TUTTLE, Senior Circuit Judge:

Ernest Griffin and 21 other black employees and former employees 1 of the United States Postal Service at Jacksonville, Florida, appeal from a decision of the district court finding no classwide or individual discrimination under Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. § 2000e-16. Appellants contend that the district court erred in excluding their challenge to written tests used in the promotion process, in excluding all disparate impact claims, and in finding no disparate treatment in promotions, awards, discipline, and details. On cross-appeal, the Postal Service argues that the district court erred in allowing a third-party complaint to serve as the administrative basis for this suit and in certifying this suit as a class action.

On August 29, 1971, Griffin filed with the United States Civil Service Commission a complaint under the third-party complaint procedure authorized by then-current regulations of the Commission and of the Postal Service. 5 C.F.R. § 713.204(d)(6) (1971); 39 C.F.R. § 747.5(a) (1971). The complaint stated:

Please accept this letter as a third party discrimination complaint against Postmaster J.E. Workman of the Jacksonville, Florida Post Office. This discriminatory complaint is based on race since qualified blacks were and are* still being systematically excluded in training and development and opportunities for advancements.

The Postal Service investigated the complaint and found no discrimination.

On July 7, 1972, plaintiffs filed a class action suit in federal district court challenging defendant’s use of discriminatory assignment and promotion methods and other discriminatory employment practices. On January 9, 1973, the district court entered an order authorizing plaintiffs to proceed as representatives of a class and dismissing that portion of the plaintiffs’ complaint which challenged the use of written tests in the promotion process. The dismissal was based on the court’s finding that plaintiffs had failed to exhaust administrative remedies as to the testing issue.

In 1976, plaintiff Griffin was fired from the Postal Service. He appealed the discharge under the then-existing regulatory scheme to the United States Civil Service Commission. In that appeal, he raised the claim that he had been discriminated against because of his race and that the action was part of a pattern and practice of racial discrimination and reprisal against *1520 those persons challenging discrimination, Griffin timely filed a supplemental complaint in the present action raising those elaims. On the district court’s order, a consolidated amended complaint was filed on November 12, 1981. This complaint again alleged discrimination against blacks in defendant’s assignment and promotion methods and other employment practices,

On September 8, 1982, the district court dismissed plaintiffs’ disparate impact claims on the grounds that plaintiffs’ pleadings had failed to put defendants on notice as to which employment practices would be challenged on this theory and that only objective, facially neutral employment practices could be challenged on a disparate impact theory. Thus, the case proceeded to trial on a disparate treatment theory.

The Jacksonville Post Office employed an average of 1,880 persons during the period covered by this lawsuit. Approximately 32 percent of these employees were black. The employees included clerks, mail handlers, city carriers, window clerks, and motor vehicle operators.

The Jacksonville Post Office promotes persons to supervisory positions almost entirely from within its work force. The process for promotion to initial supervisory positions has undergone some changes during the period covered by the lawsuit. In 1968, the Post Office used two written examinations, one for vehicle services and one for the post office branch. In order to be placed on the supervisory register and be eligible for promotion, employees had to attain a particular score on the examination. The top 15 percent of the employees on the register were placed in the “zone of consideration” and were notified of supervisory vacancies. In 1972, the zone of consideration standard was eliminated and persons who had attained a passing score on the examination were evaluated and graded by their supervisors. Those receiving an “A” rating were eligible for immediate promotion. In 1976, the examination was eliminated and employees were instead required to complete a training program as a precondition for promotion. In 1978, the Postal Service initiated the Profile Assessment System for Supervisors (PASS) which made eligibility for the supervisory registers dependent on both supervisory assessment and self-assessment. No written examination was used under the PASS program. Under all of these promotion systerns, promotion advisory boards interviewed and recommended eligible candidates for promotion, and the final selection was made by the Postmaster,

