On 16 May 2024 the European Commission opened an anti-dumping investigation into flat-rolled steel plated or coated with tin—tinplate—originating in China, and published the Notice of Initiation in the Official Journal. A customs broker forwarded a screenshot to a customer group chat. Chen, who runs a steel fastener plant in Haiyan, Zhejiang, glanced at it and scrolled past: the case was about the sheet steel used to make cans, not about his stamped and threaded parts. Three months later, another steel product case was opened, and this time it landed on his desk.
His plant is not large: 260 employees, RMB 320 million of revenue in 2024, 46% of it exported to the EU, mostly to three industrial-parts distributors in Germany and Poland. The day the second notice arrived he understood the real nature of the exercise. Anti-dumping is not litigation in the ordinary sense, where a good argument can be made at any point. It is a calendar with a handful of deadlines, each of which can only be missed once. Make yourself known within 7 days. Submit the questionnaire within 30 days. By day 90, customers start asking whether you can still supply reliably. By day 240, provisional duties are published. By day 335, the final disclosure arrives. Chen ended up with a company-specific rate of 31.6%; competitors in the same case who never came forward got the country-wide rate of 78.5%. That 46.9-point spread was not luck. It was eleven months of things done right and things done wrong, added up.
The day the notice is published, find three dates
A notice of initiation runs to dozens of pages, and most exporters read only the headline: our product is under investigation. What actually matters is three things.
First, the scope. The notice defines the product concerned, the tariff codes, and the product control numbers (PCNs) that distinguish models. The PCN decides which of your orders fall inside the case and, more practically, how every subsequent data table is organised. Chen's team ignored this at first and filed two fasteners with different surface treatments under one code. The questionnaire came back for correction, costing two weeks they did not have.
Second, the two distinct periods. The investigation period covers the most recent twelve months and is used to calculate the dumping margin. The injury period covers roughly three years and is used to assess whether the EU industry suffered material injury. Cost data, sales data and customer structure have to be prepared separately for each. Mixing them is self-inflicted damage.
Third, the deadline to make yourself known. In EU proceedings, exporting producers must come forward in writing, state whether they wish to be sampled, and submit the sampling form within 7 days of publication of the notice. Under the Chinese system run by MOFCOM, questionnaires are issued after initiation and must be completed within 37 days of the date the questionnaire is sent, under the Anti-Dumping Regulations. Which brings us to the most consequential sentence in the whole procedure: where a party fails to provide the required information, or otherwise falls within Article 21 of those Regulations, the authority may make its determination on the facts available and the best information otherwise available.
One more step gets overlooked: registration. The Commission made imports of tinplate subject to registration by Implementing Regulation (EU) 2024/2731 of 24 October 2024. Registration is the hook for retroactive collection: once provisional duties are imposed, duties can be collected on registered imports for up to 90 days before the provisional measure took effect. In plain terms, the containers shipped between initiation and the provisional duty could be taxed after the fact. Whether to keep shipping during that window is a cash-flow decision, not a test of nerve. Chen switched new customers to payment in advance and moved existing customers to FOB terms, so the exposure sat with the buyer.
Sampling and separate rates: being selected and not being selected are two different fates
When there are many exporters, the authority does not audit all of them. It selects a sample—usually two to five companies—for a full examination, gives the other cooperating exporters a weighted-average rate, and gives non-cooperators the country-wide rate. Three tiers of duty, and the gaps between them can be brutal.
In the parallel EU tinplate case, the definitive regulation (EU) 2025/1042 set the following: WISCO-Nippon Steel Tinplate at 13.1%, Shougang Jingtang United Iron & Steel at 46.8%, the other cooperating companies listed in the Annex at 24.6%, and all other imports originating in China at 62.3%. Same product, same market, and a spread of nearly five times between the best and the worst. The difference was not product quality. It was whether a company showed up inside the deadline.
