OPINION OF THE COURT
GIBBONS, Circuit Judge.We deal here with a petition by Ford Motor Co., pursuant to section 509(b)(1) of the Clean Water Act, 33 U.S.C. § 1369(b)(1) (1976), for review of the January 28, 1981 action of the United States Environmental Protection Agency (EPA) denying its petition for reconsideration of the applicability of that agency’s Electroplating Pretreatment Standards for Existing Sources, 40 C.F.R. Part 413, to integrated manufacturing facilities with combined wastestreams.1 The petition is one of a group of consolidated petitions for review directed at the EPA’s regulations applicable to the discharge of electroplating wastes to publicly operated treatment facilities. In an opinion filed simultaneously herewith, we dispose of the other consolidated petitions.2 The issues presented by this one are distinct. We deny the petition for review.
The EPA, in its general pretreatment regulations, has adopted a combined wastestream formula to adjust the discharge limit set by a categorical pretreatment standard where a wastestream containing a regulated pollutant is combined with other wastewaters prior to pretreatment by a dis-charger to a publicly owned treatment facility. 40 C.F.R. § 403.6(e) (1982). In Part II E of Judge Hunter’s opinion disposing of the other consolidated petitions for review, *57we reject challenges to the combined wastestream formula. EPA has also adopted categorical pretreatment regulations establishing numerical limits, based on BPT-level technology,3 for the electroplating point source category. 40 C.F.R. Part 413 (1982). In Parts III A and B of Judge Hunter’s opinion we reject challenges to those standards predicated upon the EPA’s methodology, and upon its consideration of the cost of segregated pretreatment of electroplating wastewaters.
When the combined wastestream formula was proposed, Ford submitted comments respecting the relationship between that formula and the electroplating pretreatment standards, as both would affect integrated manufacturing facilities which included some electroplating point sources. Ill App. 548-722. In this submission, Ford stated that “sound engineering practices dictate that combined treatment should be used at automotive plants with [wastestreams characteristic of Ford’s plants].” Ill App. 570. Ford also commented on the proposed categorical pretreatment standards for existing electroplating point sources. IV App. 769-74. After the promulgation of those categorical pretreatment regulations, on May 15, 1980, Ford filed a petition under the Administrative Procedure Act4 seeking reconsideration by EPA of the application of those regulations to integrated facilities with combined wastestreams.
Ford requested a separate industrial category for automotive manufacturing. Part II E 1 of Judge Hunter’s opinion rejects the contention that the Clean Water Act required EPA to regulate whole plants rather than operations or processes. That opinion, therefore, disposes of the principal ground on which Ford petitioned for reconsideration.
Ford also requested that rulemaking proceedings on electroplating pretreatment standards be reopened for comment because of the impact of the combined wastestream formula on manufacturers in whose plants electroplating wastewater is combined with other wastestreams. In support of the petition to reopen, Ford argued that EPA’s formula would not permit combined treatment where process effluent was mixed with other wastewaters, because the combined stream limits were technologically unattainable in many instances. Ford also contended that EPA had disregarded the economic impact of the formula on integrated facilities containing electroplating point sources. The EPA Administrator denied the petition for reconsideration.
Each agency shall give an interested person the right to petition for the issuance, amendment, or repeal of a rule.
Noting that Ford had presented technological criticisms in its comments on the proposed formula, that the agency revised the formula, and that, as adopted, it had not been finally determined at the time the petition for reconsideration was filed, the Administrator concluded that the revised formula resolved Ford’s technical objections. 46 Fed.Reg. 9476 (1981). Part II E of Judge Hunter’s opinion rejects technical challenges to the formula, and thus they are not addressed here.
Turning to Ford’s contention that the agency had failed to take into account the economic impact of the formula on integrated facilities, the Administrator wrote:
[T]he final combined wastestream formula will, in most instances, obviate the need for segregation of process wastestreams. Thus, a reconsideration of economic impact is unnecessary. Nevertheless, EPA examined the data submitted by Ford in an effort to make a rough estimate of the economic impact that would result if all integrated facilities were required to segregate their wastestreams. Since the agency believes that very few facilities will be forced to segregate, this estimate gives an exaggerated economic impact.
*58Id. at 9477. Thus what EPA did was make a worst-case assumption about the feasibility of combined wastestream treatment. The agency then obtained from Booz Allen & Hamilton, management consultants, a detailed economic analysis of the likely results of that worst case assumption.
