Memorandum and Opinion
Goldberg, Judge:This matter is before the court on the parties’ motions for judgment on the agency record made pursuant to USCIT *830Rule 56.2. Petitioner, Micron Technology, Inc. (“Micron”) and respondents: Hyundai Electronics Industries Co., Ltd. and Hyundai Electronics America, Inc. (collectively, “Hyundai”); LG Semicon Co., Ltd. and LG Semicon America, Inc. (collectively, “Semicon”); and Samsung Electronics Co., Ltd. and Samsung Semiconductor, Inc. (collectively, “Samsung”), contest various aspects of the final determination of sales at less than fair value issued and amended by the United States Department of Commerce, International Trade Administration (“Commerce”), concerning dynamic random access memory semiconductors (“DRAMs”) of one megabit (“1M”) and above from the Republic of Korea. Dynamic Random Access Memory Semiconductors of One Megabit and Above from the Republic of Korea, 58 Fed. Reg. 15,467 (Mar. 23,1993) (“Final Determination”); Dynamic Random Access Memory Semiconductors of One Megabit and Above from the Republic of Korea, 58 Fed. Reg. 27,520 (May 10, 1993) (“Amended Final Determination”). The court exercises its jurisdiction pursuant to 28 U.S.C. § 1581(c) (1988).
I. Background
On April 22, 1992, Micron filed with Commerce a petition alleging, inter alia, that Korean DRAMs of one megabit and above were being sold in the United States at less than fair value, and that home market sales of the DRAMs were taking place at prices less than the cost of production. See Dynamic Random Access Memory Semiconductors of One Megabit and Above From the Republic of Korea, 57 Fed. Reg. 21,231 (May 19, 1992) (“Initiation of Investigation”). In response to Micron’s petition, Commerce initiated an investigation of Korean DRAMs covering the period from November 1, 1991 through April 30, 1992. Final Determination, 58 Fed. Reg. 15,467, 15,468. On March 23, 1993, Commerce published its Final Determination of sales at less than fair value. Final Determination, 58 Fed. Reg. at 15,467. To correct errors contained in the Final Determination, Commerce issued an amended final determination on May 6, 1993. Amended Final Determination, 58 Fed. Reg. at 27,520.
II. Discussion
In deciding a motion for judgment on the agency record, the court assesses whether Commerce’s determination is “unsupported by substantial evidence on the record, or otherwise not in accordance with law.” 19 U.S.C. § 1516a(b)(l)(B) (1988). Substantial evidence is something more than a mere scintilla. Cerámica Regiomontana, S.A. v. United States, 10 CIT 399, 405, 636 F. Supp. 961, 966 (1986) (citations omitted), aff’d, 5 Fed. Cir. (T) 77, 810 F.2d 1137 (1987). Substantial evidence consists of “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion. ” Matsushita Elec. Indus. Co. v. United States, 3 Fed. Cir. (T) 44, 51, 750 F.2d 927, 933 (1984) (citation omitted).
In moving for judgment on the agency record, the parties raise a total of fifteen issues for the court’s review. Specifically, the parties contest *831the following aspects of Commerce’s Final Determination, as amended: (1) whether Commerce properly allocated research and development costs on the basis of aggregate rather than product-specific data; (2) whether Commerce properly allocated interest expense based upon semiconductor fixed assets; (3) whether Commerce properly identified and corrected an error related to the allocation of Hyundai’s interest expense; (4) whether Commerce properly applied a time lag when comparing cost of production and constructed value to sales; (5) whether Commerce properly included and expensed translation losses in calculating cost of production; (6) whether Commerce properly used BIA to adjust Samsung’s reported depreciation expense; (7) whether Commerce’s choice of best information available for calculating Samsung’s depreciation expenses was proper; (8) whether Commerce properly deducted U.S. direct selling expenses from Hyundai’s exporter’s sales price; (9) whether Commerce properly reclassified Semicon’s capitalized costs of facility construction and testing; (10) whether Commerce properly included a small volume of sales of “off-spec” merchandise in its calculations for Semicon; (11) whether Commerce treated Semicon’s sales of “Product 12” properly; (12) whether Commerce properly included future generations of DRAMs within the scope of its investigation; (13) whether Commerce properly verified Hyundai’s cost of production questionnaire response; (14) whether Commerce properly verified Semicon’s cost of production questionnaire response; and (15) whether Commerce properly determined that Samsung and [ ^ ] are not related parties. Each will be addressed in turn.
1. Commerce’s Allocation of Research and Development Costs:
Hyundai, Samsung, and Semicon (“respondents”), each challenge the methodology chosen by Commerce to allocate research and development (“R&D ”) costs for the purpose of calculating the cost of production (“COP”) of the subject DRAMs. Respondents raise two objections to Commerce’s allocation methodology; each will be considered in turn.
First, respondents challenge Commerce’s decision to calculate the R&D costs related to the subject DRAMs by allocating the R&D costs related to all semiconductors over the total cost of sales for all semiconductors. Commerce allocated R&D costs using aggregate data rather than product-specific data based upon its determination that, “[bjecause the general underlying technology is the same for all semiconductor products, the benefits from the results of R&D, even if intended to advance the design or manufacture of a specific product, provide an intrinsic benefit to other semiconductor products.” Final Determination, 58 Fed. Reg. at 15,472. Respondents contend that the administrative record compiled in this case does not support Commerce’s conclusion that R&D cross-fertilization in the semiconductor industry warrants the use of aggregate data to allocate R&D costs for the subject DRAMs. The court agrees.
*832To begin, the court recognizes that R&D cross-fertilization may justify allocating R&D costs on an aggregate, rather than product-specific, basis. Indeed, if substantial record evidence supports a determination that the subject merchandise benefits from R&D expenditures earmarked for non-subject merchandise, it would then be appropriate to account for such expenditures in calculating the cost of producing the subject merchandise. The court emphasizes, however, that the validity of Commerce’s methodology cannot rest on intuitive appeal alone; rather, the factual premise upon which Commerce bases its choice of methodology must be supported by substantial evidence on the record. Upon review, the court finds that in this particular instance it is not.
Significantly, in defending its choice of methodology, Commerce fails to direct the court to any record evidence of R&D cross-fertilization in the semiconductor industry. See Defendant’s Response Brief at 5-8. Instead, the defendant offers mere speculation, stating that “there is an overlap of technology between different semiconductors, the result of which is that R&D performed for one model may provide a benefit to another model.” Id. at 6-7 (emphasis added). Micron attempts to justify Commerce’s methodology by citing to a single remark offered by a Micron official at the hearing before Commerce, indicating that R&D related to one aspect of one particular product will likely benefit current and future generations of DRAMs. Micron’s Opposition Brief at 22-23. If R&D cross-fertilization in the semiconductor industry is as widespread as defendant claims, one would expect an abundance of corroborating evidence which supports and expands upon the statement offered by the Micron official. Yet, there is none. Standing alone, this statement offered by a self-interested party after the record had been closed hardly constitutes substantial evidence upon which Commerce could base so sweeping a determination.
In contrast, respondents provide ample citation to verified record evidence tending to show that the subject merchandise did not derive an intrinsic benefit from R&D related to other semiconductor products. Most notably, the bulk of Semicon’s R&D expenditures related to the subject merchandise consisted of [ ]. Semicon’s R.56.2 Brief at 17-19. In addition, Samsung notes that [ ]. Samsung’s R.56.2 Brief at 32-37. Neither Commerce nor Micron adequately address this substantial record evidence supporting respondents’ position.
For the foregoing reasons, upon reviewing the entirety of the record compiled in this case, the court concludes that Commerce’s determination that R&D related to non-subject merchandise provided an intrinsic benefit to other semiconductor products such as the subject DRAMs, is not supported by substantial evidence. As a result, the court remands this issue with instructions that Commerce recalculate respondents’ COP by allocating R&D costs on a product-specific basis. Commerce shall then amend its Final Determination accordingly.