How hard is that deadline? In the same case, a Jiangsu company asked to be included in the investigation as a cooperating exporting producer on 14 January 2025. The Commission refused, because the company had not made itself known within the deadline set out in the notice of initiation, as required by Article 17. Eight months late, and the window was closed. The refusal is printed in the body of the definitive regulation, where anyone can read it. That document, rather than any argument, is what convinced Chen's board to pay for counsel.
Not being sampled does not mean being left alone. Companies outside the sample must file a separate submission—in the EU, a cooperating non-sampled producer submission; in the US, a separate rate application. The US Department of Commerce is blunt about non-market-economy exporters: without a separate rate application, or if the application does not satisfy Commerce that you are free of government control, you fall into the China-wide rate. The EU removed the old binary of market economy treatment and individual treatment when it amended the basic Regulation in 2017, replacing it with the significant distortions approach plus analogue-country data, but cooperating exporters can still obtain an individual company rate, and the precondition is the same: complete answers that reconcile with the books.
The sample is not a lottery. Investigating authorities tend to select the largest exporters with clean records and a demonstrated willingness to cooperate. Volume is history and cannot be changed; record-keeping and willingness can be prepared. Chen ranked fourth by volume that year, right at the edge. What put him in the sample was a separate ledger he had kept for shipments through affiliated traders. When reviewers asked, he produced it the same minute.
One warning that costs real money: shipments through affiliated traders, overseas warehouses and third-country transshipment must be reported too. If undeclared volumes come to light, the whole response is treated as incomplete and the authority reverts to facts available, wiping out everything the money bought.
Questionnaires and the submission timetable: every figure will be questioned
A company questionnaire with annexes runs to 400 to 800 pages. The content divides into five blocks: corporate structure and ownership, including all affiliates; the model breakdown of the product concerned; the transaction-by-transaction domestic sales listing; the transaction-by-transaction EU export listing; and cost of production and expense allocation. The EU allows about 30 days; MOFCOM allows 37 days from the date the questionnaire is sent. Both clocks start on the date of issue, not the date you read it.
Here is where the time goes. Every transaction has to tie to an invoice, a customs declaration and a bank receipt. One mismatch and you are hunting for source documents. Chen's three accountants plus an outside firm worked six full weeks and pulled two all-nighters at the end. The first submission was rejected because exports outside the EU had been mixed into the EU sales table. The second was rejected because one affiliate sale had no written basis for its transfer price.
A few specific traps, worth internalising early:
- PCN codes must match across the questionnaire, the cost tables and the verification room. One wrong code recalculates every table.
- Domestic prices must be VAT-exclusive and must reconcile with your invoicing system. Many plants issue a different invoice format per customer; standardise before the case starts.
- Cost allocation must be explainable: the basis used, why it was chosen, and what the duty would be under an alternative basis. The authority does not demand that your method be the only correct one, but it demands that the method be consistent and verifiable.
- Do not outsource translation to people who do not know the industry. One mistranslated technical term adds twenty items to the next list of questions.
- Submit through the designated channel—MOFCOM's trade remedy platform in China, the Commission's TRON system in the EU—and make sure the electronic and paper versions are identical.
Chen's own shift in attitude is worth naming. In month one he was still hoping to gloss over weak spots. A week before verification he understood that every figure in the response would be checked against the ledger. Vagueness does not reduce work; it moves the risk from numbers that can be inspected to discretion that cannot, and discretion almost always goes against you.
Surrogate countries, surrogate values and market economy status: why your costs do not count
China is not treated as a full market economy in either US or EU anti-dumping practice. Commerce maintains a list of non-market economies that includes China, and normal value is reconstructed using costs from a surrogate country, typically India, Thailand, Brazil, Turkey or Malaysia. Since amending its basic Regulation in 2017, the EU has formally dropped the non-market economy list and works instead through country reports on significant distortions plus analogue country data. The practical effect is much the same.