Both plaintiffs and defendants relied heavily on statistical data. Plaintiffs’ statistics showed that blacks are far more likely to hold jobs at level 4 or lower and far less likely to hold jobs at or above level 7, the initial supervisory level. 2 According to plaintiffs, the probabilities that the grade distributions shown in their tables would occur by chance are one in 10,000. Plaintiffs’ statistics showed that while 35 percent of the work force is black, blacks held only 5 percent of all supervisory jobs in 1969 and only 21 percent in 1981. These statistics also indicated that blacks were promoted to supervisory positions in numbers far lower than expected from 1964 through 1976. Plaintiffs contend that the key to the under-representation of blacks in *1521 supervisory positions is their under-representation on the supervisory registers. Their statistics indicated that the probabilities that the number of blacks on the registers could have occurred by chance ranged from 15 in one hundred trillion in 1968 to 67 in 100,000 in 1977. 3

Plaintiffs’ statistics were based on the use of the entire craft work force of the Jacksonville Post Office as the applicant pool for supervisory positions. The government contends that the appropriate pool is those individuals on the supervisory registers. Using this pool, the government’s statistics showed no evidence of systemic discrimination against blacks seeking supervisory positions.

Plaintiffs’ statistics showed a consistent statistically significant over-disciplining of blacks in comparison to their numbers in the relevant work force. Blacks, who constitute 35 percent of the work force, received between 52 and 67 percent of the discipline. 4 The government concedes that blacks are disciplined more often than whites, but argues that factors other than race explain the disparity. Defendants’ statistics illustrated that black employees at the Jacksonville Post Office are, on the average, younger than white employees and take more time off from work. The Postal Service statistics also demonstrated that, controlling for the number of previous offenses, black employees did not receive more severe punishment than white employees.

In finding no discrimination, the district court held that plaintiffs’ statistical tables had negligible probative value. This finding was based in part on the court’s determination that plaintiffs’ expert had failed to control for relevant variables, such as the fact that promotions were made only from those on the supervisory registers. In addition, the court noted that defendants were able to point out errors in many of plaintiffs’ tables, that many of plaintiffs’ tables provided no raw data, and that the credibility of plaintiffs’ expert was undercut by his presentation at the last minute of a substantially new statistical report. The court found the report prepared by defendant’s expert to be a reliable and credible analysis of the promotion practices at the Post Office. The court also found that, based upon the government’s statistics, it is likely that there were different characteristics, patterns of conduct, or reactions to circumstance which explain the different levels of discipline.

Appellants introduced the testimony of 24 black class members to bring their statistical evidence to life. The district court concluded that appellants had not produced a single witness who had demonstrated a claim of discrimination. The court found that many of plaintiffs’ witnesses were not believable, that others were mistaken that they were eligible for promotion, that others were not as qualified as the employee selected, that some of the promotions challenged had gone to black employees, and *1522 that there were other non-discriminatory reasons to explain the other alleged instances of discrimination.

I. DISMISSAL OF PLAINTIFFS’ CHALLENGE TO THE WRITTEN TESTS

The district court’s order of January 9, 1973, dismissed that portion of plaintiffs’ complaint challenging the use of written tests as a condition of promotion. The court noted that the requirement of exhaustion of administrative remedies is satisfied when the issues (a) are expressly raised in the pleadings before the administrative agency, (b) might reasonably be expected to be considered in a diligent investigation of those expressly raised issues, or (c) were in fact considered during the investigation. The court held, however, that the Postal Service had not had an opportunity to consider the issue of the written test.

The starting point for determining the permissible scope of a judicial complaint is the administrative charge and investigation. The judicial complaint is limited to the scope of the administrative investigation which could reasonably be expected to grow out of the charge of discrimination. Evans v. U.S. Pipe & Foundry Co., 696 F.2d 925, 929 (11th Cir.1983); Eastland v. Tennessee Valley Authority, 714 F.2d 1066, 1067 (11th Cir.1983), cert. denied sub nom., James v. Tennessee Valley Authority, — U.S. -, 104 S.Ct. 1415, 79 L.Ed.2d 741 (1984). Griffin’s administrative complaint charged racial discrimination in that “qualified blacks were and are still being systematically excluded in training and development and opportunities for advancements.” We hold that Griffin’s complaint clearly challenged aspects of defendant’s employment practices which would reasonably include testing. The written examinations were an integral part of the promotional scheme from 1968 through 1976 because employees became eligible for promotion to supervisory positions only by attaining a passing score on the examination. Thus, the impact of the written tests should have been encompassed in a reasonable investigation of this charge of systemic discrimination in promotions.