*833Second, respondents challenge Commerce’s decision to reject Korean generally accepted accounting principles (“GAAP”), which permit amortization of R&D expenses over a three to five year period; instead, Commerce expensed respondents’ R&D costs in the year in which they were incurred. Respondents argue that Commerce’s methodology represents a departure from past practice for which the agency has failed to provide an adequate explanation. The court agrees.
Commerce will apply the GAAP of the home country if it is satisfied that those principles reasonably reflect all of the costs associated with production of the subject merchandise. See, e.g., Camargo Correa Metais, S.A. v. United States, 17 CIT 897, 899, Slip Op. 93-163 at 4 (Aug. 13,1993). In the following determinations, Commerce uniformly amortized R&D costs related to the subject merchandise rather than expens-ing such costs in the year incurred: Certain Welded Stainless Steel Pipe From the Republic of Korea, 57 Fed. Reg. 53,693, 53, 705 (Nov. 12,1992) (final determination); Polyethylene Terephthalate Film, Sheet, and Strip From the Republic of Korea, 56 Fed. Reg. 16,305,16,312-13 (Apr. 22,1991) (final determination); Certain Small Business Telephone Systems and Subassemblies Thereof From Korea, 54 Fed. Reg. 53,141, 53,149 (Dec. 27,1989) (final determination); Color Television Receivers From Korea, 53 Fed. Reg. 24,975, 24,982 (July 1,1988) (final results of admin, review); Erasable Programmable Read Only Memories (EPROMs) From Japan, 51 Fed. Reg. 39,680, 39,682 (Oct. 30, 1986) (final determination); 64KDynamic Random Access Memory Components (64K DRAMs) From Japan, 51 Fed. Reg. 15,943,15,944-45 (Apr. 29,1986) (final determination); Cellular Mobile Telephones and Subas-semblies From Japan, 50 Fed. Reg. 45,447,45,453 (Oct. 31,1985) (final determination); Cell Site Transceivers From Japan, 49 Fed. Reg. 43,080, 43,082-83 (Oct. 26, 1984) (final determination). These investigations reflect an established practice on the part of Commerce to amortize R&D expenses related to the subject merchandise in accordance with home market GAAP, which Commerce found not to be distor-tional, for the purpose of allocating such costs over the market life of the product.
Commerce may abandon or reverse an established practice or policy and continue to enjoy deference from the courts. See, e.g., Mantex, Inc. v. United States, 17 CIT 1385, 1399, Slip Op. 93-242 at 27 (Dec. 22,1993) (citation omitted). Such deference, however, is predicated upon a finding that Commerce’s decision to revise its position is based upon a reasoned analysis. See Rust v. Sullivan, 500 U.S. 173, 186-87 (citing Motor Vehicle Mfrs. Ass’n of the United States, Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 42-43 (1983)). Upon review, the court finds that Commerce has failed to articulate a reasoned analysis justifying the departure from its established practice of amortizing those R&D expenses related to merchandise under investigation.
To begin, Commerce contends that it has not deviated from its past practice in this case. Defendant distinguishes the prior investigations in *834which Commerce amortized R&D costs by arguing that the products at issue in those investigations did not have the same inherent technological characteristics of semiconductors; specifically, defendant focuses upon the fact that the semiconductor industry is “R&D intensive.” Defendant’s Response Brief at 7-8. This distinction is patently false. Defendant fails to acknowledge that in two previous semiconductor cases, i.e. EPROMs From Japan and 64K DRAMs From Japan, Commerce specifically endorsed capitalization and amortization of R&D expenses. Commerce makes no attempt to distinguish these cases, or to counter the language contained therein.1 As a result, the court finds defendant’s argument unavailing.
Commerce offered two justifications for expensing R&D costs in its Final Determination. First, Commerce concluded that amortization of R&D was inappropriate “because R&D costs can never be assigned to the proper product or time.” 58 Fed. Reg. at 15,472. As noted, however, Commerce has failed to establish that R&D related to non-subject merchandise provided an intrinsic benefit to the subject DRAMs. In addition, Commerce verified the product-specific R&D data submitted by respondents. Furthermore, the record in this case shows that the DRAM industry follows a fairly predictable three and one-half to four year life cycle.2 Notably, this product life cycle is approximated by the three to five year amortization period afforded under Korean GAAE Commerce was therefore incorrect in concluding that neither the proper product nor time period can be ascertained.
Second, Commerce concluded that amortization was inappropriate because, “in many instances [,] full amortization occurs prior to the introduction of the product in the marketplace.” 58 Fed. Reg. at 15,473. Initially, the court notes the absence of record evidence to support this statement. In any event, to the extent that this may be true, Commerce’s methodology is even more distortional because it guarantees that R&D costs will not be allocated over the commercial life of the product. In contrast, Korean GAAP provides for a reasonable allocation of the R&D costs associated with production of the subject DRAMs. Commerce’s argument fails to establish otherwise.
Finally, the court notes that Commerce itself acknowledged that R&D activities in one period will benefit future periods. See Final Determination, 58 Fed. Reg. at 15,472 (technological changes and improved manufacturing methods will benefit future production). Yet, Commerce proceeded to contravene this finding by expensing R&D costs in the year *835incurred. Commerce’s methodology is thus irreconcilable with the record compiled in this case.
For the foregoing reasons, the court concludes that Commerce has failed to articulate a reasoned analysis in support of its decision to abandon its prior practice of amortizing R&D costs associated with production of the investigated merchandise. The court also concludes that Commerce has failed to establish that Korean GAAP does not reasonably reflect the R&D costs related to the subject DRAMs. Accordingly, this issue is remanded with instructions that Commerce use amortized rather than current R&D expenses in its calculations. Commerce shall then amend its Final Determination accordingly.
2. Commerce’s Allocation of Interest Expense Based Upon Semiconductor Fixed Assets:
Samsung and Hyundai contend that Commerce erred in allocating interest expense based upon semiconductor fixed assets rather than the cost of goods sold. Specifically, Samsung and Hyundai argue that Commerce has consistently allocated interest expense on the basis of the consolidated cost of sales (“COS”) and has rejected allocation on the basis of fixed assets. Samsung and Hyundai further argue that Commerce has failed to provide a reasoned analysis justifying its departure from past practice, and that Commerce’s stated rationale actually supports allocation of interest expense on the basis of COS in this case.
Commerce acknowledges that it generally allocates interest expense based upon the consolidated cost of sales. In adopting a different methodology for this investigation, Commerce stated:
After reviewing the facts in this case, we have found that for Samsung and Hyundai, a larger proportion of total fixed assets are related to the semiconductor line of business than to other lines of business * * *. [B]ecause of this disproportional amount of fixed assets related to semiconductors, allocation of interest expense based on cost of sales would not appropriately recognize the expense related to the capital investment necessary for semiconductors compared to the other lines of business.
Final Determination, 58 Fed. Reg. at 15,471-72. Commerce therefore abandoned its COS methodology, and instead allocated interest expense to the subject merchandise on the basis of the ratio of the value of semiconductor fixed assets to company-wide fixed assets. Id. at 15,472. Although the COS methodology would appear to account for the capital intensive nature of the semiconductor industry in the form of increased depreciation expenses, the court cannot fault Commerce for rejecting this approach in favor of another plausible methodology that, in Commerce’s view, better accounts for this characteristic of the semiconductor industry. The choice between alternative methodologies, each of which is supported by substantial evidence and in accordance with law, is clearly left to Commerce’s discretion. Cf. Matsushita Elec. Indus. Co. v. United States, 3 Fed. Cir. (T) 44, 54, 750 F.2d 927, 936 (1984) (it is not the court’s function to decide that it would have made another decision *836on the basis of the evidence). Upon review, the court concludes that Commerce provided a reasoned analysis for rejecting its COS methodology, and that substantial record evidence supports Commerce’s decision to allocate interest expense based upon semiconductor fixed assets. Accordingly, this aspect of Commerce’s Final Determination is sustained.
