UNITED STATES of America, Plaintiff-Appellee, v. NEW ORLEANS PUBLIC SERVICE, INC., Defendant-Appellant

GARZA, Circuit Judge: I. FACTS New Orleans Public Service, Inc. (NOP-SI) is a private Louisiana corporation and public utility that produces, distributes, and sells electric power and natural gas to consumers in New Orleans. Until June 30, 1970, NOPSI had a contract with the National Aeronautics and Space Administration (NASA) that contained a nondiscrimination clause specifically limited to NOP-SI’s facilities at NASA’s Michoud Assembly Facility (MAF). The MAF facilities were not staffed by NOPSI employees. When the limited MAF contract expired NOPSI and NASA failed to agree on a new contract with an expanded nondiscrimination clause. Negotiations continued for an undetermined time period and NOPSI officials testified that the talks never concerned expansion of the nondiscrimination clause or rate changes. Although no contract agreement was reached, NOPSI continued to supply gas and electric service to NASA pursuant to its obligation as a public utility. Executive Order (E.O.) 11,246 prohibits employment discrimination by contractors with the federal government. The order requires that all covered government contracts contain a nondiscrimination clause and that contractors agree to take affirmative action to ensure the equal employment opportunity goals of the order are attained. E.O. 11,246, as amended, 3 C.F.R. § 339 (1964) assigns responsibility for assuring compliance with its terms to the Secretary of Labor, who promulgated guidelines codified at 41 C.F.R. § 60-1 et seq. (1983). When the events leading to this litigation occurred, the Labor Secretary had delegated enforcement responsibility to the Director of the Office of Federal Contract Compliance. 41 C.F.R. § 601.2 (1972). The OFCC assigned enforcement responsibility for various industries (including utilities) to the General Services Administration. At the time NOPSI was selected for an E.O. 11,246 compliance review, contractors were chosen in one of two ways. First, Edward Mitchell, Contract Compliance Officer for GSA, established a list of companies from which regional GSA officials could schedule compliance reviews from October 1970 through June 1971. Second, by applying certain criteria, Regional officers could select companies not on Mitchell’s list and then submit their proposed reviews to Washington, D.C. for approval. On March 15, 1971, Kenneth Patton, a Senior Assistant Regional Contract Compliance Officer, notified NOPSI that they had been selected for a compliance review. Patton proposed to visit NOPSI to examine their employment statistics and records. NOPSI refused to comply with this request. On March 15, 1972, GSA sought a preaward compliance review of NOPSI because the company was negotiating a government contract in excess of a million dollars. Again, NOPSI refused to cooperate with GSA. II. JUDICIAL PROCEEDINGS On May 17, 1973, the Attorney General sought injunctive relief to require NOPSI to permit access to its records and to comply with all other rules and regulations issued pursuant to E.O. 11,246. The District Court permanently enjoined the utility from failing or refusing to 1) comply with the executive order; 2) implement regulations; 3) allow the government to conduct compliance reviews of NOPSI. United *425 States v. New Orleans Public Service, Inc., 8 FEP Cases 1089 (E.D.La.1974). NOPSI appealed, arguing that the E.O. and its regulations could not be imposed on the company without its contractual consent and that the fourth amendment protected its records from government inspection. The Fifth Circuit rejected these arguments and held that the regulations were proper and did apply to NOPSI, and that the proposed inspections were properly limited in scope, but we vacated the district court’s injunction. The court required NOPSI’s compliance but allowed it to do so voluntarily through the regular GSA administrative process. United States v. New Orleans Public Service, Inc., 553 F.2d 459 (5th Cir.1977) . The Supreme Court then granted cert, and vacated the Fifth Circuit decision for reconsideration in light of Marshall v. Barlow’s Inc., 436 U.S. 307 (1978), 98 S.Ct. 1816, 56 L.Ed.2d 305 (1978). Consequently, the Fifth Circuit remanded the case to the district court for reconsideration in light of Barlow’s. United States v. New Orleans Public Service, Inc., 577 F.2d 1030 (5th Cir.1978) . The district court concluded that Barlow’s did not require a change in its earlier ruling, and entered an order authorizing the government to proceed against NOPSI through its administrative process. United States v. New Orleans Public Service, Inc., 480 F.Supp. 705 (E.D.La.1979). Again NOPSI appealed to the Fifth Circuit, which vacated and remanded. United States v. Mississippi Power & Light Co., 638 F.2d 899 (5th Cir.1981). In that decision the Fifth Circuit held that at least three elements are essential to justify the reasonableness of the proposed search under the fourth amendment. Id. at 908. These are 1) whether the proposed search is authorized by statute; 2) whether the proposed search is properly limited in scope and; 3) how the agency chose to initiate the particular search. Id. at 907. 