Opinion for the court filed by Circuit Judge LOURIE, in which Chief Judge MICHEL joins. Opinion concurring in the remand, dissenting in part filed by Circuit Judge NEWMAN.
LOURIE, Circuit Judge.
Deere & Company (“Deere”) appeals from the judgment of the United States International Trade Commission, which determined that the sales of European-version self-propelled John Deere forage harvesters in the United States by Intervenors Bourdeau Brothers, Inc., OK Enterprises, and Sunova Implement Co. (collectively, “Bourdeau”) did not violate § 337 of the Tariff Act of 1930 as amended, 19 U.S.C. § 1337. The Commission based its decision on a finding that, although Bourdeau infringed Deere’s trademarks through gray market importation, not “all or substantially all” of Deere’s authorized sales of harvesters in the United States were of the North American version, so Deere was not entitled to an exclusion order. See In the Matter of Certain Agric. Vehicles & Components Thereof, No. 337-TA-M87, 2008 WL 4352378 (ITC Aug. 25, 2008) (“Commission Remand Opinion"). Because the Commission improperly applied the “all or substantially all” test, we vacate and remand.
[1352]*1352Background
Deere manufactures self-propelled forage harvesters for sale in Europe, and it manufactures different self-propelled forage harvesters for sale in the United States. All of Deere’s harvesters are sold under its trademarks, including, in the United States, U.S. Registered Trademarks 1,254,389; 1,502,103; and 1,503,576. New and used Deere harvesters are sold through various distribution channels, including official1 Deere dealers and independent dealers. Independent dealers sell Deere products without any oversight from Deere. Deere has official and independent dealers in both the United States and Europe. Bourdeau is an independent dealer in the United States.
In February 2003, the Commission instituted an investigation based on a complaint filed by Deere. Deere alleged violations of 19 U.S.C. § 1337(a)(1)(C), in which Congress forbade, inter alia, importation of products that were “produced by the owner of the United States trademark or with its consent, but not authorized for sale in the United States,” often called “gray market goods.” Gamut Trading Co. v. Int’l Trade Comm’n, 200 F.3d 775, 777 (Fed.Cir.1999). Deere contended that Bourdeau and other independent and official Deere dealers based in both the United States and Europe had infringed Deere’s trademarks by unlawfully importing and selling Deere’s European-version harvesters in the United States.
In May 2004, the Commission issued a general exclusion order prohibiting importation of European-version harvesters manufactured by or under the authority of Deere bearing Deere’s trademarks, finding that they infringed Deere’s U.S. trademarks. Bourdeau appealed, and in March 2006, a panel of this court vacated in part and remanded. See Bourdeau Bros., Inc. v. Int’l Trade Comm’n, 444 F.3d 1317 (Fed.Cir.2006). We held that substantial evidence supported the Commission’s determination that there were material differences between Deere’s North American-version and European-version harvesters, supporting a finding of infringement and thus the exclusion order. Id. at 1324-25.
However, we vacated in part and remanded the Commission’s decision based on a particular requirement for recovery, under our recent decision in SKF USA, Inc. v. International Trade Commission, 423 F.3d 1307 (Fed.Cir.2005), that Deere also show that all or substantially all of Deere’s authorized domestic products are materially different from the accused gray market goods. Bourdeau, 444 F.3d at 1325-27. In other words, Deere had to show that substantially all of the Deere harvesters being sold in the United States that were authorized by Deere were North American-version harvesters, as opposed to European-version harvesters. In discussing the “all or substantially all” requirement, we stated:
As we noted in SKF, the sale by a trademark owner of the very same goods that he claims are gray market goods is inconsistent with a claim that consumers will be confused by those alleged gray market goods. “To permit recovery by a trademark owner when less than ‘substantially all’ of its goods bear the material difference ... would allow the owner itself to contribute to the confusion by consumers that it accuses gray market importers of creating.” That is, a trademark owner has the right to determine the set of characteristics that are associated with his [1353]*1353trademark in the United States; however, a trademark owner cannot authorize the sale of trademarked goods with a set of characteristics and at the same time claim that the set of characteristics should not be associated with the trademark.
Id. at 1321 (quoting SKF, 423 F.3d at 1315) (alterations in original).