Booz Allen & Hamilton used two data sources. The first was its 1976 national survey on captive plating operations, which was conducted in support of the economic analysis by EPA of categorical electroplating pretreatment regulations. The second data source was the Ford petition itself. Table V, Appendix 1 of Ford’s comments in support of that petition presents four cost estimates for different treatment options. IX App. 2002. EPA selected as the option most applicable to integrated facilities such as Ford’s the Ford Category C: an existing manufacturing plant having an existing process collecting system, which is required to segregate several wastes from that system and at the same time maintain plant production schedules. Ford’s estimate was that costs of segregation — the worst case assumption — would be between $500,000 and $750,000 a plant. EPA accepted this estimate, choosing for purposes of analysis a mid-range figure of $625,000.
The Ford data was shown for only one plant in which metalfinishing produced 30 percent of its process wastes, and the flow was 500,000 gallons per day. Booz Allen & Hamilton adjusted the cost figures for application to the captive plant data base to reflect the fact that in many plants the plant flow is lower or higher than 500,000 gallons, and in many the percentage of metal finishing wastes is greater or less than 30 percent of the wastewater. IX App. 2026. It was thus able to develop a series of tables showing the likely cost of segregation in plants in the captive plant data base. It then applied a series of decision rules with respect to management’s anticipated reaction to those costs in order to predict in which instance divestiture of captive electroplating operations would be likely.
The conclusion reached was a rough estimate that 56 out of 4700 plants in the data base might choose to divest inhouse electroplating rather than absorb the cost of segregating wastestreams. 46 Fed.Reg. at 9477. Booz Allen & Hamilton projected from these divestiture decisions disruption of employment of an estimated 5400 persons, representing $224 million in metalfinished goods. It also estimated that if all 56 plants closed their inhouse metalfinishing operations, 252 million gallons of metalfinishing wastes would be eliminated from the wastestreams of those plants. Finally, it estimated that the cost of segregating in all the remaining captive plants in the data base would be between $1.76 and $1.46 billion. IX App. 2032-33. The Administrator noted that the per plant cost of wastestream segregation used in its analysis might be too high, because its independent estimate was one-fifth lower than the Ford figures EPA utilized. The Administrator then wrote:
This analysis leads EPA to conclude that the impact of stream segregation, even assuming that every integrated facility would be required to segregate, and th[at] Ford’s high costs per plant are correct, is not sufficiently high to require a revised economic analysis or additional comment on this issue. This conclusion is strengthened by the existence of a combined wastestream formula that will make it unnecessary for most plants to segregate their wastestreams.
46 Fed.Reg. at 9477-78. It is clear, therefore, that in considering the petition for reconsideration EPA made a cost benefit analysis before determining that there was no need to reopen the rulemaking proceeding for further economic impact comment.
In Ford’s submissions in support of its petition for reconsideration Ford again urged, as it urged in its comments on the combined wastestream formula, that most integrated facilities would, as a matter of sound engineering practice, rely on combined rather than segregated treatment. IX App. 1991-93. Nor does Ford challenge the assumption that the cost of combined treatment will be less than that of segregated treatment. Indeed such lower cost is *59the chief thrust of its petition for reconsideration. EPA was on sound ground, therefore, in concluding that it could rely on the cost of segregated treatment in making a worst case assumption as to the economic impact of the categorical electroplating pretreatment regulations when applied to integrated facilities.
Ford points out that the Act requires EPA to consider the cost of the application of best practicable control technology currently available. 33 U.S.C. § 1314(b)(1)(B) (1976). It urges that the agency finding on economic impact is flawed because the Administrator made no separate finding that segregated treatment is BPT-level treatment for integrated facilities. It is clear to us, however, that in ruling on Ford’s petition for reconsideration — a petition addressed in relevant part to the economic impact of the categorical pretreatment regulations for electroplating — the Administrator assumed, hypothetically for the purposes of analysis, that segregated pretreatment of electroplating wastes would be the best practicable technology for integrated facilities. For the purpose of making a worst case economic analysis, that approach was well within the permissible range of agency discretion.
Ford also objects that EPA’s economic analysis depends upon its prediction, allegedly without factual support, that most plants would not be required to segregate wastestreams. We have noted above that the prediction is supported by the materials submitted by Ford in support of its petition for reconsideration. But in any event, Ford’s argument simply misstates EPA’s position. The agency concluded that even if every captive electroplating point source required segregation of electroplating wastewater, the total costs would be acceptable.