What that means for a factory: the fact that your steel mill price is lower than an Indian mill's price is not taken into account, while a high Indian electricity or labour cost pushes your margin up. Managers hearing this for the first time usually call it unfair. Procedure does not accept that argument. What it does accept is narrower: evidence that a particular input was bought at market prices under market conditions, so that your own purchase price rather than a surrogate value should be used. Some cases have been won on exactly that point, using purchase contracts, proof of payment, price comparisons in the source country and public global quotations for the input.
Chen made a deliberate trade-off. He abandoned two lines of argument: a full claim that his company should be recognised as operating under market conditions, which has a low success rate, a high cost and a drag on the main case; and a fight over the choice of analogue country, which is discretionary and, once decided, only reversible years later in court. He put everything into the accuracy of the transaction listings and into whether coiled steel—his single largest input—could be valued at his own purchase price. The first set the floor of the duty; the second clawed back several points. The same logic generalises: sort your arguments into those that are cheap with a predictable payoff and those that merely express how you feel.
Verification, injury defence and price undertakings: the three levers that actually move the result
On-site verification usually takes place one to two months after the response is filed. Two to three weeks beforehand you receive a verification agenda listing the tables and vouchers to be examined. The team is typically two to four people, on site for three to five days, sitting in a meeting room and asking you to pull data out of your system live.
The traps Chen hit are worth copying. Domestic invoices had no uniform format, so three random samples failed to match the filed data; an affiliate distributor's cost-plus price had no written basis, was judged unreliable, and was replaced with surrogate data, adding points to the duty; two machines had depreciation periods that did not match the declared figures and required a written explanation. The iron rule of a verification room: if you cannot answer on the spot, say you will supplement in writing. Do not guess. Guessing wrong costs far more than not knowing.
Disclosure follows verification. In the tinplate case the Commission gave parties a summary of the proposed duties and the calculation of the margins on 17 December 2024, and the final disclosure came in late April 2025—eleven months after initiation on 16 May 2024. Parties normally have about ten days to comment on the final disclosure, and late submissions are disregarded. That is the last chance to change a number, and it is precisely when many companies relax.
Injury defence is a different road: it does not attack the rate, it attacks the case. The requirements are dumping, injury, and a causal link between them; remove one and the proceeding should end. Chen filed user submissions together with two can makers and one appliance-parts manufacturer, arguing that EU capacity was insufficient, that models did not match demand, and that price increases were driven mainly by energy and logistics. The injury finding survived, but they obtained something concrete: no retroactive collection on registered imports. In theory, those retroactive duties would have exceeded their entire legal bill.
Price undertakings are the third lever and the most misunderstood. The EU model is a minimum import price, mapped to a model list, combined with quarterly reporting and acceptance of unannounced verification. The EU–China solar undertaking signed in 2013 survived five years and was terminated in 2018, with a group of companies pursued for violations committed during the undertaking period. The instrument converts a one-off response cost into permanent compliance cost, which rarely pays for a producer with many models, small individual orders and a scattered customer base. The US does not use undertakings in the EU sense at all, so the practical negotiating room in a US case lies in scope: an exclusion request for a specific model is often more valuable than an argument about the margin.
The cost of not cooperating, and when to stop spending
Not cooperating does not mean saving legal fees and carrying on. It means receiving the worst available data. Article 18 of the EU basic Regulation allows the use of facts available, and the tinplate numbers show what that means: 62.3% for all other imports from China. The US goes further with adverse facts available, the AFA doctrine, under which non-cooperating exporters in a non-market economy case fall into the country-wide rate—rates of 100%, 200% and 300% have all appeared in published determinations. The meaning of such a rate is simple: your goods no longer enter that market unless you rebuild the supply chain entirely.