In our remand instructions, we noted that Deere might have authorized some or all sales of European-version harvesters in the United States and that there was a presumption, which Deere could rebut on remand, that all sales by official Deere dealers were authorized. Bourdeau, 444 F.3d at 1327. Thus, we remanded to give Deere an opportunity to show that sales of European-version harvesters in the United States by official U.S. and European dealers were not in fact authorized by Deere and, to the extent that any such sales were authorized by Deere, that substantially all of its authorized domestic sales were nevertheless of North American-version harvesters. Id.
The Commission remanded the investigation to the administrative law judge (“ALJ”) for proceedings consistent with our decision in Bourdeau. On remand, the ALJ issued an initial decision in December 2006. In the Matter of Certain Agric. Vehicles & Components Thereof, No. 337-TA-487, 2006 ITC Lexis 862 (ITC Dec. 20, 2006) {“ALJ Remand Opinion ”). The ALJ found that the original record showed that Deere did not authorize the sales of European-version harvesters in the United States. Id. at *39-43. With respect to new evidence concerning alleged Deere financing of European-version harvesters sold by its dealers, the ALJ found that that evidence did not show authorization; hence the ALJ found infringement. Id. The ALJ also found that the number of sales that Bourdeau alleged were authorized was, in any event, so small that “substantially all” of Deere’s authorized U.S. sales were of North American-version harvesters. Id. at *47-48. Thus, even if the ALJ had agreed with Bourdeau’s contentions, any unauthorized European-version harvester sales would have been infringing. The parties each petitioned the Commission for review.
In August 2008, the Commission reversed the ALJ. Commission Remand Opinion, No. 337-TA-487. The Commission first determined that the ALJ had failed to consider whether official Deere dealers had had apparent authority to sell European-version harvesters in the United States, despite the absence of actual authority. Id., slip op. at 12-13. The Commission reasoned that, because trademark law focuses on the potential for third-party confusion in the marketplace, apparent authority is sufficient to constitute “authority” under our remand instructions. Id. at 16-17.
The Commission then found “substantial evidence” that Deere’s U.S. and European dealers had apparent authority to sell European-version harvesters. Id. at 13. The Commission explained that third parties reasonably could have and did construe Deere’s acts and omissions as condoning the importation and sale of European-version harvesters. Id. at 17. According to the Commission, our remand instructions included a presumption that all official Deere dealer sales were authorized by Deere, and Deere had failed to rebut that presumption. Id. at 19. The Commission found that Deere had presented only conclusory testimony to argue that it had not authorized its official European dealers to sell European-version harvesters in the United States, and it found that Deere had been on notice that the sales activities of its official European dealers were at issue in the remand. Id. at 19-22. Thus, the Commission presumed that all sales of Eu[1354]*1354ropean-version harvesters to the United States by Deere’s official European dealers were authorized. Id. at 22-23. (If they had not been authorized, then such sales surely would have been infringing.) It also presumed, based on a lack of evidence from Deere, “that the volume of such sales was sufficiently great that a full accounting would have demonstrated” that a substantial number of Deere’s sales of harvesters in the United States were of the European version and hence that Deere was itself contributing to consumer confusion, supporting a conclusion of non-infringement. Id. at 23-24.
The Commission also found that official U.S. Deere dealers and Deere itself sold and/or facilitated the sale of European-version harvesters in the United States. According to the Commission, Deere was aware for five years of the growing market penetration of European-version harvesters in the United States but did not discourage such sales until the end of the period. Id. at 29. Indeed, the Commission found that Deere honored European warranties on European-version harvesters that were sold in the United States. Id. The Commission further found that Deere could have stopped or curtailed gray market imports by its official dealers if it had wished to do so. Id. at 33-34. The Commission also found that Deere’s global website, machinefinder.com, which acts as an international clearing house for used harvesting equipment, would have given customers the impression that Deere condoned the international sale of its harvesters, including European-version harvesters in the United States. Id. at 43-46. Moreover, the Commission found that Deere’s credit arm, JD Credit, would have given a purchaser or dealer the impression that European-version harvester sales in the United States were authorized because JD Credit financed such sales. Id. at 46-48.
Finally, the Commission determined that not “all or substantially all” of the authorized harvesters sold in the United States were North American-version harvesters. Id. at 26-28, 49-51. The Commission reasoned that harvesters are highly specialized and expensive and therefore are sold for high prices at low sales volumes. Id. at 49. In finding low sales volumes, the Commission credited the ALJ’s finding that the total number of authorized sales of North American-version harvesters in the United States, both new and used, was approximately 4400. Id. at 49 n. 9. With such low sales volumes, according to the Commission, the introduction of even a small number of European-version harvesters could cause substantial confusion. The Commission then found such confusion.
