Petitioner International Harvester Company (Harvester) seeks review of an order of the Occupational Safety and Health Review Commission (OSHRC) finding it in violation of 29 C.F.R. § 1910.95(b)(1).1 Harvester urges reversal of the order on three *984alternate bases. First, it asserts that this action is barred by res judicata. Second, Harvester contends that the OSHRC erred in concluding that engineering noise controls are technologically feasible for use in the production engine test department of its Melrose Park facility. Finally, Harvester argues that the OSHRC erred in concluding that such controls are economically feasible for use in that department. We find Harvester’s arguments without merit and accordingly affirm the order of the OSHRC.
I
Harvester’s contention that this action is barred by the doctrine of res judicata is wholly without merit. We mention only several of many factors making application of that doctrine inappropriate in this case.
Res judicata is applicable only when “a court of competent jurisdiction has entered a final judgment on the merits of a cause of action . . ..” Commissioner v. Sunnen, 333 U.S. 591, 597, 68 S.Ct. 715, 719, 92 L.Ed. 898 (1948). The doctrine is similarly formulated in the context of administrative agency actions:
When an administrative agency is acting in a judicial capacity and resolves disputed issues of fact properly before it which the parties have had an adequate opportunity to litigate, the courts have not hesitated to apply res judicata to enforce repose.
U. S. v. Utah Construction & Mining Co., 384 U.S. 394, 422, 86 S.Ct. 1545, 1560, 16 L.Ed.2d 642 (1966). Application of these fundamental principles in this case requires rejection of Harvester’s argument.
Harvester’s assertion that the issue presented in this action “has been fully, finally and exhaustively litigated before the Review Commission,” Brief of Petitioner at 26, is without support in the record. In both prior actions cited by Harvester, the issue of Harvester’s compliance with noise exposure regulations was not litigated or decided on its merits because Harvester chose not to contest the citations. Neither administrative action resolved any relevant “disputed issues of fact.” Utah Construction, supra, 384 U.S. at 422, 86 S.Ct. at 1560. The Administrative Law Judge presiding over the 1972 action did not weigh the evidence to determine whether Harvester had failed to implement feasible administrative or engineering noise controls; the issue faced by the ALJ was whether to grant Harvester’s motion to withdraw its notice of contest. Secretary v. International Harvester Co., 2 O.S.A.H.R.C. 81, 90-92 (1972). The decision to grant the motion did not involve resolution of any factual issues and was not a decision on the merits; it has no res judicata effect in this case where the feasibility issue has been contested by Harvester and decided by the OSHRC on its merits after compilation of an extensive record. Cf. U. S. v. International Building Co., 345 U.S. 502, 505-06, 73 S.Ct. 807, 808-809, 97 L.Ed.2d 1182 (1953); Anderson, Clayton & Co. v. U. S., 562 F.2d 972, 992-93 (5th Cir. 1977), cert. denied, 436 U.S. 944, 98 S.Ct. 2845, 56 L.Ed.2d 785 (1978); Glimco v. Commissioner, 397 F.2d 537, 540 (7th Cir.), cert. denied, 393 U.S. 981, 89 S.Ct. 452, 21 L.Ed.2d 442 (1968).
Harvester, apparently in an effort to convince this court that its abatement program has been considered on its merits, repeatedly asserts that OSHA accepted its final position on abatement.2 This conten*985tion is wholly unsupported by the record. It is undisputed that OSHA did not explicitly accept Harvester’s position. See Transcript at 446. OSHA’s failure to respond to the March 18, 1974 letter cannot be interpreted as acceptance of Harvester’s unilateral decision with respect to engineering noise controls. Similarly, we cannot agree that OSHA’s decision to cite Harvester in 1974 for violation of 29 C.F.R. § 1910.95(a) indicates approval of Harvester’s abatement efforts. See Cedar Construction Co. v. O. S. A. H. R. C., 587 F.2d 1303, 1306 (D.C. Cir. 1978).3 Furthermore, OSHA’s position with respect to Harvester’s abatement efforts — whether approval, rejection or silence — is irrelevant for purposes of res judicata. OSHA was not acting as an adjudicative body either when it received Harvester’s letter or when it issued the 1974 citation. Res judicata applies in the administrative context when “an administrative agency is acting in a judicial capacity and resolvés disputed issues of fact . . ..” Utah Construction, supra, 384 U.S. at 422, 86 S.Ct. at 1560. OSHA has not acted in a judicial capacity with respect to Harvester’s abatement efforts; neither OSHA nor the OSHRC has resolved disputed issues of fact relevant to that issue. In this context, res judicata is inapplicable.
