three questions: who is the importer of record on your orders, which entry type they file and whether they can show a recent filing, and how they handle a shipment held for classification. Since US Customs and Border Protection indefinitely suspended the de minimis exemption in June 2026, shipments valued at $800 or less can no longer be released from manifest and need a proper entry with a 10-digit classification. If you sell into the EU, add a fourth question about the flat €3 per tariff sub-heading that has applied since 1 July 2026. An agent who cannot answer these is a forwarder, not an agent.
Watch SKU stability rather than a volume number, though the transition usually makes sense somewhere between 500 and 1,000 orders per month. If your top ten SKUs have held their position for two consecutive quarters, you have the forecasting confidence to import in bulk and pay duty once per container instead of once per parcel. A full switch is rarely the right move. Most merchants at this stage do better keeping proven SKUs in domestic stock while retaining the agent for new product testing, where sourcing a small run quickly is genuinely valuable and per-order duty on 30 test units is an acceptable cost.
