OPINION AND ORDER
I. Introduction and Summary
Plaintiffs, Shanghai Foreign Trade Enterprises Co., Ltd. and Shanghai Pudong Malleable Iron Plant, challenge certain aspects of a final antidumping duty determination, and the resulting antidumping duty order, that the United States Department of Commerce (“Commerce”) issued in 2003 on imported non-malleable cast iron pipe fittings from the People’s Republic of China. Shanghai Foreign Trade Enterprises is a Chinese exporter of this merchandise, and Shanghai Pudong is a Chinese producer. Anvil International, Inc. and Ward Manufacturing, Inc., domestic producers of non-malleable cast iron pipe fittings, participated as petitioners in the antidumping investigation before Commerce and have intervened in this action in support of the position of the defendant United States. The matter is before the court on plaintiffs’ motion for judgment upon an agency record, brought under Rule 56.2 of the Rules of this Court.
In their motion, plaintiffs challenge the method by which Commerce calculated the antidumping duty rate that was applied to their exports in the administrative proceedings at issue in this case. See Notice of Antidumping Duty Order: Non-Matte-able Cast Iron Pipe Fittings from the People’s Republic of China, 68 Fed.Reg. 16,765 (April 7, 2003); Notice of Final Determination of Sales at Less Than Fair Value: Non-Malleable Cast Iron Pipe Fittings From the People’s Republic of China (“Final Determination”), 68 Fed. Reg. 7,765 (Feb. 18, 2003). As is its practice, Commerce calculated the antidump-ing duty rate using “surrogate” data from a market economy country (in this case, India) in place of data pertaining to the actual production and sale of the merchandise exported from the People’s Republic of China (“China,” or the “PRC”), which Commerce considers to be a nonmarket economy country.
Plaintiffs do not contest the selection of India as the surrogate country but instead challenge Commerce’s selection of particular surrogate data from India. Plaintiffs allege, first, that Commerce improperly relied on non-industry-specific data obtained from the Reserve Bank of India to calculate the surrogate values for selling, general and administrative expenses, factory overhead, and profit. Second, plaintiffs contend that Commerce used inappropriate surrogate data to value the cost of the foundry pig iron used as a material in manufacturing the exported non-malleable cast iron pipe fittings.
This court has jurisdiction pursuant to 28 U.S.C. § 1581(c) and 19 U.S.C. § 1516a(a)(2)(A)(i). This court grants plaintiffs’ motion and remands this matter to Commerce because the findings in Commerce’s decision are not supported by substantial evidence on the record, because that decision did not provide adequate explanations for the choices of surrogate values, and because the decision did not explain adequately the departures from Commerce’s established administrative practices.
*1341 II. Background
A Determining Normal Value of Goods Produced in a Nonmarket Economy Country
Under the antidumping laws, antidump-ing duty represents the amount by which the “normal value” of the imported merchandise that was the subject of the Commerce Department’s investigation (identified as the “subject merchandise”) exceeds the “export price” for that merchandise. 19 U.S.C. § 1673. “Normal value” usually is determined by the price for which the “foreign like product” corresponding to the subject merchandise (generally, identical or like merchandise made by the same foreign producer in the same foreign country, as determined according to 19 U.S.C. § 1677(16)) is first sold, or offered for sale, for consumption in the exporting country. 19 U.S.C. § 1677b(a)(l). “Export price” usually refers to the price at which the subject merchandise is first sold, before the date of importation into the United States, by the producer or exporter outside of the United States, to an unaffiliated purchaser. 19 U.S.C. § 1677a(a).
Because it deems China to be a non-market economy country, Commerce generally considers information on sales in China and financial information obtained from Chinese producers to be unreliable for determining, under 19 U.S.C. § 1677b(a), the normal value of the subject merchandise. Accordingly, Commerce invokes a different statutory procedure for determining normal value if the subject merchandise is exported from a nonmarket economy country.
Under the substitute procedure, Commerce calculates the normal value by determining and aggregating “surrogate values” for various “factors of production” used in producing the subject merchandise, to which it also adds an amount for general expenses and profit as well as amounts for the cost of containers, coverings, and other expenses. 19 U.S.C. § 1677b(c)(l). The factors of production include, but are not limited to, labor hours, raw materials, energy and other utilities, and representative capital cost, including depreciation. 19 U.S.C. § 1677b(c)(3). The statute requires Commerce to base its valuation of the factors of production on the “best available information regarding the values of such factors in a market economy country or countries considered appropriate by the administering authority [ie., Commerce].” 19 U.S.C. § 1677b(c)(l).