The court also finds, however, that Commerce’s implementation of the methodology chosen in this case was arbitrary. Commerce’s questionnaire required respondents to allocate interest expense among their various products on the basis of COS.3 Respondents conformed their responses accordingly. Notably, Commerce did not state any change of policy in the preliminary determination, or at any time during the investigation. Indeed, Commerce verified respondents’ interest expense calculations, and did not list interest expense allocation methodology as an issue for consideration in its verification reports. The first time this issue was raised was in Micron’s pre-hearing brief before Commerce. As a result, because Commerce failed to request the appropriate data, Commerce was forced to resort to surrogate data based upon semiconductor depreciation expenses, in lieu of actual data regarding semiconductor fixed assets.
As defendant states, “[fit cannot be seriously questioned that the semiconductor industry is highly capital intensive.” Defendant’s Response Brief at 10. The basis for Commerce’s adoption of an asset-based allocation methodology was clearly evident prior to the inception of this investigation. Yet, Commerce failed to apprise respondents of its intent to depart from past practice regarding the allocation of interest expense due to the capital intensive nature of the semiconductor industry. Nor did Commerce request the requisite data for proper implementation of its alternative methodology. Instead, Commerce required respondents to submit the exact data it ultimately rejected as being inadequate for this investigation.
Commerce is presumably in the best position to know what information will satisfy its requirements. Creswell Trading Co. v. United States, 13 Fed. Cir. (T) _, _, 15 F.3d 1054, 1062 (1994). On the facts of this case, Commerce’s failure to notify plaintiffs of its intent to abandon its COS methodology and instead allocate interest expense on the basis of semiconductor fixed assets, coupled with its failure to request data regarding semiconductor fixed assets, deprived respondents of meaningful participation in the investigation as to this issue, and renders Commerce’s use of surrogate data unsupported by substantial evidence and not in accordance with law. See Usinor Sacilor v. United States, 19 CIT 711, 731, Slip Op. 95-94 at 69-70 (May 19,1995) (Commerce’s failure to request necessary information, or to notify respondents that it lacked such information, rendered its determination unsupported by substantial evidence and otherwise contrary to law); cf. Mantex, Inc. v. *837United States, 17 CIT 1385, 1399, 841 F. Supp. 1290, 1303 (1993) (court sustained Commerce’s reversal of policy in an administrative review where Commerce announced reconsideration of that policy more than two and one-half years before issuing its preliminary review, and where Commerce ultimately reversed its policy four days before issuing its preliminary review). Accordingly, the court remands this issue with instructions that Commerce reopen the record in Order to afford respondents an opportunity to present actual data regárding semiconductor fixed assets. Commerce shall then amend its Final Determination accordingly.
3. Commerce’s Identification and Correction of an Error Related to the Allocation of Hyundai’s Interest Expense:
Following publication of the Final Determination, Commerce revised Hyundai’s interest allocation in order to correct an apparent error. Specifically, Commerce concluded that Hyundai’s depreciation summary, upon which Commerce relied to estimate semiconductor fixed assets, failed to include an amount for depreciation on certain buildings which housed semiconductor production. As a result, Commerce reallocated Hyundai’s interest expense based upon the ratio of semiconductor machinery and equipment (“M&E”) to total M&E.
Hyundai objects to Commerce’s identification and correction of the alleged deficiency in its submission. As noted, however, the court remands the general issue of interest allocation in order to afford respondents an opportunity to present Commerce with actual data regarding semiconductor fixed assets. See supra part II.2. Upon remand, Commerce is directed to utilize such actual data regarding Hyundai’s semiconductor fixed assets in Commerce’s allocation of interest expense. In the event Hyundai proves unable to produce actual data, the court will address Hyundai’s objections, to Commerce’s treatment of the alleged error contained in Hyundai’s depreciation data in the court’s review of the results of remand to Commerce.
4. Commerce’s Use of a Time Lag When Comparing Constructed Value and Cost of Production to Sales:
Semicon and Hyundai challenge Commerce’s decision to use a time lag when comparing cost of production and constructed value (“CV”) to sales. Specifically, respondents allege that Commerce’s incorporation of a time lag was unnecessary and redundant because their respective cost accounting systems fully captured all historical costs. The court disagrees.
At verification, Commerce found that in the ordinary course of business Semicon’s DRAM manufacturing costs are [ ]. Confid. Doc. 174 at 12. However, for purposes of its cost response, Semi-con allocated costs upon a quarterly basis. Id. at 12-13. With regard to Hyundai, at verification Commerce found that Hyundai’s cost accounting system prepares unit costing information [ ]. Confid. Doc. 175 at 12. However, Hyundai’s cost response calculated *838product costs upon a monthly basis. Id. at 13. As a result, although respondents’ cost accounting systems may have captured all historical costs, Commerce remained unconvinced that respondents’ cost accounting properly allocated costs to specific units of output.
Commerce’s concern stems from the nature of the DRAM production process. DRAMs require approximately [ ] months to produce, from the time that the raw wafer is inserted into the process to the time that a completed, tested DRAM is ready for shipment. The cost in a given month for each stage of DRAM production is therefore not the cost of the individual DRAMs that emerge from the end of the line during that month. This distinction is important because the semiconductor industry is characterized by relatively rapidly declining per-unit costs due to increases in yield, design changes, and other efficiency improvements over time. Consequently, in order to properly match sales for a given period to the costs of manufacture related to those sales, Commerce adopted a time lag comprised of: (1) the time that completed dies were in inventory prior to being input into the assembly process; (2) the time required to assemble the dies into finished DRAMs and for testing; and (3) the average inventory period of finished DRAMs.
Upon review, the court finds that Semicon and Hyundai fail to establish that their respective cost accounting systems properly allocate costs to specific units of output. The court further concludes that Commerce’s incorporation of a time lag comprised of the three aforementioned components is supported by substantial evidence and in accordance with law. Accordingly, this aspect of Commerce’s Final Determination is sustained.
Semicon further contends that Commerce’s decision to lag costs for short inventory periods, as it did in this case, is unprecedented. Citing EPROMs From Japan, Semicon argues that even in prior semiconductor investigations in which Commerce applied a lag for work in process, Commerce refused to make an adjustment for the finished goods inventory period. EPROMs From Japan, 51 Fed. Reg. at 39,685. The court finds Semicon’s reliance upon EPROMs From Japan misplaced. Specifically, the court notes that the sole reason why Commerce did not apply a lag for the finished goods inventory period in EPROMs From Japan was because Commerce concluded that there was insufficient evidence on the record to support such an adjustment. Id. In contrast, the record in this case fully supports Commerce’s decision to include a lag for the finished goods inventory period. See, e.g., Confid. Doc. 174 at 3. Accordingly, this aspect of Commerce’s Final Determination is sustained.
Hyundai raises two additional objections to this aspect of Commerce’s Final Determination. First, Hyundai argues that Commerce arbitrarily applied a [ ] lag for all of Hyundai’s sales, rather than identifying specific periods for different types of transactions, as Commerce did for the other two respondents. The court agrees. For Samsung and Semicon, Commerce first calculated the average lag time for each production stage, i.e. inventory, assembly, and test; Commerce then cal*839culated a lag period for each type of transaction under consideration, i.e. COE] CV-ESP (exporter’s sales price), and CV-PP (purchase price). Con-fid. Doc. 203 at 5; Confid. Doc. 204 at 6-7. Commerce fails to explain why it employed a different methodology to calculate lag periods for Hyundai. Indeed, defendant makes no attempt to justify the disparate treatment accorded Hyundai with regard to this issue. See Defendant’s Response Brief at 10-14. Moreover, the court can discern no basis in the record to support Commerce’s adoption of one methodology for Samsung and Semicon, but an alternative methodology for Hyundai. See Confid. Doc. 203 at 5, ¶ 7 (indicating Commerce intended to be consistent between each respondent with regard to Commerce’s adoption of a time lag). For the foregoing reasons, the court finds that Commerce’s decision to employ a [ ] lag for all of Hyundai’s sales is not supported by substantial evidence in the record and is otherwise not in accordance with law. Accordingly, the court remands this issue with instructions that Commerce recalculate Hyundai’s lag periods utilizing the same methodology that Commerce employed for Samsung and Semicon.