Harvester’s reliance on Secretary v. Georgia Power Co., 4 OSHC (BNA) 1497 (1976), and Continental Can Co. v. Marshall, 603 F.2d 590 (7th Cir. 1979), is misplaced. In both cases, the employers had contested OSHA citations for violation of noise exposure standards and had prevailed on the merits. In this case, Harvester contested neither of the prior citations, there has been no decision on the merits, and Harvester clearly did not prevail, on the merits or otherwise. In addition, the decisions in Georgia Power and Continental Can were influenced by the possibility that OSHA might harass employers through repeated citations for the same violation. See 4 OSHC (BNA) at 1497; 603 F.2d at 596-97. AS will be discussed below, the record in this case is devoid of evidence of OSHA harassment.
Application of res judicata is inappropriate in this case for the further reason that this case does not involve the same cause of action involved in either of the two prior citations. The situation presented here is analogous to that discussed in Lawlor v. National Screen Service Corp., 349 U.S. 322, 75 S.Ct. 865, 99 L.Ed. 1122 (1955). The Supreme Court, in Lawlor, held that the fact that a subsequent suit involves essentially the same course of wrongful conduct as a prior suit does not indicate that both suits are on the same cause of action:
That both suits involved “essentially the same course of wrongful conduct” is not decisive. Such a course of conduct— for example, an abatable nuisance — may frequently give rise to more than a single cause of action. And so it is here. The conduct presently complained of was all subsequent to the 1943 judgment. In addition, there are new antitrust violations alleged here — deliberately slow deliveries and tie-in sales, among others — not present in the former action. While the 1943 judgment precludes recovery on claims arising prior to its entry, it cannot be given the effect of extinguishing claims which did not even then exist and which could not possibly have been sued upon in the previous case.
349 U.S. at 327-28, 75 S.Ct. at 868 [footnotes omitted]. Harvester’s continuing failure to comply with 29 C.F.R. § 1910.-95(b)(1) is a course of wrongful conduct *986analogous to the “abatable nuisance” described in Lawlor, giving “rise to more than a single cause of action.” Res judicata is therefore inapplicable.
Finally, we note that even where the technical requirements of res judicata have been established, a court may nonetheless refuse to apply the doctrine. This court does not adhere to a rigid view of the doctrine in the administrative context:
The sound view is therefore to use the doctrine of res judicata when the reasons for it are present in full force, to modify it when modification is needed, and to reject it when the reasons against it outweigh those in its favor.
Bowen v. U. S., 570 F.2d 1311, 1321 (7th Cir. 1978), quoting 2 K. Davis, Administrative Law Treatise 548 (1958). Res judicata must yield on occasion to competing public policies. See Spilker v. Hankin, 188 F.2d 35, 38-39 (D.C. Cir. 1951); IB Moore’s Federal Practice H 0.405[11]. This case presents such public policies. Acceptance of Harvester’s res judicata argument would mean that an employer could secure immunity from enforcement of noise exposure standards by declining to contest an initial citation for violation of the standards and paying a nominal fine. Such a position would undercut the important public policies embodied in the Act and is inconsistent with the doctrine of res judicata as well. See Lawlor, supra, 349 U.S. at 329, 75 S.Ct. at 869.