The Commission next found that 40 to 57% of the European-version harvesters sold in the United States were sold by official Deere dealers.2 Id. at 51. To arrive at its numbers, the Commission divided the number of authorized European-version harvesters by the total number of European-version harvesters sold in the United States. The Commission found that at least 141 European-version harvesters were sold in the United States by official Deere dealers. Id. at 28, 50. The Commission indicated that the record showed a possible 14 more authorized European-version harvesters sold in the United States. Id. at 27 (stating that “Stanley [1355]*1355purchased between 3 and 17 [European-version harvesters] from official European John Deere dealers” but only adding 3 to arrive at the final 141 count). The Commission found that a total of 247 to 347 European-version harvesters were sold in the United States, by both official and independent (in some cases, accused) dealers. Id. at 50-51.
Dividing the smaller number of official European-version harvesters sold in the United States (141) by the total number of European-version harvesters sold in the United States (247 to 347), the Commission arrived at its 40 to 57% number. Id. at 51. The Commission did not use as its denominator the total number of authorized harvesters sold in the United States, which would have been the number of authorized North American-version harvesters (4400) plus the number of authorized European-version harvesters (141), or 4541. The Commission thus concluded that official Deere dealers were responsible for introducing a “substantial quantity” of nonconforming goods into U.S. commerce. Id.
Deere timely appealed the Commission’s final determination that it was not entitled to recover for a violation of § 337 for gray market trademark infringement. We have jurisdiction pursuant to 28 U.S.C. § 1295(a)(6).
Discussion
Pursuant to the Administrative Procedure Act, we review the factual findings of the Commission for substantial evidence. See 19 U.S.C. § 1337(c); 5 U.S.C. § 706(2)(E). We thus will not overturn the Commission’s factual findings if they are supported by “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Finnigan Corp. v. Int’l Trade Comm’n, 180 F.3d 1354, 1362 (Fed.Cir.1999) (quoting Surface Tech., Inc. v. Int’l Trade Comm’n, 801 F.2d 1336, 1340-41 (Fed.Cir.1986)). We review the Commission’s legal determinations de novo. See 5 U.S.C. § 706(2)(A); Checkpoint Sys. v. Int’l Trade Comm’n, 54 F.3d 756, 760 (Fed.Cir.1995).
Deere argues that all or substantially all of Deere’s sales of harvesters in the United States were of the North American version, so, according to Deere, the Commission erred in determining that Bourdeau did not infringe. In other words, Deere asserts that an insubstantial number of its authorized harvester sales in the United States were of the European-version harvesters. Deere argues that the Commission erred first in its determination of the number of authorized European-version harvesters sold in the United States such that those sales did not contribute to consumer confusion. Deere argues, second, that even using the number of authorized European-version harvesters sold in the United States that the Commission found, the Commission still erred in applying the “all or substantially all” test by using the incorrect denominator (the total number of European-version harvesters sold in the United States, 247 to 347, instead of the total number of authorized harvesters sold in the United States, 4541). Thus, according to Deere, even accepting the Commission’s numbers but using the correct denominator (4541), an insubstantial number of the harvesters sold in the United States were of the European version. We address those two main arguments in turn.
A. Substantial Evidence Supports the Commission’s Determination that the Sales of European-Version Harvesters in the United States by Official Deere Dealers Were Authorized
Deere argues that the Commission erred in the first part of the Bourdeau analysis by finding that Deere had authorized any of the sales of European-version [1356]*1356harvesters in the United States. Deere first asserts that the Commission erroneously included in “authorized sales” those by official European Deere dealers to unauthorized dealers in the United States, even though this court only instructed it to look at sales “in the United States.” Also, according to Deere, basic principles of U.S. trademark law require considering only sales in the United States.
Deere also argues that it did not have an opportunity to introduce evidence that its official European dealer sales should not be considered “authorized.” Thus, Deere asserts that the Commission should have remanded again to the ALJ for more evidence instead of deciding the authorization issue in the first instance. Further, according to Deere, before the remand the ALJ had already found, and we did not reverse, that Deere was unaware that its official European dealers were selling European-version harvesters for importation into the United States. Thus, Deere asserts that law of the case precludes a finding of authorization. Moreover, Deere argues that under European law, Deere could not have stopped the importation even if it had been aware of it. Thus Deere argues that its lack of knowledge prohibited a finding of authorization.