To implement the statutory directive to add amounts for “general expenses and profit,” Commerce usually calculates separate values for selling, general and administrative (“SG & A”) expenses, manufacturing overhead and profit, using ratios derived from financial statements of one or more companies that produce identical or comparable merchandise .in the surrogate country. To calculate the SG & A ratio, the Commerce practice is to divide a surrogate company’s SG & A costs by its total cost of manufacturing. See, e.g., Manganese Metal From the People’s Republic of China; Final Results of Second Antidumping Administrative Review, 64 Fed.Reg. 49,447, 49,448 (Sept. 13, 1999). For the manufacturing overhead ratio, Commerce typically divides total manufacturing overhead expenses by total direct manufacturing expenses. Id. Finally, to determine a surrogate ratio for profit, Commerce divides before-tax profit by the sum of direct expenses, manufacturing overhead and SG & A expenses. Id. These ratios are converted to percentages (“rates”) and multiplied by the surrogate values assigned by Commerce for the direct expenses, manufacturing overhead and SG & A expenses. Id.
In this investigation, Commerce determined that financial information from pro*1342ducers of identical or comparable merchandise was unavailable or unsuitable for use as surrogate data. Based on that determination, Commerce chose to calculate the ratios based on aggregated financial information compiled by the Reserve Bank of India from a survey of 1,914 Indian manufacturing companies. Using the Reserve Bank of India data, Commerce established a rate for SG & A expenses of 25.93 percent, a factory overhead rate of 20.42 percent and a profit rate of 5.51 percent.
B. Administrative Proceedings Culminating in This Litigation
Domestic producers of non-malleable cast iron pipe fittings petitioned Commerce (and concurrently, the U.S. International Trade Commission) on February 21, 2002, seeking the imposition of antidump-ing duties on non-malleable cast iron pipe fittings from the PRC. On September 25, 2002, Commerce published an affirmative preliminary dumping determination for the period of investigation from July 1, 2001 to December 31, 2001. Notice of Preliminary Determination of Sales at Less Than Fair Value and Postponement of Final Determination: Non-Malleable Cast Iron Pipe Fittings From the People’s Republic of China (‘Preliminary Determination”), 67 Fed.Reg. 60,214 (Sept. 25, 2002). Plaintiffs and another Chinese producer, Jinan Meide Casting Co. (also a respondent in the proceedings before Commerce), filed responses alleging clerical errors in the Commerce preliminary determination. In its Final Determination, Commerce acknowledged errors in the Preliminary Determination, which it corrected in the final determination but viewed as insufficient to require an amended preliminary determination. See Final Determination, 68 Fed. Reg. at 7,766. The Final Determination assigned an antidumping rate (weighted average margin) of 6.34 percent to exports of the subject merchandise by plaintiff Shanghai Foreign Trade Enterprises, 7.08 percent to subject merchandise produced by Jinan Meide Casting Co., and 75.50 percent to all other subject merchandise from China. Id. at 7,768. After the U.S. International Trade Commission notified Commerce, on March 24, 2003, of its final determination that the industry in the United States producing non-malleable cast iron pipe fittings was threatened with injury by reason of imports of the subject merchandise, Commerce issued its anti-dumping duty order.
III. Contentions of the Parties
Plaintiffs challenge two classes of surrogate values chosen by Commerce in calculating the antidumping duty rates, and specifically the 6.34 percent antidumping duty rate that Commerce assigned to merchandise produced and exported by plaintiffs. They contend that the determinations by Commerce to use these surrogate values are unsupported by substantial evidence on the administrative record or otherwise are not in accordance with law.
A. Challenge to the Use of Reserve Bank of India Data for SG & A, Overhead, and Profit
Plaintiffs contend that Commerce’s use of the Reserve Bank of India data to calculate surrogate financial ratios for SG & A expenses, overhead, and profit was improper because the record contained a better source of financial data, specifically, the financial data of Indian producers of merchandise that plaintiffs claim to be comparable to the subject merchandise. Plaintiffs submit that the consistent prior practice of Commerce, as reflected in its regulations, is to use record evidence obtained from producers of comparable merchandise in the surrogate country and that Commerce departed from this practice without adequate explanation. Plaintiffs contend that Commerce should have used data from the financial reports of Jayas-*1343wals Neco Ltd., an Indian producer of iron and steel castings including brake rotors, and Kalyani Brakes Ltd., an Indian manufacturer of ferrous and aluminum castings for brake assemblies and other automotive parts. According to plaintiffs, Commerce should have regarded these two Indian companies as producers of merchandise comparable to non-malleable cast iron pipe fittings.
Plaintiffs object to Commerce’s use of Reserve Bank of India information because that information was not obtained from Indian producers of iron castings and instead was derived from financial data of various manufacturing enterprises in India. Specifically, the source of the Reserve Bank of India data is the 1999-2000 combined income, value of production, expenditure and appropriation account for a sample of 1,914 public limited companies in India, as reported in the June 2001 Reserve Bank of India Bulletin.
Defendant United States asserts that Commerce acted within its discretion in using the Reserve Bank of India data to determine surrogate financial ratios for SG & A expenses, manufacturing overhead, and profit. While acknowledging the Commerce preference for surrogate values derived from producer-specific data pertaining to identical or comparable merchandise, defendant contends that Commerce was compelled to rely upon broader industry groupings once it had determined that the surrogate companies identified on the administrative record either were unprofitable or did not produce identical or comparable merchandise.