1 The first two elements are questions of law and the last element one of fact. In an earlier opinion involving this case, the Fifth Circuit held that the first two elements had been satisfied. 553 F.2d 459. In Mississippi Power we remanded for a factual determination of the third element. In this regard the Court stated that: The search will be reasonable if based either on (1) specific evidence of an existing violation, (2) a showing that “reasonable legislative or administrative standards for conducting an ... inspection are satisfied with respect to a particular [establishment],” 436 U.S. at 320-21, 98 S.Ct. at 1824, 56 L.Ed.2d at 316, (quoting Camara [v. Municipal Court of City and County of San Francisco], 387 U.S. [523] at 538, 87 S.Ct. [1727] at 1736, 18 L.Ed.2d [930] at 940), or (3) a showing that the search is “pursuant to an administrative plan containing specific neutral criteria”. 436 U.S. at 323, 98 S.Ct. at 1826, 56 L.Ed.2d at 318. 638 F.2d at 907. On remand the District Court held that both the 1971 and 1972 attempts to review NOPSI were reasonable under the fourth amendment. The Court then entered an injunction requiring NOP-SI to submit to a compliance review. United States v. New Orleans Public Service, Inc., 550 F.Supp. 911 (E.D.La.1982). On January 14, 1983, the district court issued a “General Injunctive Order” permanently enjoining NOPSI from refusing to comply with section 202(5) of E.O. 11,246. That order also authorized the commencement of discovery and the court’s retention of jurisdiction for the purpose of conducting a trial concerning NOPSI’s substantive compliance with the E.O. In an uncharacteristic move, NOPSI now appeals from the lower court’s ruling. III. ISSUES Appellant/Defendant raises three points of error. It argues that the district court *426 erred in holding that the 1971 and 1972 selections were proper, and that its general injunctive order was an abuse of its discretion. After an extensive examination of the 1981 and 1974 hearings in the trial court and the depositions, affidavits, exhibits and briefs filed in connection with those proceedings, we find merit in all of its points of error. Thus, we reverse the trial court’s findings and vacate its injunction. The 1971 Selection To find that the proposed 1971 search in this case was reasonable the lower court needed to find that it was based on any one of three factors: (1) specific evidence of an existing violation, (2) a showing that reasonable legislative or administrative standards for conducting the inspection were satisfied with respect to NOPSI, or (3) a showing that the proper search was pursuant to an administrative plan containing specific neutral criteria. As to specific evidence of an existing violation, the trial court found that at the time of the review a comparison of NOPSI’s EEO-1 reports with the experienced civilian labor force in New Orleans would have shown an underrepresentation of minorities in several employment positions. Finding of Fact 23. The court also determined that discrimination claims had been filed with the EEOC, Finding of Fact 24, and that the EEOC Commissioner had charged NOPSI with violating Title VII. Finding of Fact 25. Finally, he found that as of March 1971, the EEOC had issued three decisions holding that there was reasonable cause to believe that NOPSI employment practices were in violation of Title VII. Finding of Fact 26. As to the second factor, a showing that reasonable legislative standards for conducting an inspection were satisfied with respect to NOPSI, the court found that: “[i]t was Mr. Patton’s practice, in conjunction with the scheduling of compliance reviews,” to consider specifically seven different pieces of information as part of the selection process. Finding of Fact 21 (emphasis added). The Court also stated that it “infers and finds that Mr. Patton did follow his ordinary business practice and consider the [seven] 2 factors ...” Finding of Fact 30 (emphasis added). As to the third factor under our fourth amendment test, the court found that “[t]he government developed and implemented an administrative plan containing specific neutral criteria for scheduling federal contractors within the utility industry for compliance reviews.” Finding of Fact No. 19. (emphasis added). The trial judge then lists two short Findings of Fact, Nos. 31, 32. These conclusory statements merely trace the language of two of the factors of the third element of our fourth amendment test for an administrative search. Before discussing the record evidence that refutes these two findings, we turn to the trial judge’s other findings and explain why we disagree with his reasoning. Even