Second, Hyundai points to 19 U.S.C. § 1677b(e)(l)(A) (1988), which provides that constructed value shall include the cost of materials and fabrication “at a time preceding the date of exportation of the merchandise under consideration which would ordinarily permit the production of that particular merchandise in the ordinary course ofbusiness. ” With regard to production of model HY514400, Commerce compared the prices for [ ] units sold in the United States in [ ] with the constructed value of the first [ ] DRAMs produced in [ ]. In light of this comparison, Hyundai argues that Commerce’s methodology is inconsistent with § 1677b(e)(l)(A), at least as it pertains to model HY514400. The court agrees. Commerce has failed to establish that Hyundai’s production line for model HY514400 was sufficiently established in [ ] to permit production of [ ] units in the ordinary course of business. With regard to model HY514400, therefore, Commerce is specifically directed upon remand to determine a new lag which accurately matches costs to the sales in question.
5. Commerce’s Decision to Recognize and Expense Translation Losses in Calculating Cost of Production:
Hyundai, Samsung and Semicon each challenge Commerce’s decision to include all foreign exchange translation gains and losses during the POI in its calculation of the general and administrative (“G&A”) expense component of COE Foreign exchange translation gains and losses are unrealized adjustments based upon the conversion of outstanding foreign currency monetary assets and liabilities into domestic currency (i.e. the Korean won, in this case), for purposes of inclusion on a company’s balance sheet. In contrast, foreign exchange transaction *840gains and losses are actual gains and losses experienced with respect to payments made during the POI resulting from exchange rate fluctuation between the date of transaction and the date of payment. In accordance with Korean GAAfl with respect to foreign exchange translation losses in excess of five percent of the capital stock of the respective company, respondents amortized such losses over a three to five year period. Commerce, however, rejected Korean GAAP and instead expensed all translation losses related to the POI, including losses in excess of five percent of capital stock.
Respondents first contend that translation losses are hypothetical losses only, which in no way represent actual costs of production; accordingly, respondents argue that Commerce erred by including translation losses in its COP calculations. The court disagrees. Although translation losses are unrealized, as there is no actual outflow of funds from the company, the resulting exposure to increased liability for borrowed funds caused by fluctuations in the exchange rate is by no means hypothetical. Admittedly, fluctuations in the exchange rate that occur subsequent to the POI may affect the magnitude of translation losses measured during the POI; indeed, subsequent fluctuations may eliminate translation losses entirely, such that the company may eventually recognize a transaction gain at the time the underlying liability is extinguished. Notwithstanding the contingent nature of translation losses, however, such losses are akin to an increased cost of borrowing funds that should be included in any reasonable measure of the cost climate faced by the company during the POI. Cf. Usinor Sacilor v. United States, 19 CIT 711, Slip Op. 95-94 at 14-15 (contingent nature of liabilities on loan instruments did not alter fact that such instruments were debt upon issuance and remained so until their conversion into common stock) (citation omitted). Thus, to the extent that respondents’ translation losses resulted from debt associated with production of the subject merchandise, such losses are a legitimate component of COP Significantly, Korean GAAP reflects as much by requiring that translation losses be recognized in the year in which they are incurred, to the extent such losses do not exceed five percent of the company’s capital stock. For the foregoing reasons, the court sustains Commerce’s decision to include translation losses in its COP calculations.
Second, respondents argue that Commerce has failed to establish that the companies’ translation losses are related to production of the subject merchandise. With regard to Hyundai, Commerce’s cost verification report notes that company officials indicated that all translation losses were related to borrowings for fab lines. Hyundai fails to point to any record evidence to the contrary. Accordingly, Commerce’s determination that Hyundai’s translation losses are related to production of the subject merchandise is sustained. With regard to Samsung and Semicon, however, it is unclear whether substantial evidence on the record supports the conclusion that the companies’ translation losses are related to DRAM production. Accordingly, the court remands this *841issue to Commerce. Upon remand, Commerce is directed to identify what evidence on the record supports the conclusion that the translation losses of Samsung and Semicon are related to production of the subject merchandise. Should Commerce determine that there is insufficient evidence on the record to support such a conclusion, Commerce shall exclude translation losses from the calculation of COP for Samsung and Semicon.
Third, respondents contend that Commerce erred by expensing translation losses which the companies had deferred to future years pursuant to Korean GAAP As noted, in determining whether merchandise has been sold at less than cost, Commerce will employ the GAAP of the home market of the country of exportation if Commerce is satisfied that such accounting principles reasonably reflect the cost of producing the subject merchandise. Laclede Steel Co. v. United States, 18 CIT 965, Slip Op. 94-160 at 21-22 (Oct. 12,1994) (citationomitted). Respondents argue that Commerce failed to demonstrate that Korean GAAP distorts actual costs. The court disagrees. Because translation losses relate directly to events occurring during the POI, they should not be deferred to future periods. The mere size of the loss does not alter the principle that the loss should be related to the period in which the exchange rate fluctuates. As Commerce recognized, amortization of translation losses under Korean GAAP is distortional because, if deferred, such losses “would not be appropriately matched to the sales of the company during the POI.” Final Determination, 58 Fed. Reg. at 15,471. Accordingly, the court sustains Commerce’s decision to reject Korean GAAP and instead expense all translation losses related to the POI.
6. Commerce’s Decision to Use BIA to Adjust Samsung’s Reported Depreciation Expense:
Samsung challenges Commerce’s decision to use best information available (“BIA”) to adjust Samsung’s reported depreciation expense. Commerce may resort to the best information otherwise available whenever a party refuses or is unable to produce requested information, in the requisite form, in a timely fashion. 19 U.S.C. § 1677e(c) (1988). Prior to this investigation, Samsung changed its depreciation method from the double declining balance method to the straight line method. Samsung based its questionnaire response upon its new depreciation method. However, in changing to the straight line method, Samsung did not adjust its net asset value to reflect the effect of the change. Instead, Samsung simply used the net book value of the asset as of the date of the change in method as the basis for future depreciation. Furthermore, Samsung based future depreciation upon the total useful life of the asset instead of the remaining useful life of the asset at the time of its change in depreciation method. For example, if the method of depreciation, was changed for an asset having a five-year useful life after the third year of that asset’s life, Samsung would extend the life of the asset by another five years instead of taking depreciation for the last two years of the *842asset’s original useful life. In light of these findings, Commerce concluded that the manner in which Samsung implemented its change in the method of depreciation distorted the amount of depreciation for the period of investigation. The court agrees.