Our view of the equities of this situation is not altered by Harvester’s claim that it has been the victim of OSHA harassment. This claim finds no support in the record. Prior to the citation at issue here, Harvester had been cited once for violation of 29 C.F.R. § 1910.95(b)(1) and once for violation of 29 C.F.R. § 1910.95(a). Harvester contested neither citation. Contrary to Harvester’s assertion-, the record reveals that it has failed to correct the condition which gave rise to the first citation despite the fact that approximately eight years have elapsed since it withdrew its notice of contest to that citation. Rather than harassing Harvester, OSHA has been a model of patience, cooperation and reasonableness in the face of Harvester’s continuing failure to comply with 29 C.F.R. § 1910.95(b)(1). Quite clearly, Harvester has not been the victim of the kind of administrative harassment condemned in Continental Can, supra .4
For the reasons stated above, we conclude that the present action against Harvester for violation of 29 C.F.R. § 1910.95(b)(1) is not barred by res judicata.
II
In determining whether the OSHRC erred in concluding that Harvester has failed to implement technologically feasible engineering noise controls, we must first make clear the limited scope of our review of such findings. The standard of review is found in the Act itself:
The findings of the Commission with respect to questions of fact, if supported by substantial evidence on the record considered as a whole, shall be conclusive.
29 U.S.C. § 660(a). See Marshall v. L. E. Meyers Co., 589 F.2d 270, 273 (7th Cir. 1978); RMI Co. v. Secretary of Labor, 594 F.2d 566, 570 (6th Cir. 1979). In applying this standard, we rely upon the agency’s credibility determinations:
[Ajgency credibility resolutions are essentially non-reviewable unless contradicted by “uncontrovertible documentary evidence or physical facts.”
*987Olin Construction Co. v. O.S.H.R.C., 525 F.2d 464, 467 (2d Cir. 1975) [citation omitted]. We clearly may not substitute our judgment for that of the OSHRC. See RMI Co., supra, 594 F.2d at 571.
It is important that these principles be clearly understood because Harvester’s argument on the technological feasibility issue is little more than an attempt to relitigate this dispute and to convince this court to substitute its judgment for that of the OSHRC. Harvester’s argument is primarily an attack upon the ALJ’s decision to credit the testimony of Lyle Yerges rather than that of Harvester’s witnesses. See Brief of Petitioner at 34-35, 36-37; Reply Brief at 17-18, 19 & n.6, 20. Harvester, however, has not pointed to “uncontrovertible documentary evidence or physical facts” undercutting the ALJ’s credibility resolution.5 Olin Construction, supra, 525 F.2d at 467. On the contrary, the record reveals that Yerges is an eminent acoustical engineer with extensive theoretical and practical experience relevant to the noise control problem at Harvester’s plant. Transcript at 192-99; Exhibit C-21. The ALJ was certainly justified in crediting Yerges’ testimony rather than that of Harvester’s witnesses. ALJ’s Decision at 10. As the Second Circuit noted in a closely analogous case, the ALJ’s decision to credit Yerges’ testimony is dispositive of the conflict in the evidence between his testimony and that of Harvester’s witnesses.6 See Olin Construction, supra, 525 F.2d at 466.
Yerges’ testimony clearly established that engineering noise controls are technologically feasible in Harvester’s production engine test department and that implementation of the controls would result in a substantial reduction in noise exposure for all employees in the department. In addition, the record contains Yerges’ written report,7 the Hamilton Standard report,8 and a report on test cell modernization prepared by a Harvester employee. Each of these documents provides further support for the ALJ’s conclusion. The ALJ also properly relied on evidence that Harvester employs enclosed test cells in another department of the same facility. ALJ’s Decision at 10; Transcript at 372-74.