In fact, the investigative report prepared by the Postal Service at the conclusion of its investigation contains numerous references to the written tests. The report contains copies of both the 1968 and 1971 supervisory registers, and indicates that only one black was in the zone of consideration for promotion. The investigator’s report indicates that almost half óf the 31 black employees interviewed referred to the supervisory register or the written examination. Several of these specifically indicated that they were ineligible for supervisory positions because they had failed the written examination.

We find that the testing issue was or should have been included in a reasonable investigation of the administrative complaint. We therefore reverse the order of the district court dismissing plaintiffs’ challenge to the written tests and remand for consideration of that claim.

II. DISMISSAL OF DISPARATE IMPACT CLAIMS

In the court below plaintiffs sought to rely on a disparate impact theory as well as on a disparate treatment theory. Plaintiffs sought to apply the disparate impact theory both to the final results of the multi-component promotion process and to several component parts of that process, including promotion advisory boards, awards, and discipline. In its order of September 8, 1982, the district court granted defendant’s motion to dismiss all claims by plaintiff based on a disparate impact theory. The court found that disparate impact analysis is appropriate only to challenge objective, facially neutral employment practices, and not to challenge either the cumulative effect of employment practices or subjective decision-making. The court further found that plaintiffs’ pleadings had failed to put defendants on notice as to which employment practices would be challenged on a disparate impact theory.

The district court relied on Pouncy v. Prudential Insurance Company of America, 668 F.2d 795 (5th Cir.1982), and on Harris v. Ford Motor Co., 651 F.2d 609 *1523 (8th Cir.1981). In Harris, the Eighth Circuit held that a subjective decision-making system cannot alone form the foundation for a disparate impact case. Id. at 611. In Pouncy, the Fifth Circuit stated:

The discriminatory impact model of proof in an employment discrimination case is not, however, the appropriate vehicle from which to launch a wide ranging attack on the cumulative effect of a company’s employment practices____ We require proof that a specific practice results in a discriminatory impact on a class in an employer’s work force in order to allocate fairly the parties’ respective burdens of proof at trial____ Identification by the aggrieved party of the specific employment practice responsible for the disparate impact is necessary so that the employer can respond by offering proof of its legitimacy.

Id. at 800-01.

A recent Eleventh Circuit decision referred to the Pouncy case and indicated that use of the disparate impact model to attack the excessive subjectivity of a personnel system is “troublesome.” The court stated, however:

Former Fifth Circuit precedent, however, indicates that subjective selection and promotion procedures may be attacked under the disparate impact theory. See Johnson v. Uncle Ben’s, Inc., 628 F.2d 419, 426-27 (5th Cir.1980), vacated, 451 U.S. 902, 101 S.Ct. 1967, 68 L.Ed.2d 290 (1981), modified and aff'd in part, rev’d in part, 657 F.2d 750 (5th Cir.1981), cert. denied, 459 U.S. 967, 103 S.Ct. 293, 74 L.Ed.2d 277 (1982).

Eastland v. Tennessee Valley Authority, 704 F.2d 613, 619-20 (11th Cir.1983), cert. denied sub nom., James v. Tennessee Valley Authority, — U.S. -, 104 S.Ct. 1415, 79 L.Ed.2d 741 (1984). In Eastland, the Court declined to decide whether the prior Fifth Circuit case law was distinguishable because the Court found that plaintiffs had failed to prove discrimination under the disparate impact theory.