Finally, Deere asserts that substantial evidence did not support the Commission’s determinations. According to Deere, the ALJ had considered the issue of apparent authority in its pre-remand initial determination, and none of the ALJ’s findings of fact have been overturned. Deere asserts that the Commission’s finding of apparent authority relied on independent dealer sales, contradicting the ALJ’s pre-remand findings. According to Deere, the Commission also should not have relied on the machinefinder.com website, which was informational only. Finally, Deere asserts that the Commission should not have relied on the financing of the sale of a few European-version harvesters through the JD Credit subsidiary, which Deere did not know about or control. Thus, Deere argues that the Commission’s decision should be reversed for a lack of substantial evidence.
Bourdeau and the government respond that the Commission properly accounted for sales by official European dealers to the United States. According to Bourdeau and the government, 19 U.S.C. § 1337 forbids importation, so sales by Deere’s official European dealers that result in importation should be considered in the authorization analysis. Bourdeau and the government also assert that the remand instructions from this court do not differentiate between sales by European and U.S. dealers, indicating that we intended the calculation to include sales by official European dealers.
In response to Deere’s argument that it did not have an opportunity before the Commission to introduce evidence, Bourdeau and the government argue that Deere had such an opportunity and simply failed to introduce arguments or evidence to overcome the presumption of authorization; indeed, the Commission requested additional briefing on the authorization issue. According to Bourdeau and the government, Deere introduced no documentary proof that it had tried to prohibit its official European dealers from selling European-version harvesters in the United States. Further, Bourdeau and the government argue that Deere has no support for its claim that European trademark law prohibited Deere from stopping the importation.
Finally, Bourdeau and the government respond that substantial evidence supported the Commission’s determinations. According to Bourdeau and the government, Deere actively supported the conduct of its official dealers, including their [1357]*1357sale of European-version harvesters. Further, Bourdeau and the government assert that Deere promoted European-version harvesters on its machine-finder.com website and financed their sale through JD Credit. Because the Commission reviews the ALJ’s findings de novo and makes its own findings, Bourdeau and the government assert that the Commission committed no error in making different findings from the ALJ’s.
We agree with Bourdeau and the government that the Commission permissibly found that sales by official Deere dealers of European-version harvesters in the United States were authorized. As the Commission correctly held, apparent authority constitutes “authority” for purposes of this case. Apparent authority arises from buyers’ reasonable belief, based on the acts and omissions of Deere, that sales were authorized. See Whetstone Candy Co. v. Kraft Foods, Inc., 351 F.3d 1067, 1078 (11th Cir.2003) “Apparent authority [may] exist [ ] when ... the principal knowingly permits the agent to act as if the agent is authorized, or ‘by silently acting in a manner which creates a reasonable appearance of an agent’s authority.’ ” (citation omitted); Bethany Pharmacol Co. v. QVC, Inc., 241 F.3d 854, 859-60 (7th Cir.2001); Grajales-Romero v. Am. Airlines, Inc., 194 F.3d 288, 293 (1st Cir.1999). The Commission properly looked to one of the policies underlying trademark law— avoiding third-party confusion in the marketplace — -to hold that apparent authority should suffice to find sales “authorized.” Commission Remand Opinion, slip op. at 13; see Westinghouse Elec. Corp. v. Gen. Circuit Breaker & Elec. Supply, 106 F.3d 894, 899 (9th Cir.1997) (focusing, for trademark infringement purposes, on likelihood of confusion in the minds of consumers).
Further, substantial evidence supports the Commission’s finding that Deere’s official dealers, including its official European dealers, had apparent authority to sell Deere European-version harvesters. As the Commission permissibly concluded, Deere’s acts indicated that it condoned such sales. For example, Deere promoted its machinefinder.com website, which allowed dealers to advertise European-version harvesters for sale in the United States by official Deere dealers. Although Deere argues that the website was informational only, substantial evidence supported the Commission’s conclusion that even an informational website could lead users to the conclusion that the sales it was promoting were supported by Deere. JD Credit’s financing of purchases of European-version harvesters in the United States by official dealers also supported the Commission’s conclusion that those official dealers had apparent authority to sell the harvesters. Although Deere argues that it had no control over its JD Credit subsidiary, again the Commission appropriately relied on the public’s perception of Deere’s actions in allowing such financing to take place, rather than the direct control that Deere exercised.