Harvester correctly notes that the fact that general sound control techniques exist does not establish that engineering noise controls are feasible in this case. It incorrectly suggests, however, that evidence *988of successful implementation of controls by other employers in the same industry is somehow irrelevant. Such evidence was presented through Yerges’ testimony and report and is certainly relevant. Transcript at 193, 195, 209-10, 213, 228-29, 239-40; Respondent’s Exhibit 1, at 3-4. The OSHRC cases cited by Harvester in support of its argument merely required the OSHA present evidence tending to show that such controls can be feasibly implemented in the particular facility at issue. See, e. g., Great Falls Tribune Co., 5 OSHC (BNA) 1443, 1444-46 (OSHRC 1977); Love Box Co., 4 OSHC (BNA) 1138, 1141-42 (OSHRC 1976). The evidence presented in this case established not only that engineering noise controls are generally feasible and are successfully employed by others in the industry, but also that they are technologically feasible for use in the particular facility at issue here.
Our examination of the record in this case convinces us that the OSHRC’s decision is supported by substantial evidence on the record considered as a whole. Moreover, even were we to abandon the limited standard of review properly applicable here and evaluate the evidence de novo, we would be compelled to reach the same result.
Ill
Harvester’s argument on the economic feasibility issue is governed by the principles of judicial review of OSHRC determinations outlined in Part II, supra. Several other principles should also be noted. In applying the cost-benefit test articulated in Turner Co. v. Secretary of Labor, 561 F.2d 82 (7th Cir. 1977), the fact that controls are expensive does not necessarily mean they are not economically feasible. Id. at 85.9 Evaluation of the economic feasibility of proposed controls requires that all relevant cost and benefit factors be considered. Id. Furthermore, in comparing costs and benefits, benefits to employees weigh more heavily in the calculus than costs to the employer. See RMI Co., supra, 594 F.2d at 572.
Once again, a significant portion of Harvester’s argument hinges upon rejection of Yerges’ testimony and acceptance of the testimony of Harvester’s witnesses. Harvester asserts that the controls proposed by Yerges would adversely affect production, requiring construction of more cells to maintain current output levels; it also refers to the cost of production lost during installation of the controls. Yerges specifically testified that his proposal would increase productivity and that current output levels could be maintained without construction of new cells; he also testified that gradual implementation of the controls would result in no interference with production. See Transcript at 215, 218, 228-29, 237-41, 241-42. The reports in the record, referred to in Part II, supra, support his conclusions. Our comments in Part II as to the propriety of the ALJ’s reliance upon this evidence rather than the testimony offered by Harvester are equally applicable here. Furthermore, we note that the testimony Harvester cites as support for its cost estimates is not particularly convincing.10
Harvester points to its hearing conservation program as a indication of the economic infeasibility of engineering controls, citing Turner Co., supra, as support. The hearing conservation program in this case, however, is a far cry from that in Turner Co. In that case, nine employees were exposed to excess noise. The employer required that these employees wear personal protective equipment. Various types of such equipment were provided and employ*989ees were custom-fitted by a nurse. The equipment was capable of reducing noise exposure to 80 decibels or less. No employees suffered any loss of hearing nor did any complain of hearing difficulties. 561 F.2d at 84. In the instant case, approximately 80 employees are exposed to excess noise. Harvester requires that they wear personal protective equipment. • The employees themselves select the type and size of the equipment they will use. The equipment is not custom-fitted nor are employees instructed on proper use or fitting. Transcript at 148-50, 166, 175-76, 184-85. Harvester on occasion runs out of particular sizes of ear plugs. Id. at 170-71, 184. The record contains no evidence indicating the reduction in noise exposure achieved through use of the Harvester equipment. Despite use of this equipment, Harvester employees have suffered hearing loss and complain of hearing difficulties. Id. at 152-54, 157, 168-69, 177-79, 188-89. Although Harvester urges this court to give this evidence little weight, it is essentially uncontradicted in the record. In the face of evidence indicating that Harvester’s program is somewhat less than effective, the ALJ properly accorded it little weight in his evaluation of economic feasibility.