Cooperating is not a permanent fix either. Anti-dumping measures generally run five years, and the authority publishes an expiry notice before they lapse. In a parallel example, EU measures on iron and steel fasteners from China come up for expiry around February 2026, and the Commission's notice C/2026/2827 of 29 May 2026 relates to arrangements of that kind; if no one requests an expiry review, the measures simply lapse and the market reopens. But if you are on the list, you are on the list for the review as well. And if you never cooperated in the original case, joining at review stage runs into exactly the same seven-day deadline problem. Companies lose money every year by simply not knowing a sunset review was under way.
Then there is the arithmetic, which is what owners care about. Chen's eleven months cost: RMB 1.18 million in outside counsel, RMB 260,000 in accounting and translation, and roughly RMB 400,000 in internal labour for three people over eight months—about RMB 1.84 million in total. On the other side, EU exports ran at RMB 147 million a year at roughly 12% gross margin. At the 78.5% country-wide rate, essentially all of that business disappears. At 31.6%, about 60% of the orders can be held. It is not a question of whether the money should have been spent. It is two definite numbers compared.
When should a company stop? Three signals: EU exports are under a tenth of revenue; mature suppliers for the product exist in the surrogate country; or the company is already on a non-cooperation list from a previous case. In those situations, funding a full original-investigation defence is poor value. Stopping the full defence is not the same as doing nothing, though. At minimum, file the non-sampled producer response or the separate rate application to hold a place. It costs an order of magnitude less than a sampled defence and still creates the possibility of moving from 78.5% to a middle tier. The real choice is not fight or don't fight; it is full defence or placeholder.
As for laying the groundwork, Chen now does four things. He keeps a register of trade remedy measures in every market, listing tariff codes, rates and expiry dates. He makes origin genuinely real: processing in a third country has to pass a substantial transformation test, or an anti-circumvention investigation will extend the measure to the third-country plant as well. He pushes core orders to end customers rather than intermediaries, because affiliates and middle layers are the first things stripped out during verification. And he standardises PCN coding and invoice formats for all EU and US shipments from day one, which saves two months when something happens. None of that costs much. It determines whether the day an investigation opens finds you starting from zero or starting from seventy points.
Frequently Asked Questions
What must an exporter do first after an anti-dumping notice is published?
Confirm the scope, the tariff codes and the PCN structure to establish whether you are inside the case, then come forward in writing within the legal deadline. In the EU, exporting producers must submit the sampling form within 7 days of publication of the notice of initiation. In China, MOFCOM requires a complete questionnaire response within 37 days of the date the questionnaire is sent. Neither deadline can be extended, and a missed deadline writes off the preparation already done. Archive sales invoices, customs declarations and production records by PCN, and appoint one contact each in finance, export sales and production.
If we are not selected for sampling, do we still need to respond?
Yes, and this is usually the best value step available. Exporters outside the sample must file a cooperating non-sampled producer submission, and in a US case a separate rate application, or they fall into the country-wide rate. In the tinplate case, the cooperating companies outside the sample received 24.6% while all other imports from China received 62.3%. The workload is a fraction of a fully sampled defence, but it determines whether you face a middle-tier rate or the worst one available.
How are surrogate values calculated, and can our own costs be used?
US and EU authorities normally reconstruct normal value from a surrogate or analogue country rather than accepting Chinese costs and domestic prices. If electricity, labour or a key input is expensive in that country, your duty rises accordingly. Your own purchase price is used only where you can show a specific input was bought at market prices under market conditions, supported by contracts, proof of payment, source-country price data and public market quotations. Effort is better spent on transactional accuracy and on the evidence for your largest input than on a market economy status claim.
Is a price undertaking cheaper than paying the duty?
It depends on scale and customer structure. An undertaking takes the form of a minimum import price mapped to a model list, plus quarterly reporting, disclosure of affiliates and acceptance of unannounced verification. It converts a one-off defence cost into permanent compliance cost. The EU-China solar undertaking ran for five years and was terminated in 2018, with companies later pursued for violations. Producers with many models, small orders and scattered customers rarely gain. In a US case undertakings are not the practice at all; the practical room is in scope exclusions, where removing a specific model is often worth more than arguing about the margin.