First, Samsung contends that it reported its depreciation expense in a manner consistent with both internal accounting procedures and Korean GAAP In support, Samsung directs the court to Article 111(2) of Korean GAAP which provides that “changes in accounting treatments and accounting estimates should only take effect starting with the year of change.” Confid. Doc. 187, App..Tab 40. Samsung therefore argues that it complied with Korean GAAP by not restating the value of its assets to reflect the effect of the change in method of depreciation. The court notes that Article 111(2) does not appear to preclude Samsung from adjusting the net book value of an asset when changing the method of depreciation. In any event, assuming Samsung is correct as to this issue, Samsung fails to establish that its decision to use the full useful life of the asset as the basis of depreciation, rather than the remaining useful life at the time of the change in depreciation method, is in accordance with Korean GAAP; indeed, Samsung fails to direct the court to any provision in Korean GAAP which sanctions such a result. Moreover, the court finds that Commerce was entirely justified in concluding that Samsung’s methodology, as implemented, distorted depreciation expense during the POI to the extent that Samsung used the full useful life of the asset rather than the remaining useful life at the time of the change in depreciation method. Samsung’s hybrid approach takes depreciation expense which should properly be recognized during the POI and shifts it both to previous and subsequent years. See Defendant’s Response Brief at 21-22. For the foregoing reasons, the court concludes that Commerce properly used BIA in order to adjust Samsung’s reported depreciation expense. Accordingly, this aspect of Commerce’s Final Determination is sustained.
7. Commerce’s Choice of Best Information Available for Calculating Samsung’s Depreciation Expenses:
Micron argues that Commerce’s choice of BIA for the calculation of Samsung’s depreciation expenses is not in accordance with law. More specifically, Micron asserts that Samsung failed to report that it had altered its depreciation methodology in a manner that distorted the depreciation of its assets. Therefore, according to Micron, Commerce should have adjusted Samsung’s depreciation information using information submitted by Micron, rather than relying on the information submitted by Samsung. The court disagrees.
Commerce enjoys discretion in determining what constitutes BIA in a particular case. Chemical Prods. Corp. v. United States, 10 CIT 626, 633, 645 F. Supp. 289, 295 (1986). Commerce does not necessarily have to use data supplied by plaintiff as BIA; indeed, Commerce may readily use respondents’ data as BIA. Timken Co. v. United States, 11 CIT 786, *843788-89, 673 F. Supp. 495, 500 (1987); Chemical Prods., 10 CIT at 633, 645 F. Supp. at 295. As long as substantial record evidence supports Commerce’s choice of BIA, Commerce’s choice will be sustained. Seattle Marine Fishing Supply Co. v. United States, 12 CIT 60, 71, 679 F. Supp. 1119, 1128 (1988).
Upon review, the court finds that substantial evidence supports Commerce’s decision to use information submitted by Samsung as BIA after Samsung failed to produce requested depreciation information in the requisite form. First, it appears that Commerce verified the information used in Samsung’s calculations. Confid. Doc. 203 at 2; Confid. Doc. 228, Exhibit 2. Second, Commerce adjusted Samsung’s figures to account only for depreciation during the period of investigation. Final Determination, 58 Fed. Reg. at 15,479-80; Confid. Doc. 203 at 2. Third, Commerce had reason to believe that the information submitted by Samsung, as adjusted, was more accurate than the information submitted by Micron; specifically, the calculations submitted by Samsung were based on assets used to produce DRAMs, whereas the calculations submitted by Micron were based on all of Samsung’s assets. Confid. Doc. 203 at 2; Confid. Doc. 228, Exhibit 2; Confid. Doc. 186, Attachment 1. Thus, because substantial evidence supports Commerce’s choice of BIA, the court sustains this aspect of Commerce’s Final Determination.
8. Commerce’s Decision to Deduct U.S. Direct Selling Expenses from Hyundai’s Exporter’s Sales Price:
Hyundai argues that Commerce’s decision to deduct U.S. direct selling expenses from Hyundai’s exporter’s sales price (“ESP”), instead of adding such expenses to foreign market value (“FMV”), is contrary to law. This issue has been resolved by the U.S. Court of Appeals for the Federal Circuit (“CAFC”). Koyo Seiko Co. v. United States, 12 Fed. Cir. (T) _, 36 F.3d 1565 (1994). The CAFC has held that neither the plain language of 19 U.S.C. § 1677a(e)(2) (1988) nor its legislative history preclude Commerce from adjusting ESP by deducting all selling expenses, both direct and indirect, incurred in the U.S. market. Id., 36 F.3d at 1573. Finding Commerce’s interpretation of the statute to be reasonable, the CAFC has upheld Commerce’s decision to adjust ESP to account for U.S. direct selling expenses. Id., 36 F.3d at 1573-75. It is thus settled that Commerce’s methodology is in accordance with law. SKF USA Inc. v. United States, 19 CIT 625, 631, Slip Op. 95-80 at 14 (May 2, 1995). Accordingly, this aspect of Commerce’s Final Determination is sustained.
9. Commerce’s Decision to Reclassify Semicon’s Capitalized Costs of Facility Construction and Testing:
Semicon challenges Commerce’s decision to reclassify as costs of production certain capitalized costs of facility construction and testing incurred by Semicon. Specifically, during [ ], Semicon began testing [ ]. Semicon deemed commercial production to have begun [ ]. At the same *844time, Semicon was constructing [ ]. Testing [ ]. Semicon deemed commercial production to have begun [ ]. In accordance with Korean GAAP and internal accountingprocedures, Semicon capitalized these construction and testing costs as construction in progress (“CIP”) and deferred R&D costs, respectively. Commerce verified Semicon’s production quantities and recognized that Semicon capitalized the costs at issue based upon audited financial statements. Final Determination, 58 Fed. Reg. at 15,475. Although Commerce traced the costs capitalized as CIP and R&D expenses to company documentation, Commerce disagreed that these costs should be capitalized. Id. Commerce stated: “These costs are more appropriately identified as current costs of production because they include the component costs of manufacture, i.e., materials, labor, and overhead, which should be expensed as incurred. Therefore, the Department reclassified the manufacturing costs capitalized as CIP and R&D to current costs of production.” Id.
Semicon argues that Commerce has failed to establish that Semicon’s treatment of the costs at issue pursuant to Korean GAAP is distortional. Commerce contends that “it would be distortive to exclude any manufacturing costs incurred during test production when figuring a per unit cost [ ]. ” Confid. Doc. 204 at 1. Commerce further states that “ [a] 11 costs incurred in trying to perfect and optimize [Semicon’s] [ ] are relevant current expenses in determiningper unit production costs at time of commercial production (i.e., the point in time [Semicon] feels it has achieved a desired level of efficiency in production).” Id. In short, Commerce does not challenge where Semicon drew the line between test production and commercial production; rather, Commerce refuses to acknowledge any distinction between test production and commercial production for purposes of calculating a per unit cost [ ]. Upon review, the court concludes that Commerce erred in refusing to recognize this distinction.
To the extent test production and related construction provide a benefit to current and future production, such costs are properly capitalized and amortized over the periods in which these benefits accrue. Cf. part II. 1 (finding that Commerce has failed to articulate a reasoned analysis justifying the departure from its established practice of amortizing those R&D expenses related to merchandise under investigation). Indeed, it is Commerce’s methodology which distorts costs by allocating all construction, installation, and testing costs to current period production. In sum, Semicon accounted for testing costs as R&D and amortized such costs accordingly. Semicon also accounted for related construction by transferring such costs [ ] and depreciating these amounts over the useful life of the related equipment. Commerce has-failed to establish that this treatment, which is in accordance with Korean GAAP and Semicon’s internal accounting procedures, significantly distorts costs. The court therefore concludes that Commerce’s decision to reclassify as costs of production Semicon’s capitalized costs *845of facility construction and testing is unsupported by substantial evidence and otherwise not in accordance with law.
Furthermore, the court finds that substantial record evidence supports Semicon’s determination of when it began commercial production of [ ] DRAMs. Specifically, the court notes that [ ]. In addition, [ ]. The court therefore agrees with Semicon’s description of the volume of test production of [ ] DRAMs as being de minimis. For all of the foregoing reasons, the court remands this issue to Commerce with instructions that Commerce calculate Semicon’s production costs for [ ] DRAMs without reclassifying as costs of production Semicon’s capitalized costs of facility construction and testing.
