SKF USA, Inc., SKF France S.A., SKF Aerospace France S.A.S., SKF GmbH, and SKF Industrie S.p.A. (collectively “SKF”) appeal from the final judgment of the Court of International Trade affirming the U.S. Department of Commerce’s (“Commerce’s”) revision of its model-match methodology and its ongoing zeroing methodology in calculating antidumping margins for ball bearings and related parts. SKF USA Inc., v. United States, 491 F.Supp.2d 1354 (Ct. Int’l Trade 2007). Because the Court of International Trade correctly determined that Commerce’s zeroing practice and the revision of its model-match methodology were supported by substantial evidence and were in accordance with law, we affirm.

I. BACKGROUND

SKF imports ball bearings and related parts into the United States, which have been the subject of an antidumping investigation and subsequent annual reviews by Commerce. In performing these reviews, Commerce compares the imported products with “foreign like product” as required by 19 U.S.C. § 1677b(a)(l)(B). “Foreign like product” is defined as either identical merchandise, § 1677(16)(A), or similar merchandise, § 1677(16)(B) and (C). Determinations of similar (i.e., nonidentical) merchandise are made using a model-match methodology . developed by Commerce.

Beginning with the first administrative review, Commerce utilized a “family model-match methodology,” whereby it grouped ball bearings into “families” based on exact matches of eight characteristics: 1) bearing design, 2) load direction, 3) number of rows of rolling elements, 4) precision rating, 5) dynamic load rating, 6) outside diameter, 7) inside diameter, and 8) widtb/height. Bearing models within a family were treated as equally similar to all other models sharing these eight characteristics, regardless of other product variations. When Commerce formulated the family model-match methodology in 1990, it grouped specific models of bearings into families “[t]o minimize the necessity for comparisons among an exceptional[*1376] ly large number of bearing models,” and “to limit the need for adjustments for physical differences in merchandise and the need for model matching.” J.A. at 129.

During the fourteenth administrative review, however, in a memorandum dated December 3, 2003, Commerce indicated that it was considering modifications to the model-match methodology. It explained that “[t]he family-based approach ... deviates from [Commerce’s] normal practice in that [Commerce] ha[d] considered all models within a family to be equally similar,” and that this methodology “was developed at a time ... when it was impossible for [Commerce] to identify the single most similar foreign like product” because “the technological resources available ... were far less powerful than they are now.” Id. at 359. Thus, “a more complex model-match methodology ... would have been prohibitively expensive and time-consuming.” Id. Commerce suggested changing the model-match methodology “in order to determine the single most similar comparison-market model, taking advantage of intervening technological developments,” id. at 361, but explained that it would not attempt to implement any changes to the methodology in the ongoing 2002-2003 reviews, choosing instead to “solicit comments and invite rebuttal comments” before implementing the new methodology in the 2003-2004 reviews. Id. at 364.

On July 7, 2004, Commerce issued a memorandum outlining the revised model-match methodology and indicating that although it intended to use the new methodology, it reserved the right to make further modifications based on comments received or data collected during the course of the 2003-2004 reviews. Commerce summarized its proposed methodology as follows:

We recommend that we select the single most similar comparison-market model among only those models that are in the same family, as redefined above (ie., bearing design, load direction, number of rows, and, if we determine it appropriate, precision grade), on the basis of that model that is closest in terms of inner diameter, outer diameter, width, load rating, and, if appropriate, precision grade. We recommend further that we resolve ties by selecting the single home-market model whose variable cost of manufacturing is closest to that of the U.S. model.

Id. at 439. Thus, under the new methodology, Commerce would identify models in the comparison market and the U.S. market that share four of the physical characteristics of the old methodology. After finding an identical match of these characteristics, Commerce would then identify a “most similar” model based upon a comparison of the remaining characteristics.

On May 13, 2005, Commerce issued its preliminary results for the fifteenth administrative review using the new methodology. Antifriction Bearings and Parts Thereof from France, Germany, Italy, Japan, Singapore, and the United Kingdom: Preliminary Results and Partial Rescission of Antidumping Duty Administrative Reviews, 70 Fed.Reg. 25,538 (Dep’t of Commerce May 13, 2005). Commerce also released a memorandum providing the revised model-match methodology and addressing comments received. In this memorandum, Commerce explained that “[b]ecause there is a statutory preference for using price-to-price comparisons and the technological constraints that led us to adopt the simplified family methodology are no longer operative, we find compelling reasons to change the model-matching methodology....” J.A. at 482. Under the revised methodology, Commerce “identifies] the single most similar model within the [redefined] family of bearings using a sum-of-the-deviations model-matching[*1377] methodology with a 40-percent cap on the sum of the deviations, resolving ties among bearings with the same sum of the deviations using the smallest [difference-in-merchandise] adjustment.” Id. at 492. Commerce indicated that it would “consider all arguments presented in case and rebuttal briefs and, if appropriate, revise this methodology for the final results of reviews.” Id. at 478.