if his other fact findings are correct, as a matter of law they do not satisfy the legal requirements of any one of three factors necessary to satisfy the third element of our fourth amendment reasonableness requirement. The problem is that the findings do not indicate that specific neu *427 tral criteria were applied, or that the person^) selecting NOPSI knew of the specific violations. For example, the trial court noted that a review of EEO-1 reports for the time period in question shows an under-representation of minorities, Finding of Fact 23, yet the court only found that these are “... EEO-1 reports which would have been reviewed by GSA in conjunction with scheduling NOPSI for a review in 1971...” —he did not find that they were reviewed. Finding of Fact 22 (emphasis added). The same problem plagues findings 24-27. The judge notes the existence of evidence of probable or specific violations and that there had not been any prior reviews of NOPSI — yet he never found that the GSA official who selected NOPSI considered any of this information. Finding of Fact 19 notes that GSA applied a neutral plan to select industries targeted for review. Even assuming the truth of this conclusion, it only concerns an initial step of one of the selection methods, it does not indicate that a specific company (NOPSI) was selected pursuant to neutral criteria. The judge also found that GSA “had” specific evidence that NOPSI was in violation of E.O. 11,246, Finding of Fact 33, but did not find that GSA considered the evidence in its possession. 3 The inferential approach that weakens most of the Court’s factual findings is apparent in Finding of Fact No. 30 where the Court found that: Because more than ten years have elapsed since the scheduling of the 1971 postaward compliance review of NOPSI, Mr. Patton is unable to recall specifically considering all 4 of the standards and criteria referred to in finding of fact # 21, although it was his ordinary business practice to consider these factors. Mr. Patton does not now believe he did otherwise, and the court infers and finds that Mr. Patton did follow his ordinary business practice and consider the factors listed in finding of fact # 21 when he considered NOPSI for a compliance review. Finding of Fact 30 (emphasis added). The trial Court’s findings of fact and the record indicate only what GSA officials usually did, probably did or had no reason to do otherwise — instead of showing what actually was done. We appreciate the trial judge’s problem. The defendant, who was most likely properly selected, has managed to prolong this litigation to the point where no one connected with the case can remember what criteria, if any, were actually used. Of course, GSA officials may not remember whether they applied neutral criteria not simply because of the passage of time, but because no such criteria were ever applied. Mitchell admitted that he could remember specific selections in which proper criteria were applied in the time period in question but could not remember the NOPSI case. 1981 hearing Tr., at 56-57. Moreover, he admitted that he could not remember whether the national office reviewed Patton’s selection of NOPSI, even though he had cause to remember NOPSI because it was the first company he had ever encountered which had denied it was susceptible to a review. Mitchell Depo. Aug. 25 at 47-48. Whatever the reason(s) for the officials’ lack of memory, the fourth amendment will not permit an administrative search to be justified on ex post facto suppositions, speculations or even well-founded probabilities as to whether investigators applied neutral criteria. What the government could not demonstrate actually occurred it attempts to prove through a presumption of administrative regularity and evidence of NOPSI violations that was in GSA’s possession. Even assuming that GSA properly applied its administrative standards 99% of the time *428 and that NOPSI was guilty of employment violations, the fourth amendment requires an actual showing that the proper procedures were followed at the time of the proposed inspection, not hindsight justifications based on the probability of administrative regularity and evidence of existing violations unknown to investigating officials. As the Supreme Court has noted, “the procedure of antecedent justification ... is central to the Fourth Amendment ...” Katz v. United States, 389 U.S. 347, 359, 88 S.Ct. 507, 515, 19 L.Ed.2d 576 (1967); Dow Chemical Co. v. United States, 536 F.Supp. 1355, 1359 n. 6 (E.D.Mich.1982). The Eleventh Circuit has argued that in evaluating probable cause for an administrative search, “the principal concern is that the plan be neutral.” Donovan v. Sarasota Concrete Co., 693 F.2d 1061, 1068 (11th Cir.1982). While the plan itself is very important, a determination of whether a plan was actually applied neutrally is of equal concern. Some agencies are understaffed and underfinanced, thus they may.be