10. Commerce’s Inclusion of a Small Volume of “Off-Spec” Sales in Its Calculations for Semicon:
Semicon claims that Commerce has engaged in a longstanding practice of excluding categories of sales from investigations when they involve only a small volume of sales. Semicon further claims that Commerce inexplicably departed from its longstanding practice when it decided to include a small quantity ofU.S. sales of “off-spec”4 DRAMs in the calculation of Semicon’s dumping margin. Hence, Semicon argues that Commerce’s decision to include off-spec DRAM sales in its calculations is not supported by substantial evidence on the record and is otherwise not in accordance with law. The court disagrees.
In general, Commerce does not exclude any U.S. sales from its calculation ofU.S. price (“USP”). Granular Polytetrafluoroethylene Resin from Japan, 58 Fed. Reg. 50,343, 50,345 (Sept. 27, 1993) (final results of admin, review). Although Commerce has excluded small quantities of sales from its analysis in some past cases, these cases appear to be exceptional. See, e.g., Coated Groundwood Paper from Finland, 56 Fed. Reg. 56,363, 56,365 (Nov. 4, 1991) (final determination); Gray Portland Cement and Clinker from Japan, 56 Fed. Reg. 12,156, 12,165 (Mar. 22, 1991) (final determination). These cases do not establish a long standing practice of excluding categories ofU.S. sales whenever they involve a small volume of sales.
In this case, substantial evidence supports Commerce’s decision to include U.S. sales of off-spec DRAMs in its calculations for Semicon. First, no party disputes that the off-spec DRAMs fall within the scope of Commerce’s determination. Second, Semicon sold the off-spec DRAMs in the United States during the period of investigation. Semicon Sales Verification Exhibit QV-7. Third, Micron points out that Semicon continues to produce these off-spec DRAMs and to sell them in a well established American market. Final Determination, 58 Fed. Reg. at 15,474; Confid. Doc. 193 at 13-14. Because substantial evidence supports Commerce’s decision to include the sales of off-spec DRAMs in the calcula*846tion of the dumping margin applicable to Semicon, this aspect of Commerce’s Final Determination is sustained.
11. Commerce’s Treatment of Semicon’s Sales of Product 12:
Semicon objects to Commerce’s treatment of [ ] (“Product 12”). Specifically, Semicon raises three arguments against this aspect of Commerce’s Final Determination, each of which the court will consider in turn.
First, Semicon challenges Commerce’s decision to exclude Product 12 from the home market database after determining that all of Semicon’s home market sales of Product 12 were made to a related party at less than arm’s length prices. In order to determine whether home market sales to related parties were made at arm’s length, Commerce “compared the gross unit prices of sales to related and unrelated customers. ” Final Determination, 58 Fed. Reg. at 15,469. Because Semicon sold Product 12 only to a related customer in the home market, Commerce could not compare the prices of home market sales of Product 12 to related and unrelated customers. Commerce did, however, determine that Semicon’s related customer purchased other DRAM products at less than arm’s length prices. Based upon this finding, Commerce decided to exclude all sales to that customer, including sales of Product 12, despite the fact that Commerce never specifically determined that Semicon’s related party sales of Product 12 were made on a less than arm’s length basis. Semicon challenges the reasonableness of Commerce’s decision to devise an arm’s length test on a per-customer basis instead of a per-product basis.
Commerce enjoys considerable discretion in deciding whether to include related party sales in its calculation of foreign market value (“FMV”). Usinor Sacilor v. United States, 18 CIT 1155, 1159, 872 F. Supp. 1000, 1004 (1994) (“Usinor I”). Applicable regulations provide that Commerce will consider sales between related parties if it is satisfied that such sales were made at prices comparable to the prices at which the seller sold such or similar merchandise to unrelated parties. 19 C.F.R. § 353.45(a) (1992). This court will uphold the test that Commerce selects to measure whether sales to related parties were made at arm’s length prices, unless that test is shown to be unreasonable. Usinor I, 18 CIT at 1159, 872 F. Supp. at 1004.
Upon review, the court finds that Semicon has failed to demonstrate that Commerce’s customer-based arm’s length inquiry is unreasonable. In particular, Semicon has failed to point to any record evidence that tends to show Semicon’s sales of Product 12 to its related customer were made at arm’s length prices. Semicon also fails to point to record evidence which tends to undermine Commerce’s conclusion that Semi-con’s sales of other products to its related customer were made on a less than arm’s length basis. In short, nothing on the record supports Semi-con’s position that Commerce’s methodology mistreats Semicon’s sales of Product 12 to its related customer. Cf. Usinor I, 18 CIT at 1159, 872 F. *847Supp. at 1004 (holding that Commerce’s arm’s length test was reasonable given the lack of evidence showing a distortion of price comparability). Accordingly, Commerce’s methodology is sustained in this case.
Second, Semicon contends that if sales of Product 12 are to be excluded from the home market database, then U.S. sales of Product 12 should be excluded from the margin calculation as well. Semicon argues that the exclusion of Product 12 resulting from Commerce’s arm’s length test renders Product 12 a similar model not subject to identical home market matching. Semicon’s argument is entirely without merit. Commerce was required to determine a dumping margin for all reported U.S. sales. Product 12 was among the reported U.S. sales. The result of Commerce’s arm’s length inquiry does nothing to alter Commerce’s obligation in this case; it merely casts doubt upon the reliability of the pricing data for an identical product sold in the home market. There is simply no legal support for Semicon’s claim for exclusion of U.S. sales of Product 12 from the margin calculation.
Finally, Semicon asserts that even if Commerce correctly included sales of Product 12 in its dumping margin calculation, Commerce should have used price rather than constructed value (“CV”) as its basis for comparison of FMV to USP Citing 19 U.S.C. § 1677b(b) (1988), Semi-con argues that Commerce may use CV as FMV only when sales are disregarded by virtue of having been made at less than the cost of production. Because Product 12 did not fail the cost test, but was instead excluded on the basis of an arm’s length test applied to sales of other products, Semicon argues there was no basis in law for Commerce to have used CV rather than price in determining the FMV for Product 12. The court disagrees.
19 U.S.C. § 1677b(a) (1988) provides that the price of such or similar merchandise shall form the basis for FMV if, inter alia, such merchandise is sold in the ordinary course of trade. Commerce’s conclusion that home market sales of Product 12 to Semicon’s related customer were not at arm’s length indicates that such sales were not made in the ordinary course of trade. As a result, Commerce was authorized to turn to CV pursuant to 19 U.S.C. § 1677b(a)(2) (1988); see also Cemex, S.A. v. United States, 19 CIT 587, 599, Slip Op. 95-72 at 26 (Apr. 24,1995) (citation omitted) (stating that CV should be used when Commerce has determined that an exporter’s home market prices are inadequate or unavailable). Accordingly, Commerce’s decision to resort to CV for Product 12 is in accordance with law. For all the foregoing reasons, this aspect of Commerce’s Final Determination is sustained.
12. Commerce’s Inclusion of Future Generations of DRAMs Within the Scope of Its Investigation:
Semicon contends that Commerce’s decision to include certain future generations of DRAMs within the scope of its determination is contrary to law. More specifically, Semicon claims that it was improper for Commerce to include as yet unidentified DRAMs with densities of greater *848than 16M within the scope of its determination. According to Semicon, Commerce must wait for DRAMs to come into existence and then apply certain criteria set forth at 19 U.S.C. § 1677j(d)(l) (1988) to determine whether such DRAMs fall into the same class or kind as DRAMs covered by the determination. The court disagrees.
Semicon’s argument fails because the statute on which it relies, 19 U.S.C. § 1677j(d)(l) (1988), sets forth criteria for determining whether later developed merchandise falls within the scope of an outstanding antidumping or countervailing duty order. Congress specifically intended for this statute to prevent exporters from evading outstanding orders by creating new generations of merchandise. S. Rep. No. 71, 100th Cong., 1st Sess. 101 (1987). Congress did not, however, intend for the statute to decrease Commerce’s authority to make scope decisions. S. Rep. No. 576, 100th Cong., 2nd Sess. 601 (1988).'
