TIMKEN U.S. CORPORATION and Timken Nadellager, GmbH, Plaintiffs-Appellants, v. UNITED STATES, Defendant-Appellee, 434 F.3d 1345

MICHEL, Chief Judge. Timken U.S. Corporation and Timken Nadellager, GmbH (collectively “Timken”) appeal the United States Court of. International Trade’s decision upholding the United States Department of Commerce’s antidumping duty determination and Commerce’s refusal to correct certain alleged errors associated with Timken’s home market sales that were relied upon by Commerce in computing the dumping margin for Timken’s cylindrical roller bearings exported from Germany to the United States from May 1, 1998 to April 30, 1999. Timken v. United States, No. 00-09-00454, 2004 WL 2423732 (Ct.

Int’l Trade Oct. 29, 2004). Because the Court of International Trade did not err in remanding the case to Commerce for investigation of Timken’s evidence supporting correction and because Commerce’s anti-dumping duty determination and its refusal to correct the alleged errors on remand were supported by substantial evidence and were in accordance with law, we affirm. I. BACKGROUND Like many appeals from the Court of International Trade, this appeal has a somewhat complex history. In 1999, pursuant to

Section 751 of the Tariff Act of 1930, Commerce conducted its tenth administrative review of the Antidumping Duty Orders on cylindrical roller bearings (“bearings”) from, inter alia, Germany for the period of May 1,1998 to April 30,1999. Timken imported bearings made in Germany into the United States during that period of time and thus its importations were subject to Commerce’s review. Commerce provided Timken with a questionnaire and requested Timken to identify the channels of distribution for its home market sales.

Timken identified five channels: (1) sales of bearings produced by Timken and sold by Timken from its factory to large original equipment manufacturers (“OEMs”); (2) sales of bearings produced by Timken and sold by Timken from its factory- to small OEMs; (3) sales of bearings produced by Timken and sold by Timken from its factory to distributors; (4) resales of bearings made by Timken’s affiliated marketing entity to OEMs; and (5) resales of bearings made by Timken’s affiliated marketing entity to distributors.

Timken also described the selling and marketing processes for each of the five channels of distribution. After receiving Timken’s response, Commerce verified Timken’s categorization of its sales.

Based upon the verification information, Commerce examined- Timken’s selling activities, the point in the channel of distribution at which the selling activities occurred, and the types of customers that purchased the subject bearings and eventually determined that three-of the channels of distribution identified by Timken were for home market sales by Timken itself and the remaining two were for “resales” by Timken’s affiliated marketing entity.

As a result, Commerce re-designated channel 1 as home market 1 (“HM1”), grouped channels 4 and 5 together and re-designated them as home market 2 (“HM2”), and grouped channels 2 and 3 and re-designated them as home market 3 (“HM3”). In grouping channel 2 with 3 and channel 4 with 5, Commerce essentially determined that the points in which selling activities occurred in these channels were indistinguishable. Commerce then computed a dumping margin of 61.60 percent and published its preliminary results.

See Preliminary Results of Administrative Review, Antifriction Bearings (Other than Tapered Roller Bearings) and Parts Thereof from France, Germany, Italy, Japan, Romania, Singapore, Sweden and the United Kingdom, 65 Fed.Reg. 18033 (Apr. 6, 2000). After reviewing Commerce’s results, Timken alleged that it had made certain errors, which it characterized as “clerical,” in reporting its sales. Specifically, Timken claimed that it “inadvertently and inaccurately” reported seventeen transactions in channel 1 instead of channel 2 or 3.

This mistake, Timken claimed, caused those transactions to be categorized into HM1 instead of HM3, which in turn resulted in an inaccurate dumping margin. Timken contends that the errors occurred because it relied on the customer’s name alone in categorizing the sales, rather than on whether the customer used the bearings in large-sized products or small-sized products.