Defendant contends that Commerce, based on substantial evidence on the record, properly declined to use the financial data of Jayaswals Neco Ltd. because the 2000-2001 financial statement of that company, which statement corresponded to the fiscal year overlapping the period of investigation (July 1, 2001 to December 31, 2001), showed a financial loss. In the proceeding below and in previous cases, Commerce has taken the position that financial data of a company reporting a loss are not reliable for use as surrogate values in non-market economy antidumping investigations. Although the Jayaswals financial data for 1998-1999 showed a profit, Commerce rejected the use of these data because, in its view, no party provided justification for such use. Defendant maintains that Commerce was justified in rejecting the financial data of Kalyani Brakes Ltd. because, it contends, the record did not demonstrate that this company manufactured merchandise comparable to the subject merchandise.
B. Challenge to the Use of Indian Import Statistics to Value Foundry Pig Iron
Plaintiffs argue that Commerce acted improperly in assigning what they view as an aberrantly high surrogate value to foundry pig iron, a material used in producing non-malleable cast iron pipe fittings. The value Commerce used was $0,228 per kilogram, which it derived from import data published in the Monthly Statistics of the Foreign Trade of India (“Indian Import Statistics”), using the statistics corresponding to the six-month period of investigation. Plaintiffs contend that Commerce should have determined the surrogate value for pig iron according to publicly available price information from two sources in India, as placed on the record below by plaintiff Shanghai Foreign Trade and adjusted to remove the effect of domestic taxes.
Plaintiffs view the Indian Import Statistics as unrepresentative of the true pig iron price in the Indian market. They point out that the total quantity of pig iron imported into India for the six month period, according to the Indian Import Statistics, was a mere 1,132 tons and represent*1344ed, in their estimation, less than one-tenth of one percent of Indian domestic consumption. Plaintiffs estimate that total pig iron consumption in India was at least 1.5 million tons for the six-month period, based on information in the petition identifying the output of 6,000 foundries in India. The minuscule percentage indicates, according to plaintiffs, that domestic demand for pig iron in India is satisfied almost exclusively by domestic pig iron, with the result that import prices must be viewed as an unreliable indicator of the market price.
According to plaintiffs, Commerce should have followed its practice of rejecting surrogate values obtained from import data that are shoym to be aberrational. They assert that the value chosen by Commerce is 20 percent higher than the prices for pig iron reported in an Indian domestic publication of the Joint Plant Committee, the JPC Bulletin. They further argue that the prices shown in the JPC Bulletin are corroborated by those for pig iron reported weekly on IndiaInfoline.com, a privately-owned website providing financial services and economic information regarding India.
A third objection raised by plaintiffs concerns the effect of domestic internal taxes on the prices for pig iron in the Indian market. Plaintiffs assert that Commerce typically will not include domestic taxes in calculating surrogate values and further assert that relatively high domestic taxes inflate Indian domestic pig iron prices. They argue that Commerce should base its surrogate value on the Indian domestic pig iron prices established by the JPC Bulletin and IndiaInfoline.com and then adjust these prices to remove the effect of domestic taxes. When this is done, they contend, the resulting prices are $0.15 per kilogram and $0.16 per kilogram, respectively-substantially less than the $0,228 price that Commerce used in the antidumping investigation.
Defendant maintains that Commerce’s use of the Indian Import Statistics was justified and supported by substantial evidence on the record. Commerce properly rejected the use of the JPC Bulletin and IndiaInfoline.com price information, defendant contends, because neither source discloses information on the quantity of pig iron used in deriving the reported price information and because Shanghai Foreign Trade, in urging the use of this information in the investigation, did not place on the record any such quantity information. Defendant argues that given the absence of this quantity information, Commerce was justified in concluding that it had no record evidence upon which it could conclude that the price data in the JPC Bulletin and IndiaInfoline.com were derived from statistically or commercially significant quantities.
Responding to plaintiffs’ argument that during the antidumping investigation Commerce never requested the quantity information on pig iron sales from Shanghai Foreign Trade or any other respondent and never contacted JPC Bulletin or In-diaInfoline.com to request that quantity information, defendant argues that plaintiffs, in the administrative proceeding below, had the burden of developing the record by submitting factual information. Because they did not do so, according to the argument of defendant, Commerce was well within its discretion in rejecting the price information of JPC Bulletin and In-diaInfoline.com in favor of price information gathered from official Indian import statistics.
IV. Discussion
A. Standard of Review
This court must evaluate whether the challenged findings by Commerce are sup*1345ported by substantial evidence on the record or are otherwise in accordance with law. See 19 U.S.C. § 1516a(b)(l)(B)(i). Substantial evidence is “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Consolidated Edison Co. of New York v. NLRB, 305 U.S. 197, 229, 59 S.Ct. 206, 83 L.Ed. 126 (1938); Matsushita Elec. Indus. Co., Ltd. v. United States, 750 F.2d 927, 933 (Fed.Cir.1984). The standard of review for a Commerce construction of the governing statute is not relevant as none is challenged in this case.