Commerce enjoys the authority to exercise broad discretion in defining the scope of an antidumping order. Mitsubishi Elec. Corp. v. United States, 8 Fed. Cir. (T) 45, 51, 898 F.2d 1577, 1582-83 (1990). Commerce must exercise its discretion in light of all of the facts before it, and in so doing, Commerce must seek to issue an order that will effectuate the purpose of the antidumping laws and remedy the violations found. Id. In past cases, Commerce did this by defining the scope of its antidumping orders so as to include future generations of merchandise. 3.5” Micro-disks and Coated Media Thereof From Japan, 54 Fed. Reg. 6433, 6437 (Feb. 10,1989) (final determination); Dynamic Random Access Memory Semiconductors of 256 Kilobits and Above From Japan, 51 Fed. Reg. 9475 (Mar. 19,1986) (preliminary determination). In one of these cases, Commerce noted that respondents had failed to provide sufficient evidence to indicate that future generations would fall into a different class than existing merchandise. 3.5” Microdisks and Coated Media Thereof From Japan, 54 Fed. Reg. at 6437.
In this case, the record shows that Commerce exercised its discretion in light of all of the facts before it, and issued an order that seeks to effectuate the purpose of the antidumping laws. First, the record shows that Commerce recognized that future generations of DRAMs would likely retain the fundamental characteristics of existing DRAMs, in that they would be semiconductors with dynamic random access memory. Confid. Doc. 211 at 3. Second, the record shows that Commerce recognized that new generations of DRAMs typically evolve every three and one-half to four years; hence, an order which failed to cover future generations of DRAMs would soon become meaningless. Id. at 3-4. Finally, the record shows that Semicon and other respondents failed to produce evidence indicating that future generations of DRAMs would fall into a different class than existing merchandise.5 Because substantial evidence supports Commerce’s decision to include future generations of DRAMs *849within the scope of its determination, this aspect of Commerce’s Final Determination is sustained.
13. Commerce’s Verification of Hyundai’s Cost of Production Information:
Micron argues that Commerce’s decision to rely on Hyundai’s submitted cost of production information is not supported by substantial evidence or otherwise in accordance with law because Commerce failed to verify adequately the information submitted by Hyundai. In order to prove the inadequacy of Commerce’s verification, Micron argues that Hyundai’s submitted information cannot be traced to record evidence. The court finds that Micron’s argument lacks merit.
Commerce conducts verification in order to test the information provided by a party for completeness and accuracy. Bomont Indus. v. United States, 14 CIT 208, 209, 733 F. Supp. 1507, 1508 (1990). Ordinarily, Commerce does not test every piece of information provided by a party; rather, it conducts spot testing. Monsanto Co. v. United States, 12 CIT 937, 944, 698 F. Supp. 275, 281 (1988). As long as Commerce applies a reasonable standard for verifying submitted information, and the results of Commerce’s verification are supported by such relevant evidence as a reasonable mind might accept as adequate, the court will not impose its own standard to supersede that of Commerce. Hercules, Inc. v. United States, 11 CIT 710, 726, 673 F. Supp. 454, 469 (1987).
Because Hyundai normally maintains its records in a different form than that in which it submitted its cost of production information, Commerce chose to verify Hyundai’s submitted information by attempting to tie it to information contained in Hyundai’s ordinary business records.6 Hyundai Verification Report, Confid. Doc. 175.at 3, 13-14. Indeed, substantial record evidence shows that Commerce successfully tied submitted information to information contained in Hyundai’s business records for each major phase of DRAM production, i.e. the wafer fabrication phase; the wafer probe phase; the assembly phase; and the testing phase. Confid. Doc. 175 at 23,25,26-27. In areas in which Commerce could not verify Hyundai’s submitted information, Commerce resorted to best information available. Final Determination, 58 Fed. Reg. at 15,471 (Comment 1). Upon review, the court finds that Commerce has applied a reasonable standard to verify Hyundai’s submitted information, and that substantial evidence supports the results of Commerce’s verification. Cf. Monsanto, 12 CIT at 944, 698 F. Supp. at 281 (finding that Commerce conducted an adequate verification when it traced selected items from respondents’ questionnaire response back through the production process and found that the information provided was accurate).
*850The fact that Micron cannot completely reenact Commerce’s verification of the information submitted by Hyundai does not detract significantly from the substantial record evidence that supports Commerce’s verification. Cf. Usinor Sacilor v. United States, 19 CIT _, _, Slip Op. 95-94 at 29 n.5 & 76 (May 19,1995) (holding that the mere presence of conflicting evidence on the record did not so detract from evidence relied upon by Commerce that its determination was rendered unsupported by substantial evidence) (“Usinor II”). Micron’s counsel does not enjoy the power to act as an independent investigator, and it should not attempt to duplicate Commerce’s efforts. Bethlehem Steel Corp. v. United States, 13 CIT 617, 621, 718 F. Supp. 70, 73 (1989) (citations omitted). Indeed, it is not surprising that Micron cannot duplicate Commerce’s verification using record documents because not all documents examined at verification are normally made a part of the administrative record. Kerr-McGee Chem. Corp. v. United States, 14 CIT 344, 348 n.2, 739 F. Supp. 613, 617 n.2 (1990). As Micron has failed to establish that Commerce’s decision to rely on Hyundai’s submitted cost of production information is unsupported by substantial evidence or otherwise not in accordance with law, this aspect of Commerce’s Final Determination is sustained.
14. Commerce’s Verification of Semicon’s Cost of Production Information:
Micron also argues that Commerce erred in relying upon Semicon’s submitted cost of production information because Commerce failed to verify adequately Semicon’s cost of production response. In order to establish the inadequacy of Commerce’s verification, Micron argues that submitted information cannot be traced to record evidence. The court finds that Micron’s argument lacks merit.
Commerce chose to verify Semicon’s submitted information by trac-ingit to Semicon’s business records. Confid. Doc. 174 at 30. Indeed, substantial evidence shows that Commerce successfully traced information submitted by Semicon to Semicon’s original business records for major cost elements such as materials, labor, overhead, and administrative expenses. Id. at 30-36. In areas where Commerce found that Semicon had failed to properly quantify or value costs for its submission, Commerce recalculated or revised Semicon’s submitted information. Final Determination, 58 Fed. Reg. at 15,471 (Comment 1). Upon review, the court finds that Commerce has applied a reasonable standard to verify Semicon’s submitted cost of production information, and that substantial evidence supports the results of Commerce’s verification. See Monsanto, 12 CIT at 944, 698 F. Supp. at 281.
Micron’s inability to trace submitted information back to Semicon’s business records does not render Commerce’s verification unsupported by substantial evidence. Cf. Usinor II, 19 CIT at _, Slip Op. 95-94 at 29 n.5 & 76. As noted, petitioner’s counsel does not enjoy the power to conduct its own verification, and it would be difficult for counsel to do so *851because not all documents examined at verification are máde a part of the record. Kerr-McGee Chem. Corp., 14 CIT at 348 n.2, 739 F. Supp. at 617 n.2; Bethlehem Steel Corp., 13 CIT at 621, 718 F. Supp. at 73. As Micron has failed to establish that Commerce’s decision to rely on Semi-con’s submitted cost of production information is unsupported by substantial evidence or otherwise not in accordance with law, this aspect of Commerce’s Final Determination is sustained.
15. Commerce’s Determination That Samsung and[ ] Are Not Related Parties:
Micron argues that Commerce’s determination that Samsung and [ ] are not related parties for purposes of determining constructed value is unsupported by substantial evidence on the record and otherwise not in accordance with law. More specifically, Micron asserts that because Samsung and [ ] are part of the same chaebol, or family of companies, they are related parties, as defined by 19 U.S.C. § 1677b(e)(4)(F) (1988), The court disagrees.