In this case the issues of whether the disparate impact model can be used to challenge the final results of a multi-component selection process and whether the disparate impact model can be used to challenge subjective elements of a selection process are squarely before us. We find that we are bound by former Fifth Circuit precedent to allow disparate impact challenges to the end result of multi-component selection procedures and to subjective selection procedures. 5 Further, we hold that even if these prior Fifth Circuit cases were not binding, use of the disparate impact theory to challenge the end result of multi-compo-nent selection processes and to challenge subjective elements of those processes is appropriate. We therefore reverse the order of the district court dismissing plaintiffs’ disparate impact claims.

Several decisions of the former Fifth Circuit, binding on this panel, applied a disparate impact analysis to the end result of multi-component selection processes containing subjective elements. For example, in Johnson v. Uncle Ben’s, Inc., 628 F.2d 419, 426-27 (5th Cir.1980), vacated, 451 U.S. 902, 101 S.Ct. 1967, 68 L.Ed.2d 290 (1981), modified and aff'd in part, rev’d in part, 657 F.2d 750 (5th Cir.1981), cert. denied, 459 U.S. 967, 103 S.Ct. 293, 74 L.Ed.2d 277 (1982), the court applied a disparate impact analysis to a promotion system based on the use of subjective supervisory evaluations. See also, Crawford v. Western Electric Co., Inc., 614 F.2d 1300, 1318 (5th Cir.1980) (applying disparate impact analysis to index review system involving subjective elements); Rowe v. General Motors Corp., 457 F.2d 348, 354-55 (5th Cir.1972) (applying disparate impact analysis to promotion system involving foreman’s recommendations).

Even if we were not bound by these decisions to allow application of disparate impact analysis to the end result of a multi-component promotion process and to processes involving subjective elements, we *1524 would still be inclined not to follow the decision of the current Fifth Circuit in Pouncy. 6 The Supreme Court first articulated the disparate impact model of discrimination, under which proof of discriminatory intent is not necessary, in Griggs v. Duke Power Co., 401 U.S. 424, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971). In Griggs, the Court indicated that Title VII requires “the removal of artificial, arbitrary, and unnecessary barriers to employment” which “operate as ‘built-in headwinds’ for minority groups and are unrelated to measuring job capability.” Id. at 431-32, 91 S.Ct. at 853-54. The Court in Griggs did not differentiate between objective and subjective barriers, and, in fact, the Court made frequent references to “practices” and “procedures,” terms that clearly encompass more than isolated, objective components of the overall process. 7

In the recent case of Connecticut v. Teal, 457 U.S. 440, 102 S.Ct. 2525, 73 L.Ed.2d 130 (1982), the Supreme Court held that the “bottom line” result of a promotional process could not be used as a defense to a disparate impact challenge to a particular selection procedure used in that promotion process. The Court emphasized the holding in Griggs that Title VII requires the elimination of “artificial, arbitrary, and unnecessary barriers to employment,” and again did not differentiate between objective and subjective criteria nor give any indication that a disparate impact challenge could not be made to a promotional system as a whole. See 457 U.S. at 448-452, 102 S.Ct. at 2532-2534. The Court noted the legislative history of the 1972 amendments to Title VII, 86 Stat. 103-113, which extended Title VII to federal government employees. The Court pointed out that Congress recognized and endorsed the disparate impact analysis employed in Griggs. The Court specifically cited the Senate Report (S.Rep. No. 92-415, p. 5 (1971)), which stated:

Employment discrimination as viewed today is a ... complex and pervasive phenomenon. Experts familiar with the subject now generally describe the problem in terms of ‘systems’ and ‘effects’ rather than simply intentional wrongs.

Connecticut v. Teal, 457 U.S. 440, 447 n. 8, 102 S.Ct. 2525, 2531 n. 8, 73 L.Ed.2d 130. The dissenters in Teal, while disagreeing with the Court’s conclusion that the bottom line could not be used as a defense, clearly indicated their understanding that disparate impact challenges could be made to the total selection process. The dissenters stated that “our disparate-impact cases consistently have considered whether the result of an employer’s total selection process had an adverse impact upon the protected group.” Id. at 458, 102 S.Ct. at 2537 *1525 (Powell, J., dissenting) (emphasis in original).