B. Commerce’s Decision to Use Reserve Bank of India Data
In the antidumping investigation, Commerce chose to use Reserve Bank of India data to calculate the surrogate financial ratios for SG & A expenses, manufacturing overhead and profit. As discussed above, Commerce obtained those data from the 1999-2000 combined income, value of production, expenditure and appropriation account for a sample of 1,914 public limited companies in India, as reported in the June 2001 Reserve Bank of India Bulletin. Commerce made this choice after rejecting the use of data on the record that was contained in financial statements of four Indian manufacturers, Rajesh Malleables Ltd., Rico Auto Industries, Ltd., Jayaswals Neco Ltd., and Kalyani Brakes Ltd.
The choice to use the Reserve Bank of India data was a departure from the established Commerce procedure. Commerce has included in its regulations a rule under which manufacturing overhead, general expenses and profit “normally” will be valued using “information gathered from producers of identical or comparable merchandise in the surrogate country.” The rule, codified at 19 C.F.R. § 351.408(c)(4), states as follows:
Valuation of Factors of Production. For purposes of valuing the factors of production, general expenses, profit, and the cost of containers, coverings, and other expenses (referred to collectively as “factors”) under section 773(c)(1) of the Act the following rules will apply:
(4) Manufacturing overhead, general expenses, and profit. For manufacturing overhead, general expenses, and profit, the Secretary normally will use non-proprietary information gathered from producers of identical or comparable merchandise in the surrogate country.
Although the rule allows for some deviation from the prescribed procedure by including the word “normally,” the rule does not identify an alternate method or alternate source of information.
Commerce’s own characterization of 19 C.F.R. § 351.408(c)(4) is that “[wjhenever possible, the Department has used producer-specific data. Unlike industry-specific data, which tends to be broader in terms of merchandise included, product-specific data pertains directly to the subject merchandise.” Issues and Decision Memorandum for the Final Determination in the Antidumping Duty Investigation of Non-Malleable Cast Iron Pipe Fittings from the People’s Republic of China (“Issues and Decision Memorandum”) at 19, Pub. Doc. 213 (Feb. 7, 2003). The data obtained from the Reserve Bank of India does not qualify even as “industry-specific,” as it was derived from a sample of 1,914 public limited companies in India. In Yantai Oriental Juice Co. v. United States, which involved a challenge to the use of Reserve Bank of India data to calculate Indian surrogate values for manufacturing overhead, SG & A expenses and profit in an antidumping investigation concerning Chinese apple juice concentrate (“AJC”), this Court observed that the Reserve Bank of India data “appears to bear *1346little relationship to the actual costs of an Indian AJC producer.” 26 CIT -, -, Slip Op. 02-56 at 27, 2002 WL 1347018 (June 18, 2002).
At issue in this case is the administrative decision by Commerce to deviate from its general, promulgated rule-under which it would have used information gathered from producers of merchandise identical or comparable to the subject merchandise- and to use, instead, the nonspecific information compiled by the Reserve Bank of India. This court would be required to conclude that Commerce’s decision is supported by substantial evidence on the record before it could uphold the final anti-dumping determination. The court also would need to discern in the Commerce decision a “rational connection between the facts found and the choice made.” Burlington Truck Lines, Inc. v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 9 L.Ed.2d 207 (1962); Neenah Foundry Co. v. United States, 25 CIT-,-, 142 F.Supp.2d 1008, 1014 (2001). Because its decision is a departure from its practice and the rule of 19 C.F.R. § 351.408(c)(4), Commerce in this proceeding has an additional duty “to explain its departure from prior norms.” Atchison, Topeka & Santa Fe Railway Co. v. Wichita Board of Trade, 412 U.S. 800, 808, 93 S.Ct. 2367, 37 L.Ed.2d 350 (1973); Saha Thai Steel Pipe Co., Ltd. v. United States, 19 CIT 273, 279-280, 879 F.Supp. 1331, 1336-1337 (1995).
This court is unable to sustain the Commerce decision affecting manufacturing overhead, SG & A expenses, and profit. Commerce’s decision is not supported by substantial evidence on the record and does not demonstrate a rational connection between the record evidence and the decision to use the Reserve Bank of India data. Commerce also failed to explain its departure from its rule and practice to use producer-specific data in the calculation of the surrogate values. These shortcomings result generally from the conclusory way in which Commerce addressed the issue of “comparable merchandise” manufactured in the surrogate country, India.