On September 16, 2005, Commerce published the final results of its fifteenth anti-dumping administrative review of ball bearings and parts thereof from various countries imported from May 1, 2003 to April 30, 2004. Ball Bearings and Parts Thereof From France, Germany, Italy, Japan, Singapore and the United Kingdom: Final Results of Antidumping Duty Administrative Reviews, 70 Fed.Reg. 54,-711 (Dep’t of Commerce Sep. 16, 2005) {“Final Results ”). In a related memorandum, Commerce indicated that “[t]his new model-matching methodology is much closer to [its] normal matching practice than is the family-matching methodology in that it allows [Commerce] to select the single most-similar model and allows [Commerce] to avoid rejecting reasonable price-to-price comparisons between models with slightly different physical characteristics.” Memorandum from Barbara E. Tillman, Acting Deputy Assistant Sec’y for Import Admin., U.S. Dep’t of Commerce, to Ronald K. Lorentzen, Acting Assistant Sec’y for Import Admin., U.S. Dep’t of Commerce, Issues and Decision Memorandum for the Antidumping Duty Administrative Reviews of Ball Bearings and Parts Thereof from France, Germany etc. for the Period of Review May 1, 2003, through April 30, 2004, at 19 (Sept. 12, 2005) (“Issues & Decision Memo ”). Commerce explained that it had “solicited extensive comments” and “incorporated numerous suggestions” before adopting the new methodology. Id. It concluded that “compelling reasons exist to change the model-match methodology,” that it now possessed “the technological capacity to use a more accurate methodology,” and that “[t]he new methodology is substantially more accurate than the family-matching methodology.” Id. at 20.

In calculating SKF’s weighted-average dumping margins, Commerce also employed its usual methodology in which dumped sales are not offset by non-dumped sales — a practice known as zeroing. SKF challenged these results before the Court of International Trade, which sustained Commerce’s Final Results. SKF timely appealed. We have jurisdiction pursuant to 28 U.S.C. § 1295(a)(5).

II. DISCUSSION

A. Standard of Review

We review the Court of International Trade’s decisions regarding Commerce’s antidumping determinations de novo, applying the same standard of review to Commerce’s determinations as did that court. Carpenter Tech. Corp. v. United States, 510 F.3d 1370, 1372 (Fed.Cir.2007) (citations omitted). Thus, Commerce’s antidumping determinations must be sustained unless they are “unsupported by substantial evidence on the record, or otherwise not in accordance with law.” Id. at 1373 (quoting 19 U.S.C. § 1516a(b)(l)(B)(i)).

SKF argues that the Court of International Trade erred in reviewing Commerce’s decision to change its model-match methodology for “reasonableness” rather than determining whether the change was supported by substantial evidence and otherwise in accordance with law. See 19 U.S.C. § 1516a(b)(l)(B)(i). SKF further argues that because Commerce’s own policy permits changes to model-match methodologies only when supported by “compelling reasons,” e.g., Final Results of Anti [*1378] dumping Duty Administrative Reviews and Revocation in Part of an Antidumping Duty Order, 58 Fed.Reg. 39,729, 39,765 (Dep’t of Commerce July 26, 1993), we must use a similar standard to review Commerce’s actions. We disagree. The standard articulated by the Court of International Trade, “whether Commerce reasonably determined that there were compelling reasons to revise the model match methodology,” SKF USA, 491 F.Supp.2d at 1361 (citing Hangzhou Spring Washer Co. v. United States, 387 F.Supp.2d 1236, 1246 (Ct. Int’l Trade 2005)), does not conflict with our “substantial evidence” standard of review, which requires only “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Consol. Edison Co. v. Nat’l Labor Relations Bd., 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938).