In calculating the constructed value of imported merchandise, Commerce can disregard a transaction between certain related parties if the transaction price of any element of value which Commerce must consider fails to fairly reflect the usual market price of that element. 19 U.S.C. § 1677b(e) (1988). The provision at issue, 19 U.S.C. § 1677b(e)(4)(F) (1988), specifies that “related parties” include “[t]wo or more persons directly or indirectly controlling, controlled by, or under common control with, any person.” This court has.not specified what evidence indicates that two or more persons are controlling, controlled by, or under common control with, any person, for purposes of this statutory provision.7 In other contexts, however, the court has looked for evidence of relationships defined in financiar terms when determining whether two parties control each other. See Zenith Radio Corp. v. United States, 9 CIT 110, 114, 606 F. Supp. 695, 699-700 (1985), aff’d 4 Fed. Cir. (T) 44, 783 F.2d 184 (1986).8
In this case, substantial record evidence supports the conclusion that Samsung and [ ] are not directly or indirectly controlling, controlled by, or under common control with, any person. First, Samsung does not own a significant percentage of the outstanding shares of [ ], or vice versa. Samsung Cost Verification Exhibit C0724 at 41-42. Second, no member of the Samsung chae-bol owns a significant percentage of the outstanding shares of both Samsung and [ ]. Id. Further, Micron fails to point to any record evidence indicating that the two companies share *852common officers or directors. Micron does point out that in their financial statements Samsung and [ ] state that they are part of the Samsung group, and that this group includes “numerous companies under common management control.” Pub. Doc. 118 at 8; Samsung Verification Exhibit 8D4 at 8. This conflicting evidence does not, however, alter the court’s conclusion that substantial record evidence supports Commerce’s determination that the two companies are not related for purposes of § 1677b(e) (4) (F). See, e.g., Matsushita, 3 Fed. Cir. (T) at 51, 750 F.2d at 933 (1984) (citations omitted) (noting that the possibility of drawing two inconsistent conclusions from record evidence does not prevent an administrative agency’s finding from being supported by substantial evidence). For the foregoing reasons, the court sustains this aspect of Commerce’s Final Determination.
III. Conclusion
Upon review, the court finds that the following aspects of Commerce’s Final Determination are supported by substantial evidence and in accordance with law: (1) Commerce’s decision to abandon its COS methodology, and instead to allocate interest expense based upon the ratio of semiconductor fixed assets to company-wide fixed assets; (2) Commerce’s incorporation of a time lag when comparing cost of production and constructed value to sales, which was comprised of the time that completed dies were in inventory prior to being input into the assembly process, plus the time required to assemble the dies into finished DRAMs and for testing, plus the average inventory period of finished DRAMs; (3) Commerce’s decision to lag costs for short inventory periods; (4) Commerce’s decision to include translation losses in its COP calculations; (5) Commerce’s determination that Hyundai’s translation losses are related to production of the subject merchandise; (6) Commerce’s decision.to reject Korean GAAP and instead expense all translation losses related to the POI; (7) Commerce’s decision to use BIA to adjust Samsung’s reported depreciation expense; (8) Commerce’s decision to use information submitted by Samsung as BIA after Samsung failed to produce requested depreciation information in the requisite form; (9) Commerce’s decision to deduct U.S. direct selling expenses from Hyundai’s exporter’s sales price; (10) Commerce’s decision to include sales of off-spec DRAMs in calculating the dumping margin applicable to Semicon; (11) Commerce’s decision to exclude Semicon’s sales of Product 12 from the home market database; (12) Commerce’s decision to include U.S. sales of Product 12 in its margin calculations for Semicon; (13) Commerce’s decision to resort to constructed value as its basis for comparison of FMV to USP with regard to Semicon’s sales of Product 12; (14) Commerce’s decision to include certain future generations of DRAMs within the scope of its determination; (15) Commerce’s decision to rely upon Hyundai’s submitted cost of production information; (16) Commerce’s decision to rely upon Semicon’s submitted cost of production information; and (17) Commerce’s determination that Sam*853sung and [ ] are not related parties for purposes of determining constructed value. Accordingly, these aspects of Commerce’s Final Determination are sustained.
The court also finds that the following aspects of Commerce’s Final Determination are not supported by substantial evidence on the record and are not otherwise in accordance with law: (1) Commerce’s determination that R&D related to non-subject merchandise provided an intrinsic benefit to other semiconductor products, including the subject DRAMs, such that Commerce allocated R&D costs on an aggregate rather than product-specific basis; (2) Commerce’s decision to abandon its prior practice of amortizing R&D costs associated with production of the investigated merchandise; (3) Commerce’s determination that Korean GAAP does not reasonably account for R&D costs related to the subject DRAMs; (4) Commerce’s decision to use surrogate data regarding the semiconductor fixed assets of Samsung and Hyundai, in lieu of requesting actual data; (5) Commerce’s decision to employ a [ ] lag for all of Hyundai’s sales; (6) Commerce’s decision to compare the prices for [ ] units of Hyundai’s model HY514400 that were sold in the United States in [ ] with the constructed value of only [ ] units of that model that were produced in [ ]; (7) Commerce’s conclusion that the translation losses of Samsung and Semicon are related to production of the subject merchandise; and (8) Commerce’s decision to reclassify as costs of production Semicon’s capitalized costs of facility construction and testing. Consequently, these aspects of Commerce’s Final Determination are remanded for reconsideration consistent with this opinion. Finally, the court reserves decision on Hyundai’s objections to Commerce’s treatment of the alleged error contained in Hyundai’s depreciation data, pending the results of remand to Commerce. The court’s order shall enter accordingly.
Specifically, in EPROMs From Japan, Commerce stated:
The Department applied its R&D methodology developed for products which require a significant research and development effort. The Department capitalized those R&D costs specifically associated with the product * * *. These costs are capitalized and amortized over the sales during the market life of the product.
51 Fed. Reg. at 39,682. Similarly, in 64KDRAMs From Japan, Commerce stated:
The Department’s method of accounting for research and development (R&D) expenses encompassed the historic R&D for 64K DRAMs allocated over the market life of the product* * *.
51 Fed. Reg. at 15,944-45.
Confidential Document (“Confid. Doc,”) 211 at 3.
Public Document (“Pub. Doc.") 143, section D at 15.
Off-spec DRAMs are DRAMs which are not fully functional; nevertheless, these DRAMs have certain limited applications.
The court notes that Commerce expressly recognized that at such time that respondents believe they have actual evidence indicating that a future generation of DRAMs falls into a different class than existing merchandise, respondents can malee a request for exclusion of the new DRAMs from the antidumping order. Confid. Doc. 211 at 4.
More specifically, Commerce discussed Hyundai’s accounting system with Hyundai management. Hyundai Verification Report, Confid. Doc. 175 at 15. Commerce compared Hyundai’s accounting system with the methodology employed in Hyundai’s COP submission, and it tested the mathematical accuracy of the COP submission. Id. Then Commerce attempted to trace information used in the COP submission back to Hyundai’s business records. Id.
In one case the court determined that the evidence presented failed to show that two parties were related for purposes of § 1677b(e)(4)(F). Washington Red Raspberry Comm. v. United States, 11 CIT 173, 177, 657 F. Supp. 537, 540-41 (1987). That case does not aid in the disposition of this action, however, as the court did not discuss any specific factors in assessing whether the parties were related for purposes of subsection (F).
In Zenith Radio, 9 CIT at 114, 606 F. Supp. at 699-700, the court held that Commerce is not requited to investigate relationships that do not express themselves in financial terms in order to determine whether two parties in a Japanese keiretsu control each other or are under common control for purposes of 19 U.S.C. § 1677(13).