In the investigation, respondents placed financial statements of Rajesh, Rico, Ka-lyani and Jayaswals on the record, arguing at various times that one or more of these four companies were producers of cast iron merchandise that is comparable to the subject merchandise. Commerce rejected using any of these four sets of data. The Rajesh financial data was rejected because during the period of investigation Rajesh had suffered through a long labor strike, experienced financial difficulty, and did not make a profit. No party questions that determination before this court. Commerce declined to use the Rico financial information on the ground that it could not “find any evidence demonstrating that Rico produces cast iron automobile components.” 1 Issues and Decision Memorandum, at 21. Commerce also claimed that *1347any cast-iron products represented only 1.66 percent of Rico’s raw material consumption. No party asserts in this litigation that Rico’s financial statements should have been used.
Commerce decided hot to use the financial data in the Jayaswals 1999-2000 annual report on the claim that no party argued for the use of those data. Issues and Decision Memorandum at 20. Because information on the record indicates that Jayaswals did not make a profit during that period, the record contains evidence to support that decision. However, Commerce also decided to reject the data presented in the 1998-1999 Jayaswals annual report (which showed a profit), concluding that no interested party “provided justification for using” those data. That decision, however, is unsupported by the record and in fact is contradicted by Commerce’s own findings as set forth in the Issues and Decisions Memorandum. Jinan Meide argued in the investigation for the use of financial data from Jayaswals, Rico and Kalyani in the calculation of the profit ratio and, as an alternative to the Rajesh financial data, for use of that data in the calculation of the SG & A expenses and manufacturing overhead ratios as well. Jinan Meide, noting that Rajesh made malleable cast iron pipe fittings, argued specifically that Indian cast iron brake rotor manufacturers (ie., Jayaswals, Rico and Kalyani) produced the merchandise which was the next most comparable to the subject merchandise. Commerce spe-cifieally acknowledged that Jinan Meide advanced these arguments. Issues and Decision Memorandum at 20.
Commerce decided not to use the Kalya-ni financial information on the premise that respondent Jinan Meide did not show how Kalyani “is representative of a manufacturer that produces identical or comparable merchandise.” Id. Here too, the record contradicts the Commerce premise. Commerce itself summarized, five pages earlier in the Issues and Decision Memorandum, a detailed argument by Jinan Meide presenting the reasons why the products made by Kalyani constituted merchandise comparable to the subject merchandise. Issues and Decision Memorandum at 15 (“JMC [Jinan Meide] states that these cast iron brake rotors are made with strikingly similar materials, methods, foundry equipment, and finishing procedures as the subject merchandise. JMC contends that the similarities in the production processes of brake rotors and pipe fittings outweigh the differences in their end uses.”).2
In the investigation, the petitioners favored the use of the Reserve Bank of India data and urged Commerce to reject the use of financial data from the Indian brake rotor producers. The Issues and Decision Memorandum describes petitioners’ position that brake rotors are not comparable merchandise because they do not share the same physical characteristics (meaning size and shape) as pipe fittings and do not share the same end use. Issues and Ded *1348 sion Memorandum at 17. Missing from the document, however, is an analysis setting forth Commerce’s own findings and reasoning on this issue. The Commerce treatment of the issue presents little more than paraphrases of the contentions of the parties and the conclusory statements, contradicted by the record, that no party “provided justification” for use of the Jay-aswals information and that no party showed how the Kalyani products were comparable to the subject merchandise. Most notably, Commerce fails to discuss why merchandise made by Jayaswals and Kalyani, including in particular cast iron brake rotors, is or is not comparable to the subject merchandise.
To determine if a product produced by a company in the surrogate country is comparable, Commerce’s established practice is to apply a three-part test that examines “physical characteristics, end uses, and production processes.” Issue and Decision Memorandum at 19, citing Certain Cased Pencils from the People’s Republic of China,; Final Results and Partial Rescission of Antidumping Duty Administrative Review, 67 Fed.Reg. 48,612 (July 25, 2002) (“Pencils Final Results”) and accompanying Issues and Decision Memorandum at Comment 5. Neither the Federal Register notice announcing the Final Determination nor the Issues and Decision Memorandum provides reasons why Commerce, in this case, departed from its practice by omitting an analysis of its application of the three-part test or another such test. As a result, the Commerce decision, failing to address the record evidence concerning Indian producers of cast iron products, does not adequately explain why Commerce considered Reserve Bank of India data preferable to the company-specific data for its surrogate value analysis. The court’s understanding on this point is not furthered by the statement in the Issues and Decision Memorandum that the Reserve Bank of India information “contains a number of potentially comparable producers of pipe fittings.” Issues and Decision Memorandum at 22. This assertion, which is not further explained or justified by any reference to record evidence, seems incongruent with the generalized nature of the Reserve Bank of India data as derived from a broad sampling of Indian companies. It also invites questions concerning which publicly owned companies in India included in the Reserve Bank of India compilation are “potentially comparable producers of pipe fittings” and why Commerce did not consider using financial data from those producers for calculating SG & A expenses, manufacturing overhead and profit.