B. Analysis

1. Commerce’s Revised Model-Match Methodology

SKF disputes the legitimacy of Commerce’s new model-match methodology, the appropriateness of Commerce’s decision to disturb the prior family-match methodology, and the retroactive application of the new methodology to products covered by the fifteenth review. We address each of these arguments in turn.

With respect to the new methodology, the parties disagree about the purpose of a model-match methodology itself. SKF fundamentally argues that the model-match methodology should be designed solely to compare “the most physically and commercially similar products,” regardless of whether such matching consists of price-to-price comparisons or constructed values. Domestic Producer Timken U.S. Corporation (“Timken”) and the government maintain that the new methodology more accurately reflects the intent of 19 U.S.C. § 1677(16)(B) to identify a single most similar model, see Timken Co. v. United States, 630 F.Supp. 1327, 1337 (Ct. Int’l Trade 1986) (holding that “[t]he spirit if not the letter of [the statute] obligates the agency to also ascertain what constitutes the most similar merchandise,” and that “if values can be found only for items of ‘similar’ merchandise, the value of the item most similar to that under appraisement should be adopted”), and that the family-match methodology runs counter to Commerce’s normal practice. SKF responds that the statute does not require the selection of a single most similar model and that the statute does not preclude the appropriate use of constructive value.

The statute provides:

The term “foreign like product” means merchandise in the first of the following categories in respect of which a determination for the purposes of part II of this subtitle can be satisfactorily made:

(A) The subject merchandise and other merchandise which is identical in physical characteristics with, and was produced in the same country by the same person as, that merchandise.

(B) Merchandise—

(i) produced in the same country and by the same person as the subject merchandise,

(ii) like that merchandise in component material or materials and in the purposes for which used, and

(iii) approximately equal in commercial value to that merchandise.

(C) Merchandise—

(i) produced in the same country and by the same person and of the same general class or kind as the subject merchandise,

(ii) like that merchandise in the purposes for which used, and

[*1379] (Hi) which the administering authority determines may reasonably be compared with-that merchandise.

19 U.S.C. § 1677(16). SKF maintains that the hierarchy set forth in the statute does not require Commerce to select a single most similar model and points to prior affirmance of the family-match methodology as evidence of the correctness of its statutory interpretation. See SKF USA Inc. v. United States, 876 F.Supp. 275, 279 (Ct. Int’l Trade 1995) (noting that “the statute does not require Commerce to use a methodology that identifies the greatest number of matches of similar merchandise,” and that the family-match methodology “was within the broad discretion [Commerce] is granted to determine ‘similar merchandise’”). The relevant question, however, is not whether the statute requires such an approach, but rather whether Commerce’s interpretation is permissible.

We have previously observed that this statute “is silent with respect to the methodology that Commerce must use to match a U.S. product with a suitable home-market product.” Koyo Seiko Co. v. United States, 66 F.3d 1204, 1209 (Fed.Cir.1995). Under Chevron, U.S.A., Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984), if a statute is silent or ambiguous with respect to a specific issue, we “must defer to an agency’s reasonable interpretation of a statute even if [we] might have preferred another.” Koyo Seiko Co. v. United States, 36 F.3d 1565, 1570 (Fed.Cir.1994). Furthermore, “[d]eference to an agency’s statutory interpretation is at its peak in the case of a court’s review of Commerce’s interpretation of the anti-dumping laws.” Id. Accordingly, we have previously held that Congress has granted Commerce considerable discretion to fashion the methodology used to determine what constitutes “foreign like product” under the statute. Pesquera Mares Australes Ltda. v. United States, 266 F.3d 1372, 1384 (Fed.Cir.2001) (citing Koyo Seiko, 66 F.3d at 1209).

Commerce’s interpretation of the statute merits deference in this case. SKF alleges that the increased number of price comparisons achieved under the new methodology corresponds with a decrease in the accuracy of the matches themselves. But as noted by the government, SKF fails to identify a single instance of an unreasonable match resulting from the new methodology. Cf. Koyo Seiko Co. v. United States, 516 F.Supp.2d 1323, 1335 (Ct. Int’l Trade 20Ó7) (noting that “[plaintiffs ... each argue through specific examples that the Department’s new methodology led to the comparison of dissimilar bearing products”). And as Timken argues, it is reasonable that although the new methodology allows up to a 40 percent total deviation in dimensions and load rating, the methodology yields more accurate results because it matches the most similar product rather than merely pooling several models that matched as to eight characteristics but could vary significantly in price or cost, due to differences in materials for certain components or added features.