The record also indicates that the engineering controls proposed by OSHA will yield significant savings over time in terms of increased productivity, reduced maintenance expenses, energy conservation and reduced manpower requirements. Transcript at 214-15, 217-18, 228-29, 237-38, 239-40.11 Harvester argues that this evidence is irrelevant because the fact that an employer can afford engineering controls does not make them economically feasible. This evidence is extremely relevant, however, not because it shows that Harvester can afford the controls, but rather because it shows that the true economic cost of the controls may be far lower than the initial capital expenditure. It is clear that all relevant cost and benefit factors enter into determination of economic feasibility. See RMI Co., supra, 594 F.2d at 572; Turner Co., supra, 561 F.2d at 85. Just as Harvester vigorously contends that the adverse effect of engineering controls on production, if established, is an element which increases cost, so the positive effect of these controls on production, clearly established in the record, is an element which reduces cost. Harvester’s argument is without merit.
The benefits to employees from implementation of engineering noise controls will be substantial. At present, approximately 80 employees are exposed to noise levels many times the Table G-16 eight hour maximum. The use of personal protective equipment has not prevented hearing loss and hearing difficulties among these employees. Implementation of engineering controls will bring all 80 employees within permissible noise exposure levels. ALJ’s Decision at 11; Transcript at 210-12; Respondent’s Exhibit 1 at 4-5.
We have considered Harvester’s challenges to various specific factual findings of the OSHRC and find them without merit.12 On the basis of our examination of the record in this case, we conclude that the decision of the OSHRC that engineering controls are economically feasible for use in Harvester’s production engine test department is supported by substantial evidence on the record considered as a whole and is in accordance with our decision in Turner Co., supra. The evidence as to feasibility, economic and technological, is so substantial *990that no other result is possible. The observation of the Second Circuit in Olin Construction, supra, is appropriate here:
In the light of the record in this regard and the strong direct evidence of the Secretary, both oral and documentary, the preponderance of the evidence not only supports the finding but it is so overwhelming that it would be a miscarriage of justice to rule otherwise.
525 F.2d at 467. Accordingly, we affirm the order of the OSHRC.
. The regulation promulgated by the Occupational Safety and Health Administration (OSHA) provides:
When employees are subjected to sound exceeding those listed in Table G-16, feasible administrative or engineering controls shall be utilized. If such controls fail to reduce sound levels within the levels of Table G-16, personal protective equipment shall be provided and used to reduce sound levels within the levels of the table.
Table G-16 — Permissible Noise Exposures
Sound level
dBA slow
Duration per day, hours response
8 00
g 92
4 95
3 ............................ 97
2 ............................ 100
1 y2 .......................... 102
1 ............................ 105
V2 ........................... 110
14 or less..................... 115
29 C.F.R. § 1910.95(b)(1) & Table G-16 [footnote to Table G-16 omitted].
. Harvester also repeatedly asserts that it has fulfilled its duty to implement feasible engineering or administrative noise controls. Indeed, much of its res judicata argument hinges upon its contention that it has abated the violation which was the subject of the first citation. See, e. g., Brief of Petitioner at 30; Reply Brief at 10, 13. This contention is unsupported by the record. The record reveals that Harvester itself unilaterally determined that its duty had been fulfilled. It unilaterally decided that the Hamilton Standard proposal was not economically feasible and, in fact, did not even submit that proposal to OSHA for its evaluation. It also unilaterally decided not to install sound absorbing curtains. No adjudicative body has ever concurred in Harvester’s decisions and; contrary to its contention, neither has OSHA. The only decision on this issue, other than Harvester’s, is that of the OSHRC now before us, finding that Harvester failed to implement feasible engineering noise controls. As will be *985discussed in Parts II and III, infra, that decision is adequately supported by the record.
. Harvester’s attempt to distinguish Cedar Construction is unpersuasive. See Reply Brief at 8. The decision to cite Harvester for violation of 29 C.F.R. § 1910.95(a) was in no way inconsistent with Harvester’s continuing obligation to comply with 29 C.F.R. § 1910.-95(b)(1). OSHA’s failure to cite Harvester for violation of that regulation in 1974 did not grant Harvester permanent immunity from compliance with it and certainly was not a decision of an adjudicative body resolving disputed issues of fact entitled to res judicata effect. The 1974 citation cannot be interpreted as acknowledgment by OSHA that Harvester had abated the violation.