In some past cases in which Commerce has applied its three-part “comparable merchandise” test to two classes of products made using similar materials and production processes, it has found comparability despite differences in shape, size and end use. See Notice of Preliminary Determination of Sales at Less than Fair Value and Postponement of Final Determination: Lawn and Garden Fence Posts From the People’s Republic of China, 67 Fed.Reg. 72,141, 72,145 (Dec. 4, 2002) (rejecting use of Reserve Bank of India data after finding circular steel pipe to be comparable to steel fence posts because they have similar production processes and material inputs); see also Glycine from the People’s Republic of China: Final Results of New Shipper Administrative Review, 66 Fed.Reg. 8,383 (Jan. 31, 2001) and accompanying Issues and Decision Memorandum at Comment 7 (finding that similarity in production processes of glycine, a food additive, and phenylglycine, a toxic ingredient in dyes, outweighed any difference in the final end use of the products); see also Pencils Final Results and accompanying Issues and Decision Memorandum at Comment 5 (finding that wooden cabinets, doors and handicrafts were comparable to *1349pencils based on similarities in production and rejecting use of generic Reserve Bank of India data). In some cases, Commerce has given the term “comparable” an expansive interpretation. See Tapered Roller Bearings and Parts Thereof, Finished or Unfinished, From Romania: Final Results of Antidumping Duty Administrative Revietv, 62 Fed.Reg. 37,194, 37,199 (July 11, 1997) (“As the Department noted ... in defending the use of data from the Turkish pipe and tube industry, the term “comparable” encompasses a larger set or products than “such or similar.” Thus we have supported the use of pipe industry data in earlier reviews of this proceeding as being sufficiently ‘comparable’ to tapered roller bearings.”). If steel fence posts are comparable to steel pipes and pipe fittings, if a food additive is comparable to a toxic dye ingredient, if pencils are comparable to furniture, and if bearings are comparable to steel pipes, then Commerce must explain, in the context of its established practice, how cast iron pipe fittings are not comparable to cast iron brake rotors.
With regard to the Jayaswals financial statements, Commerce also apparently overlooked evidence that Jayaswals may qualify as a producer of comparable merchandise other than brake rotors. The record indicates that Jayaswals has a diversified casting business: “Jayaswals produces iron and steel castings, including drainage pipes and cylinder heads, that weigh 500 grams to 5 tonnes.” Def’s Mem. in Opp’n to Pis.’ Rule 56.2 Mot. for J. upon the Agency R. at 32 (Sept. 26, 2003) (citing 1999-2000 Jayaswals Annual Report, Pub. Doc. 1, Ex. 20). If Commerce determines that brake rotor manufacturers produce comparable - merchandise, then Jayaswals would appear to qualify as a brake rotor producer. If Commerce determines that brake rotors are not comparable merchandise, it also must consider whether Jayaswals would qualify as a producer of identical or comparable merchandise based on its larger casting business, which apparently includes pipes and other products.
In summary, Commerce’s decision not to use the data contained in the 1998-1999 Jayaswals and the 2000-2001 Kalyani financial statéments fails because it is not supported by substantial evidence. It also fails because it lacks a rational connection between its conclusion and the evidence in the record and also lacks a justification for the departure from Commerce’s rule and past practice.
Accordingly, the court remands this case to Commerce for correction of the inadequacies in its determination concerning the surrogate values for SG & A expenses, manufacturing overhead and profit. On remand, Commerce either must follow the general rule of 19 C.F.R. § 351.408(c)(4) by calculating these values using “non-proprietary information gathered from producers of identical or comparable merchandise in the surrogate country,” or it must provide an explanation sufficient to justify -its use of information that falls short of that standard. That explanation must be grounded in evidence on the record and must explain the rational connection between the record evidence and the conclusion reached. Commerce must determine whether cast iron brake rotor manufacturers produce merchandise comparable to the subject merchandise. If it concludes that they do not, then it must state its reasons for that conclusion and justify its determination that a product made with similar materials and production processes is not comparable to the subject merchandise. Even if Commerce determines that brake rotors are not comparable merchandise, then it must explain why Jayaswals, which the record indicates to have a significant iron casting business, *1350is not a producer of comparable merchandise.
C. Valuation of Pig Iron
Commerce obtained its surrogate value of $0,228 per kilogram (10.99 Rupees per kilogram) for pig iron, a primary material in the manufacturing of non-malleable cast iron pipe fittings, from Indian Import Statistics corresponding to the six-month period of investigation. As discussed previously, plaintiffs challenge this surrogate value on various grounds, alleging in particular that it is based on a quantity of pig iron, 1,132 metric tons for the six-month period, that is so small as to be statistically and commercially insignificant when viewed against the total Indian domestic consumption of pig iron.