tempted to take short cuts. In the case at bar, for example, Mitchell admitted that in early 1971 the compliance program was an understaffed, interim organization with a paucity of resources. Moreover, agencies and public officials occasionally harass companies which have been the subject of long standing investigations. First Alabama Bank of Montgomery v. Donovan, 692 F.2d 714, 720 (11th Cir.1982) (“... this Court finds itself somewhat sympathetic to the bank’s claim [of harassment by the Labor Dept.]”). Clearly a neutral plan standing alone does not afford protection; it must be applied properly to preserve fourth amendment rights. We note that requiring actual proof of what was done will not be unduly burdensome or expensive for administrative agencies. Instead of trusting their memories, officials should keep a written record of the criteria they use in each particular case. In the case at bar, for example, every quarter Patton sent Mitchell a list of companies he had selected for review. If this schedule had stated that he had utilized GSA guidelines and included examples where the established criteria justified an investigation of a company (e.g., review of EEO-1 reports showed minority underrepresentation) then the fourth amendment would have been satisfied. Although oral testimony of the officials involved can prove that neutral criteria were used, agencies are well advised to make proper records and maintain them. We note that the propriety of the 1971 selection could possibly have been determined in this case if GSA had not destroyed NOPSI selection documents pursuant to general government routine. See Mitchell Depo. Aug. 25, at 36, 55-56. Appellee has argued that an agency is entitled to a presumption of regularity, a presumption that government officials selected NOPSI in accordance with neutral criteria then in effect. We recognize this presumption and also have no doubt that the officials involved in this case acted in the utmost good faith. 5 Nevertheless, “that presumption is not to shield [an agency’s] action from a thorough, probing, in-depth review.” Citizens to Preserve Overton Park v. Volpe, 401 U.S. 402, 416, 91 S.Ct. 814, 823, 28 L.Ed.2d 136 (1971); Sierra Club v. Sigler, 695 F.2d 957, 964 (5th Cir.1983). Our thorough review of the record indicates that the 1971 selection occurred at a time of transition for both the compliance program and the GSA officers involved. In October of 1970 Mitchell was assigned responsibility for reorganizing GSA’s federal contract compliance program under the E.O. It was at this time that he established a contract compliance schedule, which was sent to newly designated regional compliance officers that were part of what Mitchell termed an “interim organization.” 6 He noted that because he was in the process of organizing the program, the national office was forced to conduct reviews “to the best *429 of our ability using quick analysis.” Mitchell Depo. Aug. 26, at 48. He also admitted that the program was unable to reach its goals because of the “paucity of [its] resources.” Mitchell Depo. Aug. 26, at 16. Finally, he testified that at times the national office was “almost flooded with requests for pre-award clearance.” Mitchell Depo. Aug. 25, at'32. Mr. Patton assumed his position in the regional office in November of 1970 and may not have had time to become aware of all the factors he was supposed to apply or may not have had time to apply them. He testified, for example, that when NOPSI was scheduled that he had no specific directives from the national office, other than the listing of industries for which GSA had primary responsibility. U.S. Exhibit 3(D), at 19. We are not dealing with a case where a government agency had been applying the same plan and selection guidelines for several years. The agency and its officials were understaffed, underfinanced and in a transition stage. Moreover, the record reveals that the newly established procedures were not always followed. Mitchell Depo. Aug. 25, at 31-32. On these facts the presumption has been amply rebutted. Additional Record Evidence Indicating. That The 1971 Selection was Improper The previous findings of fact, even if true, could not sustain a search as a matter of law because they did not indicate that specific neutral criteria were followed. We now turn to Findings of Fact 31 and 32, which if true would justify the proposed administrative search. Here the district court found that GSA followed reasonable administrative standards pursuant to an administrative plan containing specific neutral criteria. When one examines the record in detail one understands why the trial judge’s Findings of Fact 31 and 32 are erroneous. Although Patton formally notified NOPSI that it had been scheduled for a review, the record evidence does not reveal who selected NOPSI. Patton could have scheduled NOPSI because it was one of the companies on the list selected by Mitchell or he could have chosen NOPSI himself by applying the seven factors he normally considered. 