The new model-match methodology not only reflects a reasonable interpretation of the statute but also comports with our precedent. In CEMEX, S.A. v. United States, we observed that:

Section 1677(16) sets up a hierarchy for identifying “such or similar merchandise.” Commerce examines each category in order, and once merchandise is presented that meets the criteria stated by a category, the price of this merchandise in the home country becomes the foreign market value. See 19 U.S.C. §§ 1677(16), 1677b(a)(l). The first class of merchandise is identical merchandise, the next class is nonidentical merchandise that is made by the same producer[*1380] in the same country and is similar in value to the merchandise under investigation, and the third class is merchandise made by the same producer in the same country and used for the same purposes as the merchandise under investigation. See 19 U.S.C. § 1677(16).

133 F.3d 897, 902-03 (Fed.Cir.1998) (internal footnote omitted). In light of this statutory structure, we concluded that “[t]he plain language of the statute requires Commerce to base foreign market value on nonidentical but similar merchandise ... rather than constructed value when sales of identical merchandise have been found to be outside the ordinary course of trade.” Id. at 904. Thus, Commerce’s decision to seek out product matches based on the most similar products rather than constructed values does not contravene the statute.

With respect to Commerce’s decision to revise the model-match methodology after 14 annual reviews utilizing the family-match methodology, we have specifically affirmed changes to model-match methodologies by Commerce where reasonable. See Koyo, 66 F.3d at 1211. In Koyo, Commerce had changed from a “greatest-single-deviation methodology ... to a ‘sum-of-the-deviations’ methodology for matching U.S. [tapered roller bearings (‘TRBs’) ] with home market TRBs ... without using any limitation or ‘cap’ on the deviation of any one criteria.” Id. at 1207. The Court of International Trade affirmed Commerce’s switch in methodologies, but imposed a ten percent cap to limit the permissible deviation. Id. at 1208. We stated that “our inquiry [was] limited to determining whether Commerce’s model-matching methodology, without the ten percent cap later required by the court, is reasonable,” and we concluded that it was. Id. at 1210.

In this case, Commerce provided adequate explanation of the reasonable “compelling reasons” motivating its determination, as detailed supra. Although SKF argues that in implementing the previous model-match methodology, Commerce did not specifically suggest that it was constrained by technological limitations, contemporaneous documentation from Commerce indicates that motivating factors in adopting the family-match methodology included “minimiz[ing] the necessity for comparisons among an exceptionally large number of bearing models,” and “limit[ing] the need for adjustments for physical differences in merchandise and the need for model matching.” J.A. at 129. These concerns were obviated by the availability of “all new margin-calculation programs on personal computers that are several orders of magnitude faster, vastly more sophisticated, yet far easier to use than the offsite mainframe computer [Commerce] used in the 1988-90 administrative reviews.” Id. at 360. Based on Commerce’s increased technological capacity, combined with its desire to fashion a model match methodology more in keeping with its ordinary practice of selecting a single most similar model rather than pooling values of product families, we affirm its decision to revise the model-match methodology.

Even assuming that Commerce was justified in modifying the model-match methodology, however, SKF analogizes to Shikoku Chemicals Corp. v. United States, 795 F.Supp. 417 (Ct. Int’l Trade 1992), to argue that we should not permit Commerce to apply the modified model-match methodology retroactively. In Shikoku, the Court of International Trade found that reliance interests precluded a “late stage” change in model-match methodology. Id. at 422. Appellees counter that in Shikoku, “[t]here [was] no serious challenge to plaintiffs’ contention that they actually relied on the old pricing method,” id. at 420, and that SKF ignores the inher[*1381] ently retrospective nature of the anti-dumping duty statutory scheme. See 19 C.F.R.. § 351.212(a) (“Unlike the systems of some other countries, the United States uses a ‘retrospective’ assessment system under which final liability for antidumping and countervailing duties is determined after merchandise is imported.”); Abitibi-Consol. Inc. v. United States, 437 F.Supp.2d 1352, 1361 (Ct. Int’l Trade 2006) (“The absence of certainty regarding the dumping margins and final assessment of antidumping duties is a characteristic of the retrospective system of administrative reviews designed by Congress.”). We agree. While Commerce must comply with the notice provisions of the statute, “[cjhanges in methodology, like all other antidumping review determinations, permissibly involve retroactive effect.” Koyo Seiko, 516 F.Supp.2d at 1334 (citing 19 U.S.C. § 1675(a)(2)). Appellees also argue that Commerce provided even greater notice and opportunity to comment than provided by statute, see 19 U.S.C. § 1677m(g), thus undermining any reasonable reliance on the family-match methodology by SKF.