Ford also urges that EPA underestimated the cost of segregation. We find this argument unpersuasive, when compared with the contents of Ford’s petition for reconsideration and the supporting documents submitted with it. It is from this source, primarily, that EPA’s cost data was derived. Additionally, Ford takes issue with the adjustments made by Booz Allen & Hamilton to Ford’s cost estimates. The Booz Allen & Hamilton report explains the methodology by which it adjusted the Ford cost estimates for application to the data base for other captive electroplaters. Reliance on that explanation is a matter well within the area of agency expertise.
Finally, the dissent contends5 that EPA failed to take into account the incremental benefit to water quality that would flow from the investment of the difference in cost of segregated versus combined waste treatment. This contention is meritless. The agency made a cost benefit analysis when it promulgated the categorical pretreatment standards for electroplating. In Part III B 2 of Judge Hunter’s opinion we conclude that the agency did not act arbitrarily or capriciously in making that analysis of the overall effect of the standards. In considering Ford’s petition for reconsideration, EPA made the. permissible assumption that those standards would, at integrated facilities, require segregated treatment. Nothing presented in the petition for reconsideration casts any doubt upon the sufficiency of EPA’s overall cost benefit analysis. Nothing in the Clean Water Act requires that a cost benefit analysis be made industry by industry, or plant by plant, or treatment technique by treatment technique.
We conclude, therefore, that we must deny the Ford Motor Company petition to review the EPA’s denial of its petition for reconsideration of the application of the Electroplating Pretreatment Standards for Existing Sources, 40 C.F.R. Part 413, to integrated manufacturing facilities.
JAMES HUNTER, III, Circuit Judge, dissenting.The majority denies Ford Motor Co.’s petition to review the EPA’s refusal to recon*60sider its decision to apply the Electroplating Pretreatment Standards for Existing Sources, 40 C.F.R. Part 418 (1982), to integrated manufacturing firms with combined wastestreams. I respectfully dissent.
The Administrator, in my view, acted arbitrarily and capriciously by failing to consider adequately the economic cost to integrated electroplating facilities, and the resulting effluent reduction benefit, when he promulgated his 1979 combined electroplating standards. The history of those standards is instructive. The Administrator originally directed his sampling visits to segregated or easily segregable plants, J.App. at 1036, thus excluding data from integrated facilities in calculating the proposed standards, see 44 Fed.Reg. 52,602, 52,608 (1979). After the electroplating standards were promulgated, the Administrator proposed to amend the general pretreatment regulations to add a combined wastestream formula. 44 Fed.Reg. 62,264 (1979) ; see National Association of Metal Finishers v. EPA, 719 F.2d 624, 650-56 (3d Cir. 1983) [hereinafter referred to as “Metal FinishersHe also amended section 413.01 of the electroplating standards to make them inapplicable to integrated facilities until the combined wastestream formula became effective. 45 Fed.Reg. 19,246 (1980) .
On January 28, 1981 the Administrator promulgated his revised combined wastestream formula and denied Ford’s petition to reconsider the applicability of the EPA’s electroplating standards to integrated facilities. 46 Fed.Reg. 9442-43, 9476 (1982). Further, the Administrator refused to reopen the electroplating record to consider the economic impact of the standards on integrated facilities. Instead he attempted to justify the electroplating standards by undertaking to demonstrate that the costs would be justified even if all integrated facilities were forced to segregate their electroplating wastestreams.
The majority is correct in holding that the Administrator may use the cost of segregating the wastestreams at all presently integrated facilities as the measure of the best practicable control technology currently available (“BPT”). In so doing, however, the Administrator must also conduct the cost-benefit analysis mandated by the Act, see Metal Finishers, 719 F.2d at 661-66, with reference to the costs and the effluent reduction benefits of segregation. He may not bolster his position by making unsupported claims that the application of the combined wastestream formula will obviate the need for segregation at most facilities and thus result in lower costs.
Similarly, the Administrator was justified in basing his cost-benefit analysis on cost figures supplied by Ford in its petition for reconsideration, rather than conducting an independent analysis. The Administrator erred, however, in departing from the Ford estimates upon which he purports to rely without offering sufficient explanation or justification for his revised figures.