Timken divided these errors into three groups: (1) sales of bearings shipped to several large OEMs for use as replacement parts (“Replacement Parts Sales”); (2) sales of bearings shipped to a division of a large OEM for use in small electric tools (“Small Electric Tools Sales”); and (3) sales of bearings shipped to a prototype center of a large OEM for use in prototypes (“Prototype Sales”). Timken requested Commerce to correct these errors before issuing the final results.

In support of its request, Timken submitted a “case brief,” which included exhibits consisting of invoices and corresponding purchaser orders for the seventeen allegedly miscategorized sales. Commerce applied the test highlighted in Certain Fresh Cut Flowers from Colombia; Final Results of Antidumping Duty Administrative Review, 61 Fed.Reg. 42833 (Aug. 19, 1996), in deciding whether to substitute Timken’s new information and recalculate the applicable dumping margin.

The first requirement of the so-called “Colombian Flowers Test” is that the error be of a “clerical” nature, not a “methodological error, a substantive error, or an error in judgment.” Commerce found that Timken’s errors were not “clerical” but instead were due to “error[s] in judgment.” The second requirement of the “Colombia Flowers Test” is that Commerce must be satisfied that the new evidence provided in support of the “clerical” error is reliable.

Commerce found that the new information that Timken submitted in its case brief conflicted with information already of record and that the corrections were not offered at the earliest opportunity, thus allegedly preventing Commerce from ascertaining their reliability through the verification process. For these reasons, Commerce declined to correct the asserted errors and issued its final results without any changes.

See Final Results of Administrative Review, Antifriction Bearings (Other than Tapered Roller Bearings) and Parts Thereof from France, German, Italy, Japan, Romania, Singapore, Sweden and the United Kingdom, 65 Fed.Reg. 49,-219 (Aug. 11, 2000). Timken appealed to the Court of International Trade. See Timken U.S. Corp. v. United States, 318 F.Supp.2d 1271 (Ct. Int’l Trade 2004) (“Timken I ”).

The Court of International Trade rejected Commerce’s rigid application of the “Colombia Flowers Test,” even though it agreed with Commerce that Timken’s error was not “clerical,” because the Court of International Trade was concerned that Commerce’s application of the test would render a grossly erroneous dumping margin in this case. Id. at 1277. The Court of International Trade also noted that it was unclear what evidence of record Commerce thought contradicted Timken’s new evidence and that any corrective information would necessarily be in conflict with information originally submitted to Commerce.

Id. at 1279. Accordingly, the Court of International Trade remanded the case to Commerce for further investigation. On remand, Commerce noted its disagreement with the Court of International Trade’s order at the outset. Remand Determination, slip op. at 4. Because Commerce posited that the errors in dispute were not “clerical,” as determined by the Court of International Trade, these errors did not require correction and there was no need for any further investigation.

Despite this position, Commerce of course followed the Court of International Trade’s order and reviewed Timken’s new evidence, ultimately determining that corrections were not warranted because “the information [Timken] submitted does not substantiate the existence of the alleged error, and, thus, reclassifying [Timken’s] sales in a different level of trade would not result in a more accurate dumping margin.” Id., slip op. at 8. In reaching that conclusion, Commerce specifically evaluated the new information that Timken submitted for the three groups of allegedly miscategorized sales.

Starting with the first group, Commerce found that Timken’s argument that the Replacement Part Sales should have been categorized under channel 3 conflicted with the information of record. Commerce noted that Timken described the bearings it sold under channel 3 in its verified questionnaire as those listed in its “standard catalog,” not replacement parts. Id., slip op. at 16. It consequently found that Timken failed to provide a sufficient explanation of why replacement parts should be classified under channel 3.

Commerce also observed that the customers of the Replacement Part Sales were large OEMs, not distributors — the customers Timken originally identified for channel 3 sales. Id., slip op. at 16-17. Lastly, Commerce noted that Timken did not provide evidence showing the bearings were, in fact, used as replacement parts. “The merchandise,” Commerce reasoned, “very well may have been used for the normal production activities of the large OEMs.” Id., slip op. at 17.