. Harvester argues that Continental Can is not distinguishable merely because it involved sixteen citations, 603 F.2d at 593 n. 3, as opposed to the three citations in this case. The number of citations, however, is certainly a relevant factor in determining whether an employer has been the victim of harassment. At any rate, Continental Can is distinguishable on far more persuasive grounds. The employer there contested the initial citation, litigated it on its merits and prevailed on the issue of economic feasibility. OSHA issued repeated citations for the same violation even though the employer had been found to be in compliance with the regulation. No adjudicative body has found Harvester in compliance with the noise exposure standard involved here; thus, OSHA has not cited Harvester repeatedly for conduct previously litigated on its merits and found lawful.
. Harvester emphasizes the fact that Yerges conducted a relatively brief inspection of its facility in connection with this proceeding, apparently in an effort to challenge his credibility. The record reveals, however, that Yerges also visited the facility in 1973 and “worked up .
a complete study of noise controlled, essentially automated, test cells for that area.” Transcript at 201. The 1973 visit was in connection with preparation of the extensive and detailed Hamilton Standard report which Harvester concedes Yerges “played a major role in preparing." Brief of Petitioner at 34. Yerges had also visited the facility in 1945. Transcript at 201--02. Harvester’s emphasis on the length of Yerges’ visit does not undercut the ALJ’s credibility ruling.
. Harvester points to three items of testimony by its witnesses which it claims were uncontroverted. See Petitioner’s Brief at 36-37; Reply Brief at 19. Each of the three cited items, however, is in direct conflict with Yerges’ testimony as well as other evidence in the record. See Transcript at 208, 212, 215, 218, 228-29, 232, 237-41. Thus, the ALJ did not ignore “wholly unrebutted” testimony but rather, on the basis of his credibility finding, chose between conflicting evidence. We cannot say that this choice was improper.
. Harvester asserts on several occasions that Yerges’ report failed to discuss maintenance costs and the effect which his proposal would have on production during and after installation. See, e. g., Reply Brief at 19. Yerges specifically addressed each of these matters in his testimony. See, e. g., Transcript at 214, 215, 228, 229, 241-42.
. It is most interesting to note that although Harvester contends the ALJ erred in concluding that it has failed to implement technologically feasible engineering controls, it concedes that the Hamilton Standard proposal is technologically feasible:
[T]he cell modernization concept set forth in the Hamilton Standard report is certainly technologically feasible, .
Brief of Petitioner at 34 n.29. Contrary to Harvester’s portrayal of the record, id. at 23-24, 34 n.29, 35, Yerges specifically declined to say that this proposal was not economically feasible. See Transcript at 226-27.
. Of course, it is also true that controls are not economically feasible merely because an employer can afford them.
. The witness through which Harvester presented evidence that Yerges’ proposal would require construction of new cells housed in a new building conceded that he had not surveyed existing space in Harvester’s plant nor had he studied in detail utilization of existing test cells. Transcript at 401-02. Other testimony suggests that existing cells are not fully utilized. Id. at 453-54. Harvester’s witness also indicated limited personal knowledge of the factual basis for the estimates he presented. See, e. g., id. at 397-99.
. Harvester’s own witness testified that the Hamilton Standard proposal, rejected by Harvester as too expensive, would yield a long term “positive increment factor.” Transcript at 446; see also id. at 204. The cell modernization proposal prepared by a Harvester employee, although not involving total cell enclosure, further supports the conclusion that substantial long term savings are likely. See Transcript at 390-93; Respondent’s Exhibit 7 at 1, 3 — 4, 6-9.
. In particular, the ALJ’s reference to 26, rather than 36, test cells is nothing more than a typographical error. ALJ’s Decision at 10. There has never been any dispute in the record or among the parties as to the correct number of test cells. Indeed, elsewhere in his decision the ALJ indicates that there are 36 test cells. Id. at 4.