1981 Hearing Tr., at 39. If Patton selected NOPSI on his own the selection can be justified only if he utilized neutral criteria 7 that prevented him from exercising unbridled discretion. Patton testified that he routinely considered seven factors in scheduling reviews. Patton admitted, however, that he did not give particular weight to any one of these factors but after reviewing them he would use his own “generalized judgment” to select contractors for compliance reviews. U.S. Exhibit 3(D), at 44-45. Most significant, he can not remember whether he actually considered any of those factors in connection with NOPSI’s selection. U.S. Exhibit 3(D), at 21-23. After a regional officer selected companies for a compliance review a quarterly list would go to the national office in Washington where that decision was to be reviewed. The record does not indicate that there was a thorough review at the national office. Patton testified that in the “great majority” of cases his recommendations were accepted. U.S. Exhibit 3(D), at 16. Mitchell testified that he “overwhelmingly” approved the regional officer’s selections. Mitchell Depo. Aug. 25, at 37. Moreover, Mitchell did not require regional officers to *430 state in their quarterly schedules the reasons why they had selected a particular company because “I certainly took the assumption that the director of mine was following the guidelines that they had.” Mitchell Depo. Aug. 25, at 52. Without such information, we doubt that the national office conducted much more than a pro forma review. The reports did usually contain the number of employees a selected company had and what SIC code it was in, Mitchell Depo. Aug. 25, at 54, but this is not enough information for a thorough review to prevent the unbridled exercise of discretion by a regional compliance officer. Finally, even if the national office sometimes conducted its own review of employment statistics and other criteria concerning suggested reviews, without the benefit of the regional officer’s reasons, we note that there is no evidence in the record that the national office undertook such a review in NOPSI’s case. Indeed, a reasonable inference from the record is that the national office did not have time for thorough reviews. Mitchell testified that in 1971 the national office was understaffed and some of the regular reviews submitted could not be reviewed. When this occurred he said that the regional officers had authority to make changes in the schedules of companies to be reviewed. Mitchell Depo. Aug. 25, at 32. Patton may have continued his efforts to investigate NOPSI even though the national office did not review Patton’s initial selection of NOPSI (if Patton did in fact select NOPSI). Of course, NOPSI could have been on the list of companies initially selected by Mitchell and sent to the regional offices for scheduling of reviews. If that was the case, Patton’s scheduling of NOPSI would not have necessitated review by the national office. Even assuming that Mitchell applied neutral criteria in formulating his list, the record does not reveal whether NOPSI was on this list. The list was destroyed and neither Mitchell nor Patton remembers whether NOPSI was on the list. Such facts raise serious questions about the discretion Patton wielded. Moreover, there is insufficient evidence in the record to indicate that the selection was not the product of “unbridled discretion” of the type that the Supreme Court has refused to countenance. As this court has stated: “[i]t is important that ‘the decision to enter and inspect ... not be the product of the unreviewed discretion of the enforcement officer in the field.’ ” Mississippi Power & Light, 638 F.2d at 907-08 (quoting See v. Seattle, 387 U.S. 541, 545, 87 S.Ct. 1737, 1740, 18 L.Ed.2d 943, 947 (1967). In any event, we hold that Findings of Fact Nos. 31 and 32 are clearly erroneous. Because the trial court’s conclusions of law 7-10 are based on clearly erroneous fact findings they are incorrect and we need not go through the mechanical exercise of discussing them individually. The 1972 Selection In 1972 NOPSI was again selected by GSA, this time pursuant to the following regulation: Each agency shall ... state at the outset of negotiations for each negotiated contract, that if the award, when let, should exceed the amount of $1,000,000 or more, the prospective contractor ... will be subject to a compliance review before the award of such contract. No such contract shall be awarded unless a pre-award compliance review of the prospective contractor is conducted within twelve months prior to the award. 