Turning to the second group, Commerce found Timken did not provide factual information to substantiate the claim that the customer of the Small Electric Tools Sales, a division of a large OEM, should independently be considered a small OEM under channel 2 instead of a large OEM under channel 1. Id., slip op. at 14-15. As to the third group, Commerce found that Timken’s argument that the Prototype Sales should have been categorized under either channel 2 or 3 conflicted with the information of record.

Significantly, it noted that when Timken described the selling function of channel 1 in its verified questionnaire, it specifically mentioned prototype sales, but did not do so for either channel 2 or 3. Id., slip op. at 12-13. Next, Commerce noted that the customer of the Prototype Sales was a large OEM. “That fact alone,” Commerce reasoned, “supports the original channel 1 classification.” Id., slip op. at 13.

It further noted that only two of the invoices submitted by Timken for the Prototype Sales contained the word “Muster,” the German equivalent of “sample,” thus suggesting that the remaining Prototype Sales did not actually involve prototypes and were instead regular sales to a large OEM. Id. Finally, Commerce noted that Timken did not provide any evidence showing that the alleged prototypes were not used in producing original equipment. Id., slip op. at 14.

Timken again appealed to the Court of International Trade, which upheld Commerce’s remand determination not to make the requested corrections on the grounds that Timken’s corrective information was not sufficiently supported whereas Commerce had verified Timken’s original information. Timken U.S. Corp. v. United States, No. 00-09-00454 (Ct. Int’l Trade Oct. 29, 2004) (“Timken II ”). The Court of International Trade observed that Commerce met its burden of reviewing the disputed classifications and the corrective information submitted by Timken.

It thus concluded that Commerce provided a reasonable explanation for its determination that the dumping margin applicable to Timken’s bearing imports was as accurate as possible and that there was no need to make any corrections. Id., slip op. at 24. . Timken timely appeals the Court of International Trade’s decision upholding Commerce’s remand determination. We have jurisdiction over this appeal pursuant to 28 U.S.C. § 1295(a)(5). II.

DISCUSSION We review a decision of the Court of International Trade.in an antidumping duty case de novo, applying the same standard of review to Commerce’s determination as the Court of International Trade applies. See Co-Steel Raritan, Inc. v. Int’l Trade Comm’n, 357 F.3d 1294, 1309 (Fed.Cir.2004).

Under that standard, we are obligated to affirm the Court of International Trade’s decision unless Commerce’s determination is “unsupported by substantial evidence on the record, or otherwise not in accordance with law.” 19 U.S.C. § 1516a(b)(1)(B)(i) (2000). “Substantial evidence” means “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Consol. Edison v. Nat’l Labor Relations Bd., 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938).

Notably, “the possibility of drawing two inconsistent conclusions from the evidence does not prevent an administrative agency’s finding from being supported by substantial evidence.” Consolo v. Fed. Mar. Comm’n, 383 U.S. 607, 620, 86 S.Ct. 1018, 16 L.Ed.2d 131 (2001). Timken argues that Commerce’s decision not to reclassify its allegedly miscate-gorized sales is not supported by substantial evidence.

Timken asserts that the new information it submitted in its case brief at the preliminary results stage adequately supports the reclassification of its seventeen allegedly miscategorized sales from channel 1 to either channel 2 or 3. More specifically, for its Replacement Part Sales, that the customer was a large OEM, Timken contends, does not automatically mean that all sales were for bearings that would be incorporated into other goods; a large OEM may also purchase bearings to serve as replacement parts.

In such case, Timken points out that the nature of sales and the associated selling activities are by definition different. It thus argues that the market in which its bearings were sold governs the classification, not the identity of the customer, contrary to Commerce’s determination. For its Small Electric Tools Sales, Timken claims that the proper classification is ascertainable by resort to its descriptions of channel 1 and channel 2.

Timken points out that it described channel 1 as involving “sales to automotive and other large OEMs” and channel 2 as involving “sales to smaller OEMs such as machinery and other equipment producers.” Based upon these descriptions and the fact that the customer sells small electric tools, Timken asserts that the Small Electric Tools Sales necessarily fall under channel 2; that the customer is a division of a large OEM is irrelevant because the only point of distinction between channel 1 and channel 2 is the size of the goods sold.