The Commission also did not err in including sales by Deere’s official European dealers in the United States in its calculation. As Bourdeau and the government point out, authorized sales by European dealers of European-version harvesters introduced into the United States must be included in total “authorized sales” because, absent such authorization, they would constitute “importation into the United States ... of articles that infringe” a U.S. trademark. 19 U.S.C. § 1337(a)(1)(C). We also agree with the government and Bourdeau that, under our remand instructions, sales by Deere’s official European dealers to the United States should be counted in the calculation of authorized European harvesters. Our earlier opinion never distinguished between U.S. and European Deere dealers. In[1358]*1358stead, our remand instructions stated that, in general, “the [Commission] must presume that sales by authorized dealers were in fact authorized by Deere,” and that “Deere bears the burden of proving that sales of European forage harvesters by its authorized dealers were not authorized sales.” Bourdeau, 444 F.3d at 1327. Indeed, we consistently referred to both types of official dealers throughout the Bourdeau opinion. See, e.g., id. at 1325 (“Appellants asserted that more than fifty used European forage harvesters were sold to them by authorized Deere dealers in the United States and in Europe.”); id. (On the machinefinder.com website, “the worldwide network of authorized Deere dealers” allows “United States consumers [to] search for European forage harvesters offered for sale by authorized Deere dealers in Europe.”).
Finally, we agree with Bourdeau and the government that Deere did in fact have an opportunity to introduce evidence regarding authorization of its official European dealer sales. As Bourdeau and the government point out, the Commission even requested additional briefing on the authorization issue. As for Deere’s claim that European trademark law precluded Deere from stopping the importation, the Commission permissibly found that the “claim was not sufficiently developed during the remand proceeding to permit the Commission to draw any conclusions as to what is permissible under EU law.” Commission Remand Opinion, slip op. at 26 n. 5. We further agree with Bourdeau and the government that the Commission was permitted to look beyond the ALJ’s findings, as it reviews all of the ALJ’s findings de novo. See 5 U.S.C. § 557(b) (giving agencies “all the powers which it would have in making the initial decision except as it may limit the issues on notice or by rule”); 19 C.F.R. § 210.45(c) (implementing 5 U.S.C. § 557(b) by allowing the Commission to “make any findings or conclusions that in its judgment are proper based on the record in the proceeding”); see also Kay v. Fed. Commc’ns Comm’n, 396 F.3d 1184, 1189 (D.C.Cir.2005) (holding that an agency need not accept any of the ALJ’s findings, “even if the ALJ’s findings rested on his evaluation of the credibility of the witnesses”). Thus, the Commission was within its power to find facts beyond those found by the ALJ and to request briefing on the issue of apparent authority.
The dissent argues that we need not discuss and decide the question of authorization of the sales of European-version harvesters in the United States by official Deere dealers. The dissent also questions our reasons for concluding that the sales were authorized. However, as conceded by the dissent, the “all or substantially all” calculation requires dividing the number of authorized European-version harvesters sold in the United States, the numerator, by the total number of authorized harvesters sold in the United States, the denominator. In order to divide the numerator by the denominator, we need to have calculated the numerator. Thus, in order to enable the Commission to reapply the “all or substantially all” test on remand, we must give the Commission guidance on whether it appropriately calculated the numerator. We therefore must decide whether the sales were authorized by Deere. Furthermore, although the dissent may disagree with our reasoning, we must affirm if the Commission’s decision on authorization was supported by substantial evidence, and we believe it was.
B. The Commission Misapplied the “All or Substantially All” Test
Deere argues that, irrespective of the Commission’s finding that Deere authorized certain European-version harvester sales in the United States, the Commission [1359]*1359erred in its application of the “all or substantially all” test. According to Deere, the Commission failed to follow our remand instructions in Bourdeau, which required a comparison of authorized sales of European-version harvesters in the United States (which the Commission found to be 141) with total authorized harvester sales in the United States (the sum of the authorized North American-version harvesters, which the Commission found to be 4400, and authorized European-version harvesters, which the Commission found to be 141, or a total of 4541). Thus, according to Deere, even using the Commission’s findings of authorized sales, only 3.1% (141/4541) of the authorized harvesters sold in the United States were of the European version.