SKF alleges that it relied upon the old methodology, but provides no evidence to support that assertion and argues only that it “could have adjusted its practices” had it known sufficient details about the new methodology. However, SKF does not dispute that Commerce has consistently found that SKF continues to sell at dumped prices. Thus, SKF cannot properly analogize its situation to that in Shikoku, where “[tjhe record contained] evidence that plaintiffs adjusted their prices in accordance with methodology consistently applied by Commerce in an attempt to comply with United States antidumping law.” Shikoku, 795 F.Supp. at 420. Having failed to establish detrimental reliance on the old methodology, and given the inherently retroactive nature of the anti-dumping statutory scheme, see 19 U.S.C. § 1675(a)(2), we reject SKF’s argument that it was entitled to greater notice than that provided by Commerce in accordance with — and exceeding the requirements of — 19 U.S.C. § 1677m(g). See NSK Ltd. v. United States, 510 F.3d 1375, 1385 (Fed.Cir.2007) (noting that “Commerce ... publicly stated that it w[ould] change its methodology only after parties have had a meaningful opportunity to comment on the proposed change”).

Finally, SKF challenges Commerce’s decision to consider lubricant types in assessing identical products. Yet SKF fails to explain why we should disregard the Court of International Trade’s observation that SKF “failed to raise this issue during the open comment period pri- or to publication of the final determination.” SKF USA 491 F.Supp.2d at 1363. Nor does SKF’s brief argument on this point undermine Commerce’s justifications for its inclusion of lubrication type in making this determination. See Issues and Decision Memo at 37 (explaining the decision not to ignore differences in lubrication and cautioning that, for example, “a comparison of a U.S. model with a high-performance grease to a home-market model with a standard grease could mask dumping”); see also Pesquera Mares, 266 F.3d at 1384 (holding that “Commerce has considerable discretion in defining ‘identical in physical characteristics’ ”). Thus, we reject SKF’s arguments related to lubricant type.

2. Commerce’s “Zeroing” Methodology

SKF also challenges Commerce’s ongoing “zeroing” practice as a distortive misapplication of the anti-dumping laws. We have addressed the practice of “zeroing” numerous times, however, and have unequivocally upheld this practice. For example, in one recent opinion we explained:

Occasionally, the price charged for the subject merchandise in the United States is greater than the price charged[*1382] for the same merchandise in the home market. This results in a negative dumping margin for that merchandise. In these situations, Commerce sets the negative dumping margins to zero when calculating the weighted average dumping margin. By doing so, the sum of the dumping margins calculated on the individual transactions is not reduced by the negative amount of the dumping margins. This practice is referred to as “zeroing” and has been repeatedly upheld by this court.

NSK, 510 F.3d at 1379 (emphasis added).

We have reviewed SKF’s arguments regarding zeroing and find them unpersuasive. SKF fails to raise any argument not fully resolved by our established precedent. See Corus Staal BV v. United States, 502 F.3d 1370, 1374 (Fed.Cir.2007) (holding that Commerce’s “policy has not changed with respect to the retrospective application of the zeroing methodology and that a remand to Commerce ... would therefore serve no useful purpose”); Corus Staal BV v. Dep’t of Commerce, 395 F.3d 1343, 1349 (Fed.Cir.2005) (“We will not attempt to perform duties that fall within the exclusive province of the political branches, and we therefore refuse to overturn Commerce’s zeroing practice based on any ruling by the WTO or other international body unless and until such ruling has been adopted pursuant to the specified statutory scheme.”); Timken Co. v. United States, 354 F.3d 1334, 1342 (Fed.Cir.2004) (finding that “the statute does not directly speak to the issue of negative-value dumping margins,” and holding that “Commerce based its zeroing practice on a reasonable interpretation of the statute”). Accordingly, we need not revisit this issue today.

III. CONCLUSION

For the foregoing reasons, we conclude that the Court of International Trade correctly determined that Commerce’s modification of the model-match methodology and its application of the zeroing policy in this case are supported by substantial evidence on the record and are otherwise in accordance with law: Accordingly, the decision of the Court of International Trade is

AFFIRMED.