Plaintiffs also assert that the surrogate value chosen by Commerce is substantially higher than prices for pig iron shown in two Indian domestic references for pig iron prices. During the investigation, Shanghai Foreign Trade placed on the record two such sources: the JPC Bulletin, an Indian government publication of market prices in six major cities in India, and the website IndiaInfoline.com. Plaintiffs identified a pig iron price of 9.12 Rupees/kg. (7.21 Rs/kg. excluding excise tax) based on JPC Bulletin and a price of 9.852 Rs/kg. (7.79 Rs/kg. excluding excise tax) based on IndiaInfoline.com. Mem. in Supp. of Mot. for J. on the Agency R. under Rule 56.2 filed by Pis. {“Pis. ’ Br. ”). at 25-26 (July 7, 2003). Plaintiffs urge that Commerce use these two sources, exclusive of the excise tax, to calculate the pig iron surrogate value.
The governing statute grants considerable discretion to Commerce in choosing among surrogate values for the factors of production. See, e.g., Nation Ford Chemical Co. v. United States, 166 F.3d 1373, 1377 (Fed.Cir.1999). Nevertheless,the statute requires that “the valuation of the factors of production shall be based on the best available information regarding the values of such factors in a market economy country or countries” that Commerce considers “appropriate.” 19 U.S.C. § 1677b(c). In addition, it is Commerce’s duty to ensure that the antidumping rates are as accurate as possible. See Rhone Poulenc, Inc. v. United States, 899 F.2d 1185, 1191 (Fed.Cir.1990).
Consistent with the statutory mandate to use the best available information, Commerce must evaluate all data in the record to determine reliability. See Olympia Industrial, Inc. v. United States, 22 CIT 387, 390, 7 F.Supp.2d 997, 1001 (1998) (“Commerce has an obligation to review all data and then determine what constitutes the best available information or, alternatively, to explain why a particular data set is not methodologically reliable.”). In fulfilling this duty, Commerce’s practice is to discard as unreliable proposed surrogate market values that are aberrational compared to other market values on the record. See Pencils Final Results, 67 Fed. Reg. 48,612, and accompanying Issues and Decision Memorandum at Comment 4 (citing Heavy Forged Hand Tools, Finished or Unfinished, With or Without Handles, from the People’s Republic of China; Final Results of Antidumping Duty Administrative Reviews, 60 Fed.Reg. 49,251, 49,-253 (Sept. 22, 1995) {“Hand Tools Final Results”)).
Commerce has a preference for using import statistics to value material inputs because they are “publicly available published information” and do not include domestic taxes or subsidies. See Hand Tools Final Results, 60 Fed.Reg. at 49,252. However, if the import statistics are based on a small quantity of imports for the period of investigation, the Commerce practice is to determine if the’ price for those imports is aberrational. See Shakeproof Assembly Components Div. of Ill. Tool Works, Inc. v. United States, 23 CIT *1351479, 485, 59 F.Supp.2d 1354, 1360 (1999). If the price is aberrational, Commerce will consider the statistics unreliable and use a different source. See Final Determination of Sales at Less Than Fair Value: Certain Cuh-to-Length Carbon Steel Plate From the People’s Republic of China, 62 Fed.Reg. 61,964, 61,981 (Nov. 20, 1997) (“For pig iron, we were unable to use the Indian Monthly Statistics as we determined that the import price was aberrational because the Indian data was based on a very small quantity and was almost two times the price of the Indonesian pig iron.”); see also Hand Tools Final Results, 60 Fed.Reg. at 49,253 (Commerce’s practice is to check import statistics against “sources of market value if the total quantity imported under a specific category was small, and, if the value was found to be aberrational, i.e., too high or too low, [Commerce has] chosen another surrogate value.”).
The Commerce decision that the Indian import data was the “best available information” from which to calculate a surrogate value for pig iron, as set forth in the Issues and Decision Memorandum, suffers from two shortcomings. Commerce does not “explain its departure from prior norms.” Atchison, Topeka & Santa Fe Railway Co. v. Wichita Board of Trade, 412 U.S. at 808, 93 S.Ct. 2367. Nor does Commerce present a “rational connection between the facts found and the choice made.” Burlington Truck Lines, Inc. v. United States, 371 U.S. at 168, 83 S.Ct. 239. A Commerce decision to rely on potentially aberrational data without explanation and contrary to its own practice is not based on substantial evidence and cannot be sustained. See Shakeproof Assembly Components, 59 F.Supp.2d at 1360.
Commerce’s explanation of its decision to use the Indian Import Statistics is con-clusory and inadequately supported. Commerce claimed that it was “not persuaded” to stop using the import statistics and that Shanghai Foreign Trade failed to show how the JPC Bulletin and Indialn-foline.com data “are a more accurate representation of competitive prices in the Indian market.” Issues and Decision Memorandum at 26. Commerce also indicated that JPC Bulletin and Indialnfo-line.com do not disclose the amount of pig iron sold in the period and that Commerce, therefore, had no evidence that the prices are “derived from statistically or commercially significant quantities.” Id. There is little in the decision, beyond these conclusory allegations, to support the choice to usé the import data.3 Com*1352merce’s decision to use the Indian Import Statistics suffers from the same flaw that Commerce alleges as a basis for its rejecting plaintiffs’ alternatives. The Commerce decision fails to establish that the small amount of pig iron imported by India during the period of investigation was statistically or commercially significant and demonstrates no apparent consideration of that issue. Commerce did not address the issue whether the Indian Import Statistics were based on too small a sample to be reliable. Commerce did not explain its decision to deviate from its past practice, under which it normally would ensure that a small quantity of imports did not produce a price that is aberrational relative to other sources of market value. Before Commerce can choose among various values to select the most accurate, it must, consistent with its practice, discard those that are unreliable. In this case, Commerce summarily discarded the alternatives as flawed but did not evaluate the reliability of its own choice.