For its Prototype Sales, Timken argues as it did for the Replacement Part Sales that customer identity does not control classification; rather, the market in which the product was sold dictates whether a particular sale should be categorized in channel 1, 2, or 3. Hence, that the customer of the Prototype Sales was a large OEM is, Timken claims, of no import in deciding whether the Prototype Sales were erroneously categorized in channel 1.

Further, it asserts the fact that some invoices contained the word “Muster,” while others did not, is not decisive since all Prototype Sales were shipped to the customer’s prototype center, as shown on all the invoices. Finally, Timken contends that the quantity of bearings involved in the allegedly miscategorized sales confirms that they were erroneously included in channel 1. Timken points out that each sale consisted of several hundreds of bearings; sales placed in channel 1, by contrast, involve several thousands of bearings.

As such, Timken claims quantity alone demonstrates that its bearing sales were erroneously categorized in channel 1. The government responds by arguing that Commerce’s decision not to reclassify Timken’s allegedly miscategorized sales is supported by substantial evidence. Beginning with the Replacement Parts Sales, the government asserts that Timken neither mentioned replacement parts in its descriptions of its channels of distribution nor offered an explanation for why such sales are correctly classified in channel 3 as opposed to channel 1.

Turning to Small Electric Tools Sales, the government points out that Commerce correctly determined that Timken provided no evidence to substantiate its claim that the division of a large OEM, which manufactures small electric tools, should be considered a small OEM. Moreover, the government contends that Commerce correctly noted that while sales to a single customer may be classified in different channels of distribution, Timken failed to demonstrate facts sufficient to show that such bifurcation is necessary for the Small Electric Tools Sales.

Finally, the government argues that if Commerce were to reclassify all of the Prototype Sales on the basis of only two invoices, it essentially would place unverified information over conflicting, verified information. Even assuming Timken is correct that the sales were truly for prototypes, the government claims that such sales were properly included in channel 1 because Timken mentioned prototype sales only in its description of channel 1, not in its descriptions of channel 2 and 3.

As an alternative ground of affirmance, the government asserts that the Court of International Trade erred in Timken I by remanding the case to Commerce for investigation of Timken’s new information. Because the Court of International Trade had agreed with Commerce that Timken’s alleged errors were not “clerical,” the government contends that the Court of International Trade should have simply affirmed Commerce’s original refusal pursuant to the “Colombia Flowers Test.” Accordingly, the government argues that Commerce’s original antidumping duty determination and first refusal to correct should stand.

Timken replies to the government’s alternative argument by disputing both Commerce’s and the Court of International Trade’s characterization of its errors as “errors in judgment.” Timken claims that its errors are “clerical” and thus subject to correction under the “Colombia Flowers Test.” A. Before addressing Timken’s argument concerning its new evidence, we focus on the government’s challenge to the Court of International Trade’s remand order in Timken I since this issue potentially is dispositive of the appeal without requiring minute examination of the old and new information on various sales.

On the merits, we disagree with the government. Essentially, it contends that the Court of International Trade erred in remanding the case for further investigation solely because it had concurred with Commerce’s determination that Timken’s errors were not “clerical” in nature. , In making this argument, the government seems to advocate a bright-line rule regarding what kinds of errors may be corrected in the context of antidumping duty determinations.

Specifically, it' argues that once Commerce calculates a dumping margin, an error in the underlying data supplied by an importer to Commerce may be corrected only where it is “clerical” in nature. Surprisingly, counsel for Timken appeared to agree with the government’s contention at oral argument. The government does not, however, identify any statute or regulation on point. Instead,.it relies solely on our decision in Alloy Piping Products, Inc. v. United States, 334 F.3d 1284 (Fed.Cir.2003), and its own determination in Certain Fresh Cut Flowers from Colombia, 61 Fed.Reg. at 42834.