41 C.F.R. § 60-1.20(d) (1972). As a matter of law, the trial court found that this regulation properly applied to all contractors who in 1971 and 1972 were engaged in negotiations for a government contract of $1 million or more and that it was proper for the government to conduct a review at a time when NOPSI was providing services to the government. Conclusion of Law No. 11. Appellee argues that since the parties were contemplating the consummation of a written contract a government review was justified. In addition, appellee posits that the mechanical nature of section 60-1.20(d) precludes any possibility of the type of “unbridled discretion” disapproved of in Barlow’s. *431 NOPSI contends that both Patton and Mitchell understood that the regulation applied to situations in which a new contractor would be supplying new and additional services exceeding $1 million to the United States or when an old contractor would be supplying new and additional services exceeding $1 million. NOPSI argues that it should not be placed in either of these categories because it was the sole source of electric and gas services in New Orleans and had supplied such services to NASA and the federal government for many years. As such, NOPSI maintains that it was not providing “new and additional services” within the meaning of 41 C.F.R. § 60-1.20(d). Appellant also opines that the pre-award regulation is not merely mechanical but involves some discretion by GSA officials. It contends that while the regulation provides that an agency “must” conduct a review, in practice only a fraction of contractors awarded million dollar contracts are reviewed. It notes that OFCC officials have privately admitted that the agency is incapable of performing all the “required” reviews, and thus maintains that some discretion is used in selecting which contractors will undergo pre-award review. See B. Schlei & P. Grossman, Employment Discrimination Law 883 n. 87 (2nd ed. 1983); see, e.g., Mitchell Depo. Aug. 25, at 31-32. Finally, appellant maintains that there were no ongoing negotiations when GSA attempted the 1972 review. Although rate schedules were discussed after the 1964 contract expired appellant points out that NOPSI officials had made it clear early on that an expanded nondiscrimination clause was not negotiable. An examination of the record reveals that the 1972 selection was improper. First, the record does not indicate that negotiations were taking place when NOPSI was officially selected on March 15, 1972. Second, even if selection occurred at the same time as the negotiations we find that the record fails to reveal who selected NOP-SI and why. The trial judge held that NOPSI was negotiating with NASA. Conclusion of law No. 12. We note, however, that in Finding of Fact 35 he found that negotiations had ceased. An examination of the record reveals 1974 testimony from one NASA official who claims that negotiations occurred at a meeting sometime in February of 1972. 1974 Hearing Tr. at 170. Two NOPSI officials involved in the discussions claim that no negotiations about the contract or rate schedule occurred. 1974 Hearing Tr. at 156; 1981 Hearing Tr., at 111, 114. Moreover, NOPSI argues that no negotiations could have occurred about rate schedules because these schedules were proposed by NOPSI and then approved or rejected by the New Orleans City Council. Once rates were approved NOPSI had no power to alter them and could only advise a user as to the most advantageous rate schedule to use. U.S. Exhibit 1(D), at 214-15. Even if we were inclined to follow the judge’s conclusion of law No. 12 (because he observed the demeanor of the witnesses involved), there is no way to determine from the record when NOPSI was selected. Obviously some GSA official decided to schedule NOPSI for a review before the official notification occurred on March 15, 1972. Nevertheless, even assuming we believe the GSA testimony that negotiations did occur sometime in February, we do not know whether NOPSI was selected during these negotiations or after their conclusion. When Patton wrote NOPSI concerning their March 15, 1972, selection he stated that he had been advised that “GSA is now negotiating with your firm.” U.S. Exhibit 2(e), at H 2 (emphasis added). When deposed, however, Patton could not recall consulting with anyone, and had no knowledge as to the reasons behind GSA’s request for a pre-award review. U.S. Exhibit 3(D), at 72, 79. He also testified that at the time he sent the letter, he would not have even known whether NOPSI was supplying services to NASA. Id. at 73. Thus, the record concerning the 1972 selection does not reveal when NOPSI was negotiating, who determined NOPSI was negotiating or why NOPSI was found to be negotiating. *432 At best, the only certain conclusion we can draw is that NOPSI was selected in 1972 while an “ongoing contractual relationship existed.” 8 After the 1964 contract expired on June 30, 1970, NOPSI, as required by its public utility charter, continued to provide services as it always had. Doing so, however, does not bring it within the ambit of “new and additional services” in the section 60-1.22(d) regulation. NOP-SI was neither a new contractor providing new services nor an old contractor supplying new and additional services. Moreover, Patton has testified it would be unusual to conduct a pre-award review of a sole source contractor. U.S. Exhibit 3(D), at 78. The foregoing facts prove that the trial judge’s Findings of Fact Nos. 46 and 47 are clearly erroneous. 