Deere further argues that 3.1% is an insignificant number under the “all or substantially all” test. Deere illustrates that argument by pointing to Warner-Lambert Co. v. Northside Dev. Corp., 86 F.3d 3 (2d Cir.1996), in which the Second Circuit found that authorization of 4.4% of sales was a small enough number to allow a trademark owner relief. Thus, according to Deere, if 4.4% is small enough to be considered insignificant, 3.1% is even more insignificant, such that all or substantially all (96.9%) of Deere’s authorized harvesters were of the North American version.
Deere finally asserts that the Commission erred by comparing the number of authorized European-version harvesters (which the Commission found to be 141) with total European-version harvesters in the United States regardless of source (which the Commission found to be between 247 and 347), to arrive at 40 to 57% (141/347 = 40% and 141/247 = 57%). According to Deere, the Commission’s denominator of 247 to 347 instead of 4541 violates the reasoning behind the “all or substantially all” test. Deere argues that the test is intended to avoid consumer confusion based on the differences between gray market goods and authorized goods. The Commission did not state that it was precluded from applying the correct test due to insufficient evidence, and Deere asserts that the government cannot now argue that there was insufficient evidence.
Bourdeau and the government respond that Deere failed to submit comprehensive proof of its official European dealers’ sales to the United States or the total number of used harvesters sold. Bourdeau and the government assert that, because Deere had the burden of proof, the Commission permissibly inferred a large ratio.
The government further argues that, even using the Commission’s number of authorized European-version harvesters, which was 141, the Commission properly found it to be substantial given the total number of European-version harvesters sold in the United States (between 247 and 347). In other words, the government asserts that Deere does not deserve to prevail when it caused a substantial portion of the consumer confusion that it now complains of. The government also argues that the Commission reasonably concluded that the introduction of even a small number of European-version harvesters would cause consumer confusion. According to the government, there is no benchmark ratio of 4.4%, as Deere argues; instead, the government argues that under SKF, “substantially all” depends on the facts of the case, and that this court entrusted such fact-finding to the Commission.
Finally, Bourdeau asserts that the percentage of nonconforming sales is only one factor that Deere must establish to meet the “all or substantially all” test. According to Bourdeau, under SKF, Deere also must show that the infringing sales measurably diminish the final value of an already partially devalued mark. Bourdeau asserts that Deere failed to establish any [1360]*1360diminishment in the value of its trademark as a result of Bourdeau’s sales.
We agree with Deere that the Commission misapplied the “all or substantially all” test. In SKF, we reasoned that
the consuming public, associating a trademark with goods having certain characteristics, would be likely to be confused or deceived by goods bearing the same mark but having materially different characteristics. Conversely, then, a trademark owner’s argument that consumers would be confused by gray goods lacking an asserted material difference from the authorized goods is inconsistent with the owner’s own sale of marked goods also lacking that material difference from its own authorized goods. To permit recovery by a trademark owner when less than “substantially all” of its goods bear the material difference from the gray goods thus would allow the owner itself to contribute to the confusion by consumers that it accuses gray market importers of creating.
423 F.3d at 1315 (quotation marks and citation omitted). Thus, applying that reasoning to this case (in Bourdeau), our remand instructions stated that Deere would prevail if it could establish by a preponderance of the evidence “that the number of sales of European forage harvesters was so small that substantially all of Deere’s sales in the United States were of North American forage harvesters, such that substantially all of the authorized sales were of goods bearing the asserted material differences.” 444 F.3d at 1327.
Our remand thus set out a test for the Commission to follow. The question was whether “substantially all of the authorized sales,” ie., the sum of authorized North American-version and authorized European-version harvester sales, were of North American-version harvesters. The denominator therefore should have been total authorized sales, not total European-version harvester sales, in the United States.
Deere also points out that the Commission actually found both the number of authorized North American-version and the number of authorized European-version harvester sales, and it did not assert that either number was based on insufficient evidence. Contrary to Bourdeau and the government’s arguments, the Commission was therefore not entitled to infer, based on its findings, that the ratio was “large.” Indeed, the Commission added up all of the authorized European-version harvesters sold in the United States, compiling numbers from different sources, and found that “the record indicates that at least 141” were authorized. Commission Remand Opinion, slip op. at 27-28; see id. at 49-51. The Commission indicated that the record showed a possible 14 more authorized European-version harvesters sold in the United States. Id. at 27. Thus, the Commission found the number of authorized European-version harvester sales in the United States to be between 141 and 155. The Commission credited the ALJ’s estimate of approximately 4400 as the number of Deere’s authorized sales of North American-version harvesters in the United States. Id. at 49 n. 9; see ALJ Remand Opinion, 2006 ITC Lexis 862 at *59-60. Thus, the Commission found the total number of authorized harvester sales in the United States to be between 4541 (141 + 4400) and 4555 (155 + 4400).