Such reliance is, however, misplaced. In Alloy Products, after receiving the final results of Commerce’s antidumping duty determination, the importer sought correction of what it described as a “clerical” error made when it provided its original sales information to Commerce. The importer argued that Commerce was absolutely and expressly required to make the correction pursuant to 19 U.S.C. § 1673d(

e) and 19 C.F.R. § 351.224. Commerce declined, and both the Court of International Trade and this court affirmed. We recognized that Commerce was not required to correct the “clerical” errors made by the importer because the statute and regulation cited by the importer apply only to “clerical” errors committed by Commerce itself. 334 F.3d at 191. Contrary to the government’s implication, however, we did not broadly bar Commerce from correcting “clerical” errors or, for that matter, methodological errors, errors in judgment, or substantive errors made by an importer.

Instead, our decision in Alloy Products turned on the fact that the importer sought correction of errors at the final result stage, not at the preliminary result stage as in this case. For this reason, the facts in Alloy Products are readily distinguishable from those implicated here. Separately, a determination by Commerce, like Certain Fresh Cut Flowers from Colombia, is not binding on this court. We also disagree with the restrictiveness of the “Colombia Flowers Test.” That test, which Commerce states is based on our decision in NTN Bearing Corp. v.

United States, 74 F.3d 1204 (Fed.Cir.1995), imposes six conditions which must be met before Commerce will correct an error made by an importer. The first condition is that the “error in question must be demonstrated to be a clerical error, not a methodological error, an error in judgment, or a substantive error.” Certain Fresh Cut Flowers from Colombia, 61 Fed.Reg. at 42834. It is perplexing how Commerce could have derived this first condition from NTN Bearing when our holding in that case, as in Alloy Products, was premised exclusively on timeliness.

Indeed, upon careful review, we do not even find any dicta suggesting a distinction between “clerical” and other types of errors, and Commerce cited none. More specifically, after receiving the International Trade Administration’s (“ITA”) preliminary antidumping duty determination, NTN realized that it made two “clerical” errors caused by typographical mistakes. These clerical errors, NTN asserted, caused a substantial increase in its dumping margins. NTN offered corrective information to the ITA and sought to have corrections made before the ITA issued its final determination.

The ITA refused because NTN made its request after the deadline for submitting new information had expired and because its new evidence did not adequately demonstrate that errors had occurred. Notably, the ITA did not challenge NTN’s characterization of its errors as “clerical.” The Court of International Trade affirmed that refusal. We reversed, holding that the ITA abused its discretion in refusing to consider NTN’s corrective information because of untimeliness. NTN Bearing, 74 F.3d at 1208.

At no point in our analysis did we distinguish “clerical” errors from other errors such as methodological errors, substantive errors, or errors in judgment. We simply reasoned: NTN responded in a timely manner to the preliminary determination, pointing out ITA’s clerical errors, as well as its own. ITA corrected its own but not NTN’s errors. Correction of NTN’s errors would neither have required beginning anew nor have delayed making the final determination. A straightforward mathematical adjustment was all that was required.

Failure to make it resulted in the imposition of many millions of dollars in duties not justified under the statute. Id. We also noted that while a tension between finality and correct result may arise in some instances, such was not the case in NTN Bearing because NTN requested corrections to be made at the preliminary results stage. Id. “[Pjreliminary determinations,” we observed, “are ‘preliminary’ precisely because they are subject to change.” Id.

To the extent that Commerce cites NTN Bearing for the source of its distinction between “clerical” and other errors, we surmise that Commerce erroneously extrapolated one where none was stated or intended. We emphasize that our ruling in NTN Bearing was based solely on timeliness, not on the type of error.

Additionally, we hold that Commerce is free to correct any type of importer error — clerical, methodology, substantive, or one in judgment — in the context of making an antidumping duty determination, provided that the importer seeks correction before Commerce issues its final results and adequately proves the need for the requested corrections. In light of this holding, we need not address Timken’s argument that its errors were “clerical” rather than “errors in judgment” as found by Commerce and affirmed by the Court of International Trade.