The Administrator did not adhere to the data contained in the Ford petition. Rather, he lowered dramatically the projected cost of segregating all electroplating wastestreams by altering the Ford data. See 46 Fed.Reg. 9477 (1981). Under the data that Ford submitted, a “typical plant” with a combined wastestream flow of 500,000 gallons per day, of which 30% is electroplating wastewater, must spend $500,000-$750,-000 to segregate its 150,000 gallon-per-day electroplating wastestream. See J.App. at 1984, 1998-2002. In adopting that data the Administrator picked the midpoint of the cost figures, $625,000, as the cost for the typical plant. J.App. at 2026. On the assumption that the cost of segregating the electroplating wastestream is proportional to the amount of electroplating wastewater that must be segregated,1 the Administra*61tor then set up an equation to predict the cost for plants with different total flows and different percentages of electroplating wastes. Id. at 2026, 2028. In his equation, however, the Administrator assumed that it would cost $625,000 only if electroplating comprised 45% of the plant’s combined flow of 500,000 gallons per day, rather than 30%, as suggested by Ford’s data. The Administrator explained his use of the 45% assumption thus:
Not all plants are at a 30% flow ratio between metalfinishing and other wastes. Given that the [Administrator’s] sample showed 45% to be the average split between metalfinishing and other flows that is the value to be used as the new industry norm.
J.App. at 2026. Ford admits that the Administrator can use 45% as the industry norm, but argues that he had to adjust the $625,000 cost norm accordingly. Ford notes that by using the 45% assumption the Administrator cut the segregation cost of every plant by one-third — for example, segregation costs for Ford’s “typical facility” are reduced to approximately $410,000. In other words, absent that assumption the total cost of segregation for all integrated facilities would be approximately $2.6 billion.
I am concerned about the Administrator’s failure to articulate an adequate explanation for his alteration of the Ford data. The Administrator gave a reason for his use of the 45% figure, but did not mention the consequential reduction in the $625,000 cost estimate for the typical plant. While he said that “the cost per plant of wastestream segregation presented by Ford and used in EPA’s analysis may also be too high,” 46 Fed.Reg. 9477 (1981), the Administrator based his cost estimate on the premise that “Ford’s high costs per plant are correct.” Id. By altering the estimates on which he purported to rely, and thus reducing costs by $800 million, all without logical explanation, the Administrator casts grave if not fatal doubts on his consideration of segregation costs.2
The Administrator’s efforts to compare the additional cost of segregating wastestreams with the incremental effluent reduction benefit of segregated pretreatment were also deficient.3 In denying Ford’s petition for reconsideration, the Administrator did not claim that any effluent reduction benefit would be gained by segregating all integrated facilities. It appears that he viewed the cost of segregation as an added cost of removing the 140 million pounds per year of toxic pollutants which he had predicted would be removed by compliance with the 1979 standards. It may be that some fraction of that 140 million pounds per year constitutes the incremental effluent reduction benefit from segregation and segregated treatment. No such benefit was articulated by the Administrator, however.
I am disturbed by the Administrator’s failure to mention whether any incremental effluent reduction benefit will result from the segregation of all integrated facilities. *62Although the Administrator chose to rely in his segregation cost calculation on data from Ford’s typical plant, he ignored data suggesting that the typical plant’s effluent reduction benefits from segregation were out of proportion to the costs. Given the Administrator’s obligation to consider the additional costs in relation to the incremental benefits, the segregation costs stand without effluent reduction justification. See American Iron & Steel Institute v. EPA, 526 F.2d 1027, 1048 (3d Cir.1975) {“AISI I”) (remanding where no evidence that the Administrator had considered a relevant factor); Metal Finishers, 719 F.2d at 661-66.
The Administrator’s unexplained one-third reduction in segregation costs, and his failure to claim any additional effluent reduction benefit from such segregation, are disturbing enough. Of even more concern, however, is the Administrator’s ultimate conclusion that the additional cost of $1.76 billion dollars, 56 firms and 5400 jobs “is not sufficiently high to require a revised economic analysis or additional comment on this issue.” 46 Fed.Reg. 9477 (1981). To put that statement in perspective it is necessary to consider the base to which those costs were added.
The Administrator was required to determine the best practicable technology when he promulgated the 1979 electroplating standards, considering the costs in relation to the effluent reduction benefits. The Administrator attempted to do so, and determined that BPT could remove 140 million pounds of toxic pollutants per year at the cost of $1.34-plus billion, 737 electroplating operations, and 12,584 jobs. The Administrator at that time termed the cost “significant,” and outlined the steps he had taken to reduce that economic impact. Despite those measures, he candidly admitted that “the projected economic impact of these standards [is] a major concern.” 44 Fed. Reg. 52,591 (1978).