We have repeatedly held that subjective practices such as interviews and supervisory recommendations are capable of operating as barriers to minority advancement. See, e.g., Johnson, 628 F.2d at 426; Rowe, 457 F.2d at 359; Miles v. M.N.C. Corp., 750 F.2d 867, 871 (11th Cir. Jan. 13, 1985). Exclusion of such subjective practices from the reach of the disparate impact model of analysis is likely to encourage employers to use subjective, rather than objective, selection criteria. Rather than validate education and other objective criteria, employers could simply take such criteria into account in subjective interviews or review panel decisions. It could not have been the intent of Congress to provide employers with an incentive to use such devices rather than validated objective criteria.

Likewise, limiting the disparate impact model to situations in which a single component of the process results in an adverse impact completely exempts the situation in which an adverse impact is caused by the interaction of two or more components. This problem was recognized in the recent Eighth Circuit decision in Gilbert v. City of Little Rock, Ark., 722 F.2d 1390 (8th Cir. 1983), cert. denied, — U.S.-, 104 S.Ct. 2347, 80 L.Ed.2d 820 (1984). The Court there held that the district court’s finding of no discrimination under a disparate impact theory was incorrect because “the district court neglected to adequately consider the interrelationship of the component factors and, more specifically, whether the oral interview and performance appraisal factors ... had a disparate impact____” Id. at 1397-98.

Finally, we note that the Uniform Guidelines on Employee Selection Procedures, 29 C.F.R. § 1607, developed by the four federal agencies with responsibility for enforcing Title VII, interpret the disparate impact model to apply to all selection procedures, whether objective or subjective. The Guidelines define the selection procedures to which a disparate impact analysis applies as follows:

Any measure, combination of measures, or procedure used as a basis for any employment decision. Selection procedures include the full range of assessment techniques from traditional paper and pencil tests, performance tests, training programs, or probationary periods and physical, educational, and work experience requirements through informal or casual interviews and unscored application forms.

29 C.F.R. § 1607.16(Q).

We therefore reverse the order of the district court dismissing plaintiffs’ disparate impact claims and remand to that Court for consideration of the plaintiffs’ disparate impact challenges to the final result of the defendants’ overall promotion process and to specific components of that process whether subjective or objective.

III. DISPARATE TREATMENT IN PROMOTIONS, DETAILS, DISCIPLINE, AND AWARDS

After the district court eliminated plaintiffs’ disparate impact claims, plaintiffs proceeded to trial on a disparate treatment theory. After consideration of plaintiffs’ statistical and anecdotal evidence, the district court held that plaintiffs had failed to prove any class-wide discrimination.

In a disparate treatment case proof of discriminatory motive or intent is essential. International Brotherhood of Teamsters v. United States, 431 U.S. 324, 335-36 n. 15, 97 S.Ct. 1843, 1854-55 n. 15, 52 L.Ed.2d 396 (1977). In an action alleging class-wide discrimination plaintiffs must “establish by a preponderance of the evidence that racial discrimination was the company’s standard operating procedure— the regular rather than the unusual practice.” Id. at 336, 97 S.Ct. at 1855. A prima facie case of disparate treatment may be established by statistics alone if .they are sufficiently compelling. East-land, 704 F.2d at 618 (11th Cir.1983). The prima facie case is enhanced if the plaintiff offers anecdotal evidence to bring “the cold number's convincingly to life.” Teamsters, 431 U.S. at 339, 97 S.Ct. at 1856.

Once plaintiff establishes a prima facie case of disparate treatment, the burden *1526 shifts to defendant to rebut the inference of discrimination by showing that plaintiffs’ statistics are misleading or by presenting legitimate non-discriminatory reasons for the disparity. The defendant does not have to persuade the court that it was actually motivated by the proffered reasons. It is sufficient if the defendant raises a genuine issue of fact as to whether it discriminated. Texas Department of Community Affairs v. Burdine, 450 U.S. 248, 254, 101 S.Ct. 1089, 1094, 67 L.Ed.2d 207 (1981). If the defendant carries this burden, the presumption raised by the pri-ma facie case is rebutted, and the plaintiff must prove that the reasons offered by the employer were pretextual. Id. at 256, 101 S.Ct. at 1095.