Finally, the convenience of the bright-line rule that Commerce advocates is troubling. We suspect that it is not uncommon for an importer to make errors in reporting its sales information to Commerce pursuant to an administrative review. We also suspect that most such errors are methodological, substantive or, as in this case, errors of judgment, rather than “clerical” errors.

Oddly, it nevertheless appears that Commerce has not issued any regulation addressing whether an importer can correct errors in the information it has submitted, let alone restricted the types of importer errors that are eligible for such correction. By urging this court to embrace Commerce’s determination in Certain Fresh Cut Flowers from Colombia limiting correction to just “clerical” errors, the government seemingly aims to save itself from having to evaluate corrective information in most cases, regardless whether correction is sought at the preliminary results stage or the final results stage.

This court, however, has never discouraged the correction of errors at the preliminary result stage; we have only balanced the desire for accuracy in antidumping duty determinations with the need for finality at the final results stage. We thus have no choice but to conclude that Commerce’s position here amounts to ipse dixit, which is a very weak basis for any rule.

Accordingly, because Timken sought correction of its errors after Commerce issued the preliminary results, but before it issued the final results, we conclude that the Court of International Trade in Timken I did not err in remanding the case to Commerce for an analysis of Timken’s new evidence. B. Turning to the heart -of Timken’s appeal, we disagree with Timken that the new evidence it submitted in its case brief, which consisted of invoices and, in most instances, corresponding purchase orders, is strong enough to require reclassification.

At the outset, we note that the government contends in general that Commerce would be forced to prefer unverified information over conflicting, verified information if it were to make Timken’s requested corrections. On appeal, the government particularly raises this argument with respect to Timken’s Prototype Sales. We are struck by the fallacy of such a position; Commerce declines to verify the new information and then rejects it because it is unverified. If Commerce had reason to doubt Timken’s corrective information, then it could have, and perhaps should, have, performed a second verification.

Indeed, the government, we note, does not cite any statute or regulation barring such a second verification. We now focus on Timken’s allegedly miscategorized sales, beginning with the Replacement Part Sales for which there were eight total sales and three different customers, ie., three sales to the first customer, one to the second customer, and four to the third customer. For sake of clarity, we consider the Replacement Part Sales for each of the three customers separately.

For the first customer, one of the purchase orders contained a handwritten notation in German, which Timken contends translates as: “The parts (from the purchase order) are exclusively destined for repair-purposes ... The total annual demand for this purpose are 26 units.” Timken explained that the notation applied to the other two purchase orders as well and was added by Mrs. Kleiner, an employee of Timken. According to Timken, Mrs. Kleiner contacted the customer in preparation for the case brief to confirm that the purchased bearings were not used in producing original equipment.

Timken does not, however, present any evidence to substantiate her communication with the customer. Commerce was thus in the position of having to accept her notation, knowing full well that it was produced specifically for the purpose of proving the alleged miscategorization. Viewed in this light, we hesitate to fault Commerce for rejecting this evidence over evidence for these particular sales which it previously verified in conjunction with Timken’s original submission.

For the second customer, the invoice contained a handwritten German word, which was not easily readable, followed by an equal sign and the words “Replacement Parts.” That invoice, along with the corresponding purchase order, also contained a “649” code followed by the words “shipping point.” A code sheet provided with the two documents listed “central location for replacement parts” for that code. It is, nevertheless, entirely unclear who supplied the handwritten information or when it was added to the invoice.

As with the aforementioned Replacement Part Sales, the notation quite possibly was added after Timken decided to argue that it erred in categorizing the sale. Additionally, Timken did not present any evidence about the sale showing that the bearings were, in fact, used as spare parts and were not incorporated into original equipment produced by the customer. Consequently, we conclude that Commerce reasonably rejected this evidence over previously submitted evidence for this customer.