After Ford’s petition highlighted the Administrator’s failure to include the extra costs to integrated facilities in his 1979 estimate, he attempted to calculate those additional costs. He estimated that the total capital investment to industry would increase by over 130%, the total number of jobs lost would increase by over 40% and an added 25% of the largest captive electroplating operations would close, presumably to remove the same quantity of toxic pollutants. Incredibly, the Administrator then stated that he found no need to reexamine his cost-benefit analysis or even to subject his cost estimate to public comment.
It is difficult indeed to understand how the Administrator, having promulgated standards already purporting to represent the best practicable control technology currently available, could announce that the immense costs of meeting those standards had so vastly increased, without any additional effluent reduction, and yet fail to reconsider those standards.4 The Administrator’s failure to reconsider standards that doubled in cost upon closer inspection “itself was arbitrary and capricious.... These errors cannot be ignored as minimal.” National Renderers Association v. EPA, 541 F.2d 1281, 1288 (D.C.Cir.1976); see also American Iron & Steel Institute v. EPA, 568 F.2d 284, 310 (3d Cir.1977) (“AISI II”) (“whatever it costs” is not adequate consideration). I cannot conclude that the Administrator properly determined the addi*63tional cost and the marginal effluent reduction benefit of segregation, considered the cost in relation to the benefit, and concluded that the cost was not wholly out of proportion to the benefit. Because the Administrator failed to consider those relevant factors, I would remand.
. See 46 Fed.Reg. 9476-9478 (1981) (dated Jan. 13, 1981).
. National Association of Metal Finishers v. EPA, 719 F.2d 624 (3rd Cir.1983).
. BPT-level technology is used in Judge Hunter’s opinion, and here, as a short reference to the best practicable control technology currently available, a statutory standard found in 33 U.S.C. §§ 1311(b)(1)(A), 1314(b)(1) (1976).
. 5 U.S.C. § 553(e) (1976) provides:
. Ford’s briefs do not press this argument raised by the dissent. See Ford Brief at 18-28; Reply Brief at 3-7.
. Ford attacks the Administrator’s assumption that the cost of pretreatment is proportional to the flow, claiming that the effect of fixed costs is thus ignored. See J.App. at 2026. As Ford itself notes, however, the Administrator determined that the flow was the “key driver” of costs. J.App. at 1622-24. Given that Ford’s data made no distinction between fixed and variable costs, I cannot say that the Adminis*61trator was arbitrary and capricious in making that simplifying assumption.
. EPA offers two reasons why the Administrator’s reduction of the $625,000 estimate should be considered harmless. First, it asserts that the Administrator could have picked the lowest per-plant cost provided by Ford, rather than the $625,000 figure. Brief for Respondent (No. 79-2256) at 170 & n. *. Even had the Administrator chosen the lowest per-plant cost in the cited range of $500,000-$750,000, without altering Ford’s 30% figure, segregation costs would have topped $2.1 billion. The $375,000 per-plant cost proposed by EPA’s counsel was not even considered by the Administrator. Second, EPA notes that the Administrator compared Ford’s assumptions to his own and found that no more firms would close their electroplating operations than the 56 predicted under his own assumptions. Id. at 171; see J.App. at 2030-31. Of course, that does not alter the $800 million difference between the two assumptions. In any case, such post-hoc rationalizations cannot and do not supply the explanation required of the Administrator. See Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1029-31 (D.C.Cir.1978).
. Ford cites data from a typical Ford plant showing that segregating all of its combined wastestreams and separately pretreating each segregated stream would only increase the total daily removal of metal wastes from 292 to 296 pounds, and alleges that the cost per pound is much higher. J.App. at 1994-96.
. I recognize that the Administrator may have had reasons to conclude that his cost estimate was exaggerated. He expressed his belief that the combined wastestream formula would allow most indirect dischargers to meet pretreatment standards without the necessity of segregation, and he noted that Ford’s per-plant segregation cost “may be too high.” 46 Fed.Reg. 9477-78 (1981). Perhaps the Administrator could have shown that few integrated electroplaters would need to segregate to meet the 1979 standards, or that the per-plant segregation cost was much lower than Ford’s figures. The Administrator might have been able to establish that segregation resulted in effluent reduction benefits not wholly out of proportion to his $1.76 billion estimate. The point, of course, is that the Administrator chose not to make such efforts. Instead, he based his cost estimate on complete segregation and, purportedly, on Ford’s figures. As that estimate is the only one the Administrator undertook, it is the only one we can review.