The court’s examination of the record reveals indications that the 1,132 metric tons of pig iron imported into India during the period of investigation are not commercially significant. First, plaintiffs submitted for purposes of valuing the factors of production the 2000 Indonesian import statistics. See Pub. Doc. 95, Dickstein, Shapiro, Morin & Oshinsky, LLP Letter to Commerce, June 21, 2002. Those statistics show that Indonesia imported 107,542 metric tons of pig iron (excluding 30,774 metric tons from the PRC) in 2000. When divided in half to represent a six-month period equivalent to the period of investigation, this amount indicates that Indonesia imported approximately fifty times the amount of pig iron imported into India. Second, Jayaswals, which consumes and produces pig iron, produced 384,176 metric tons of pig iron in 1998, according to its annual report. See Pub. Doc. 175, O’Mel-veny & Myers Letter to Commerce, Nov. 4, 2002, Ex. 3B. The Jayaswals data indicates that the amount imported into India was one-half of one percent of half the annual amount produced by just one Indian domestic company.4
In addition to the indications on the record that the India Import Statistics were based on a commercially insignificant quantity of pig iron, the record contains indications that the price for pig iron obtained from those statistics is aberrational relative to other sources for determining market value. If an adjustment is made for the effect of excise taxes, as urged by plaintiffs, the JPC Bulletin price for pig iron is 66 percent of the Indian Import Statistics price and the IndiaInfoline.com price is 71 percent of that price. In addition, Indonesian import statistics for 2000 priced pig iron at $0.13, which constitutes only 56- percent of the Indian Import Statistics price. The court finds that Commerce failed to justify the departure from its usual practice of using import statistics only after concluding that they are based *1353on commercially and statistically significant quantities. Commerce also failed to explain its disregard of record evidence indicating that the 1,132 metric tons of pig iron imported into India during the period of investigation may be too small a quantity to support a reliable determination of market value in the surrogate country. Moreover, Commerce does not address whether the value it chose is aberrational relative to other record evidence of the market value of pig iron.5 Had Commerce considered that evidence and the evidence that the Indian Import Statistics were not based on a sufficient quantity, it then would have been in a position to make the determination the statute requires, ie., whether the value it chose was “based on the best available information.”
On remand, Commerce’s analysis must address whether the price for pig iron obtained from the Indian Import Statistics is based on a statistically or commercially insignificant quantity. To do this, Commerce must state its method for determining what is an insignificant quantity. If Commerce concludes that the quantity is insignificant, then it must determine if the Indian Import Statistics price is aberrational relative to other market-based sources for pig iron prices. Commerce must state how it determines what qualifies as an aberrational price relative to those other sources. If Commerce concludes that the value obtained from the Indian Import Statistics is unreliable because it is aberrational relative to other sources for pig iron prices, then Commerce must fulfill its statutory obligation to use the best available information by looking to other sources to value pig iron. If those alternative sources are drawn from domestic information from India, Commerce must address plaintiffs’ argument that domestic excise taxes should not be included in the pig iron price established by Commerce. If necessary, Commerce should re-open the record to establish a market value to compare to the Indian Import Statistics price or to obtain another source for valuing pig iron.
V. Conclusion and Order
Upon consideration of plaintiffs’ Rule 56.2 Motion for Judgment upon an Agency Record, plaintiffs’ briefs in support of said motion, and defendant’s and defendant-in-tervenors’ opposition thereto, upon all relevant papers and proceedings had herein, and upon due deliberation; it is hereby
ORDERED that determinations by the United States Department of Commerce (“Commerce”) in the Antidumping Duty Order: Novr-Malleable Cast Iron Pipe Fittings from the People’s Republic of China) 68 Fed.Reg. 16,765 (April 7, 2003), and the Final Determination of Sales at Less Than Fair Value: Non-Malleahle Cast Iron Pipe Fittings From the People’s Republic of China, 68 Fed.Reg. 7,765 (Feb. 18, 2003), are remanded for proceedings consistent with this opinion and order; and it is further
ORDERED that Commerce shall have ninety (90) days, until July 8, 2004, to complete and file its remand determination; plaintiffs shall have thirty (30) days from that filing to file comments, and Commerce and defendant-intervenors shall *1354have twenty (20) days after plaintiffs’ comments are filed to file any reply.