For the third customer, the purchase orders all contained the typed word “er-satzteile,” followed by an equal sign and the handwritten words “spare parts.” While we do not doubt that German word “ersatzteile” correctly translates as “spare parts” or the equivalent thereof, Timken again did not provide any evidence to substantiate its claim that the bearings actually were used by the customer as spare parts.

Commerce particularly noted the absence of such evidence in its determination: “The merchandise very well may have been used for the normal production activities of the large OEMs.” Remand Determination, slip op. at 17. As a final consideration applicable to the eight Replacement Part Sales in general, Timken did not offer a persuasive explanation of why, if the sales were really for spare parts, they should be categorized under channel 3 as opposed to channel 1.

Indeed, in defining its channels of distribution in response to Commerce’s original questionnaire, Timken did not mention replacement part sales in either channel 1 or channel 3. It consequently has no real basis for reclassifying the Replacement Part Sales in channel 3 as opposed to channel 1 or, for that matter, channel 2. Small OEMs under channel 2 are, after all, just as likely to purchase replacement parts as large OEMs under channel 1 or distributors under channel 3.

Therefore, since any customer may require and purchase replacement parts, we conclude that customer identity is not a conclusive way for categorizing Timken’s sales, but serves as a sound starting point. Here, the parties do not dispute that the three customers who purchased the Replacement Part Sales were all “large OEMs” classified under channel 1. As a result, we conclude that the evidence supporting Commerce’s original determination is substantial and that it was justified in refusing to reclassify the Replacement Part Sales from channel 1 to channel 3.

As to the Small Electric Tools Sales, there were four distinct sales, all to a single customer. One of the four invoices contained a handwritten notation showing an asterisk beside the customer name followed by the comment “product division electric tools.” Because Timken’s argument as to these sales substantially depends on whether the words “large” and “small” modify the size of the producer or the size of the products sold, we are compelled to reach that distinction in this context.

If we agree with Timken that the words “large” and “small” modify the size of the goods produced, then Timken is necessarily correct that its Small Electric Tools Sales were miscategorized under channel 1 when they should have been categorized under channel 2. We do not, however, accept Timken’s position. Although Timken is in a better position than Commerce to understand and define its various channels of distribution and customers, we nevertheless observe that Timken’s argument is not supported by its own descriptions of its channels of distribution.

If anything, Timken’s descriptions favor Commerce’s position that the words “large” and “small” modify the size of the producers.

Throughout its descriptions for channel 1 and channel 2, Timken consistently placed the adjectives “large” and “small,” respectively, before the noun “OEMs.” Moreover, Timken did not clearly describe the goods produced by the OEMs under channel 1 or 2, let alone define those goods as being “large” or “small.” Commerce’s analysts, like the average reader, would thus grammatically interpret the adjectives “large” and “small” to modify the noun “OEMs.” It is entirely counterintuitive, based solely upon the syntax, to read the adjectives as modifying the products manufactured by the OEMs.

We, consequently, must reject Timken’s contention that a “large” or “small” OEM means an OEM that produces “large” or “small” products. As such, Timken’s argument falls apart. Hence, we conclude that Commerce’s refusal to reclassify the Small Electric Tools Sales from channel 1 to channel 2 is supported by substantial evidence. With regard to the Prototype Sales, Timken made five separate sales to a single customer.

Two of the invoices associated with those sales contain the German word “Muster.” The other three invoices, although missing the word “Muster,” were sent to the same location as the other two sales, presumably the customer’s prototype testing facility since the shipping address for each sale contained the word “Prototypen.” Timken’s argument that these sales should be classified under channel 2 or 3 based upon the “Muster” notation and/or shipping address, however, seems contradicted by Timken’s own description of its channels of distribution.

As Commerce noted, when discussing sales under channel 1, Timken explained: [Timken] will develop and supply prototypes (“Muster”) which are specially produced for the customer to test, and if the customer approves these units, the parties estimate the quantities to be bought and negotiate a fixed selling price. Thereafter, [Timken] provides further sample units (designated by the suffix “M41”), which are produced under normal production conditions, and if the customer approves these samples, [Timken] is fully approved to supply.