Using the ratio that was dictated by our remand instructions and using the Commission’s findings, we conclude that a total of 3.1 to 3.4% of the authorized harvesters sold in the United States were European-version harvesters, or that 96.6 to 96.9% of the authorized harvesters sold in the United States were of the North American version. In other words, if one takes the [1361]*1361Commission’s lower-end finding, 141/ (4400 +141) = 3.1% of the authorized harvesters sold in the United States were European-version harvesters, and if one accepts the Commission’s highest numbers, 155/(4400 + 155) = 3.4% of the authorized harvesters sold in the United States were European-version harvesters.
Those figures may be insubstantial. However, that is for the Commission, not this court, to determine on the basis of all of the relevant facts. We therefore remand for the Commission to determine whether 3.1 to 3.4% is an insubstantial percentage, such that substantially all of the authorized harvesters sold in the United States were of the North American version. The cutoff as to what is to be considered “substantially all” is a question of fact. See SKF, 423 F.3d at 1317 (“The determination [ ] by the Commission that ... all or substantially all of the goods were not [the materially different foreign version is a] finding [ ] of fact, subject to substantial evidence deference by this court.”). On the one hand, we note the Commission’s indication that the “relatively high prices and low sales volumes” of Deere’s harvesters might lead to the conclusion that “the introduction of even a small number of [European-version] harvesters into the U.S. market could cause substantial confusion by consumers and have a significant impact on the marketplace.” Commission Remand Opinion, slip op. at 49-50. On the other hand, only 3.1 to 3.4% of the authorized harvesters sold in the United States being nonconforming might not cause confusion. We also note that the ALJ considered the “all or substantially all” test and found, using a reasonable numerator and denominator, that the percentage of authorized sales in the United States that were of North American-version harvesters was large enough to meet the “all or substantially all” test. See ALJ Remand Opinion, 2006 ITC Lexis 862 at *46-65. Furthermore, in SKF, we discussed what percentage might be considered “substantially all.” We disagreed with Martin’s Herend Imports, Inc. v. Diamond & Gem Trading USA, Co., 112 F.3d 1296 (5th Cir.1997), which held “that the sale of even one authorized item lacking a material difference defeats infringement.” SKF, 423 F.3d at 1316. We determined that “[i]nstead, the ‘all or substantially all’ benchmark recognizes that something less than 100% compliance will suffice and certainly permits a small amount of nonconforming goods. A single sale of a nonconforming item typically should not defeat a trademark owner’s protection.” Id. We also cited Warner-Lambert, 86 F.3d 3. SKF, 423 F.3d at 1311, 1316. In Warner-Lambert, the court indicated that, if “only 4.4%” of the plaintiffs cough drops were found to be nonconforming goods, that would favor a finding of infringement. 86 F.3d at 8 n. 1. Thus, as a general matter, 95.6% might well be considered to be “substantially all.”
Furthermore, because of our reversal of the Commission’s decision, the Commission is not entitled to rely, as it did in the previous remand proceeding, on the ratio of authorized European-version harvesters to the total number of European-version harvesters sold in the United States. Although one could conceive of a test that determines whether the trademark owner contributed less than, as much as, or more than the accused infringer to consumer confusion, the contribution cannot be relevant unless enough authorized products are sold to likely cause consumer confusion. Thus, the benchmark is instead whether consumers would likely have been confused by the trademark owner’s actions. As the Commission itself noted in the authorization context, “the focus of trademark infringement law is on the potential for third party confusion in the marketplace.” Commission Remand Opinion, slip op. at 13. Our remand instructions focused on the potential for con[1362]*1362sumer confusion, and therefore the Commission should determine in this remand whether 96.6 to 96.9% is “substantially all.”
Conclusion
We have considered the parties’ remaining arguments and do not find them persuasive. Accordingly, the judgment of the Commission is
VACATED and REMANDED
Opinion for the court filed by Circuit Judge LOURIE, in which Chief Judge MICHEL joins.
Opinion concurring in the remand, dissenting in part filed by Circuit Judge NEWMAN.