9 Similarly, we disagree with conclusions of law 11-14. Contrary to Conclusion No. 11, we have noted that GSA officials did at times exercise discretion as to which companies to select under § 60-1.-20(d) because they lacked the manpower to conduct all pre-award reviews. As to No. 12, we have pointed out that the record is unclear as to whether negotiations actually occurred. Moreover, even if they did, there is no evidence as to whether they were occurring at the time NOPSI was selected, or who determined they occurred. In conclusion of law No. 13 the trial court attempts to justify selection on the basis of GSA’s policy concerning “ongoing relationships.” As noted earlier, this is an insufficient basis under the statute because it governs only a new contractor or an old contractor providing new and additional services. Finally, our examination of the record contradicts conclusion No. 14. The record clearly fails to reveal that the 1972 selection was pursuant to neutral criteria and not the result of unbridled discretion of a GSA official. Ill Injunction Since we hold as a matter of law that the 1971 and 1972 selections of NOPSI violated the fourth amendment we vacate the trial court’s general injunctive order without prejudice to the agency seeking a review in a proper administrative and legal manner. REVERSED. Footnotes [1] For a survey of fourth amendment criteria courts have applied to administrative searches examine Note, Marshall v. Barlow’s, Inc: Are an Employer’s Fourth Amendment Rights Protected?, 16 Cal.W.L.Rev. 161 (1980); Administrative Searches, 77 Mich.L.Rev. 1291 (1979); Note, Camara See and Their Progeny: Another Look at Administrative Inspections Under the Fourth Amendment, 15 Colum.J. of L. & Soc. Prob. 61 (1979). [2] (1) determine that the contractor was in an industry which had been assigned to GSA by Order No. 1; (2) review the contractor’s most recent EEO-1 reports; (3) determine the available minority representation in the employer’s labor market from the relevant census data and compare it to the employer’s EEO-1 reports; (4) compare the contractor’s representation of minorities and females with the representation of other companies in the industry; (5) determine whether the contractor had been previously reviewed, and, if so, when. The purpose of this procedure was to ensure that companies were scheduled for review who either had not been reviewed or had been reviewed several years in the past; (6) consult with the EEOC to determine the number and nature of outstanding complaints; and (7) determine if other contractors within the geographic area had been reviewed. This was done so as not to concentrate compliance reviews in any one geographic area. [3] If Patton selected NOPSI we note that he testified that he neither had nor considered any evidence of substantive violations of E.O. 11,-246 by NOPSI. U.S. Exhibit 3(D), at 70. If Mitchell selected NOPSI we note that the record fails to indicate that he was aware of any violations by NOPSI. [4] As we read the record, he is unable to recall any of the standards or criteria he used to select NOPSI. [5] Indeed their testimony is refreshingly candid as well as destructive to their case. [6] The “formal system” was not established until June 17, 1971. Mitchell Depo. Aug. 26, 1981, at 47. [7] Although the factors Patton routinely used appear to be specific, neutral and fair we have substantial doubts that they constituted a valid administrative plan. In OSHA inspection warrant cases, for example, courts have required OSHA to produce evidence of a national or general plan. E.g., Donovan v. Wollaston Alloys, Inc., 695 F.2d 1, 5, 8 (1st Cir.1983); Erie Bottling Corp. v. Donovan, 539 F.Supp. 600 (W.D.Pa.1982). In addition, another district court recently held that OSHA selection criteria and procedures must be promulgated pursuant to the APA as amended, 5 U.S.C. § 553 (1976). In The Matter of Establishment Inspection of Kast Metals Corp., 11 O.S.H.C. 1266 (W.D.La. 1983). The record does not indicate that Mitchell approved of Patton’s use of the seven factors. We need not reach this issue, however, because we find that there is insufficient evidence to show that Patton actually considered the seven factors if he did select NOP-SI. [8] This court has already ruled that a contractual relationship exists between NOPSI and NASA, and that NOPSI is a government contractor. 553 F.2d at 469. We now reach the issue of whether NOPSI came within 41 C.F.R. 60-1.22(d). [9] We do not disagree with fact findings 34-45. Their validity does not affect our decision, however, because they do not concern whether specific neutral criteria was involved.