Timken did not make a similar disclosure for either channel 2 or 3, thereby suggesting that it consciously did not contemplate selling prototypes under channel 2 or 3. What is more, the customer is an “automobile and other large OEMs,” the very customer defined for channel 1 sales. Thus, this evidence, especially taken together, demonstrates that Commerce was reasonable to refuse to reclassify the Prototype Sales from channel 1 to channel 2 or 3.

Lastly, Timken’s argument that the small quantity of bearings involved in each of the seventeen disputed transactions weighs in favor of reclassifying the sales under either channel 2 or 3 is unavailing. None of the channels of distribution, as defined by Timken, contain any numeric limitations. Despite this, Timken claims that the average quantity for a channel 1 sale exceeds 14,000 units, but fails to substantiate its claim with any evidence. Furthermore, we presume that the sale of prototypes likely would only include a few units.

The customer, after all, would want to confirm that the prototypes function for them intended usage before buying commercial quantities. Thus, that the seventeen allegedly miscategorized sales involved a small number of bearings compared to the number typically involved in a commercial sale under channel 1 does not alone require reclassifying the Prototype Sales from channel 1 to channel 2 or 3. Nor does it when taken in view of Timken’s other new information.

In sum, we hold that Timken’s case before Commerce basically involves a failure of proof; none of the new evidence that Timken submitted adequately supports its position that it initially misclassified seventeen sales. Commerce calculated the dumping margin applicable to the bearings Timken imported in the United States during the review period based upon Timken’s original information, which it verified. Although this is a close case, we hold that Commerce’s original determination was, on the balance of the old and new evidence, supported by substantial evidence and was not contrary to law.

The Court of International Trade, therefore, properly affirmed Commerce’s antidumping duty determination based on the original information and its refusal to accept Timken’s corrective information. III. CONCLUSION We conclude that the Court of International Trade in Timken I correctly remanded the case to Commerce for investigation of Timken’s new evidence because Timken sought correction after Commerce issued the preliminary results, but before it issued the final results and such corrections are not limited to “clerical” errors only.

Because we conclude that Commerce had sufficient factual basis for refusing to make the corrections on remand and because its original duty determination was supported by substantial evidence and in accordance with law, we affirm the Court of International Trade’s decision in Timken II upholding Commerce’s rejection of Timken’s corrective information and its antidumping duty determination based on Timken’s original evidence. AFFIRMED .

Commerce and Timken interpret the definition of "large OEMs” under channel 1 and "small OEMs” under channel 2 differently based upon whether the words “large” and "small" refer to the size of the OEM or the size of the product produced by the OEM. Commerce takes the former position, while Timken takes the latter.

Indeed, in Commerce’s own words: It is important to recognize that the term "large original equipment manufacturer,” which connotes the level of trade in which such sales were grouped in the final results of review, does not refer to producers of “large original equipment.” Rather it refers to large producers of original equipment. That is, the term "large” modifies the type of manufacturer, not the type of product. Final Results of Redetermination Pursuant to Court Remand, Timken U.S. Corporation and Timken Nadellager, GmbFt v.

United States, No. 00-09-00454, slip op. at 7 (June 7, 2004) ("Remand Determination ”). Timken does not so clearly explain its view. However, we are able to infer its position based upon the arguments it made before Commerce, the Court of International Trade, and this court on appeal. .

Section 1673d(

e) states in pertinent part: “The administering authority shall establish procedures for the correction of ministerial errors in final determinations within a reasonable time after the determinations are issued under this section.”

Section 351.224(

e) provides in relevant part: "The Secretary will analyze any comments received and, if appropriate, correct any significant ministerial error by amending the preliminary determination, or correct any ministerial error by amending the final determination or the final results of review (whichever is applicable).” . We note that the parties agreed to keep the identities of Timken's customers, among other things, confidential under the terms of a protective order. We refer only to the customers generally, because of the protective order and because their identities are not critical to the disposition of this appeal anyway.