This Court may not reverse the decision of the district court unless plaintiffs establish that the court’s findings of fact, whether of subsidiary or ultimate fact, are clearly erroneous or that the court erred as a matter of law. Pullman-Standard v. Swint, 456 U.S. 273, 285-90, 102 S.Ct. 1781, 1788-91, 72 L.Ed.2d 66 (1982); Giles v. Ireland, 742 F.2d 1366, 1374 (11th Cir. 1984). We now consider the court’s findings as to each of the four challenged practices under this standard.

A. Promotions

Although, as noted above, the district court was critical of plaintiffs’ statistics for several reasons, a key reason for the court’s discrediting of the statistics concerning promotion to first level supervisor, was that plaintiffs used the craft work force as the labor pool rather than using those employees on the supervisory registers. In making this criticism, the court erred as a matter of law because it excluded the first step of the promotion process, that of getting on the supervisory register. During the time period covered by this suit, that first step involved both objective elements, such as passing the written examination, and subjective elements, such as supervisory evaluations. When promotions to supervisory positions are made almost exclusively from the internal work force and when the primary qualification for promotion is experience in the craft work force, the appropriate comparison is to this work force rather than to those on the supervisory register, who have already been screened by the agency through the use of various procedures. See Johnson, 628 F.2d at 425; Carroll, 708 F.2d at 192. Because it is unclear whether the district court would have found plaintiffs’ statistics sufficient to establish a prima facie case if it had considered the craft work force to be the appropriate labor pool, we must remand this issue for further proceedings.

On remand if the plaintiffs establish a prima facie case of disparate treatment, the burden will shift to defendant to rebut the inference of discrimination by showing that plaintiffs’ statistics are misleading or by presenting legitimate non-discriminatory reasons for the disparity. Burdine, 450 U.S. at 254, 101 S.Ct. at 1094. The district court relied in part on defendant’s statistics concerning the percentages of employees on the supervisory registers to find plaintiffs’ statistics misleading. Appellants contend that defendant cannot rely on the supervisory registers to rebut the presumption of discrimination unless the written tests used in determining eligibility for the registers have been validated in compliance with Griggs. The Postal Service argues that it can rebut a prima facie case by articulating a reason which is non-discriminatory on its face and that it is not required to prove or validate its reason.

The necessity of validating selection devices identified by employers to rebut a prima facie case of disparate treatment whs exhaustively analyzed in a recent opinion by Judge Wright for the District of Columbia Circuit. In Segar v. Smith, 738 F.2d 1249 (D.C.Cir.1984), the Court discussed the nature of the burden on a defendant in a pattern or practice disparate treatment case as follows:

The defendant must at least make a “clear and reasonably specific showing,” based on admissible evidence, that the alleged nondiscriminatory explanation in fact explains the disparity. Burdine, supra, 450 U.S. at 253-255, 101 S.Ct. at 1093-1095. In the context of an individu *1527 al plaintiff’s claim of disparate treatment, the Court in Burdine suggested that a defendant need do no more than make such an articulation. Id. at 253-254, 101 S.Ct. at 1093-1095. Though the principles on which Burdine is based are fully applicable to pattern or practice cases, the specific definition of the rebuttal burden on an employer in an individual plaintiff’s disparate treatment case should not be unthinkingly applied to class actions such as the present case____

In the context of an individual’s suit, the bare articulation of a legitimate nondiscriminatory explanation generally suffices to undermine a plaintiff’s initial proof ... because the plaintiff’s prima facie case will typically consist of the low-threshold showing of McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973).

When a defendant in a pattern or practice class action offers such an explanation, the circumstances differ in two crucial ways. First, to make an initial showing of disparate treatment in such cases the plaintiff class will typically have presented statistical evidence showing pervasive disparities and eliminating most, if not all, potential nondiscriminatory explanations for the observed disparities. Though the employer is not required to meet a burden of persuasion in rebutting the disparate treatment claim, the nondiscriminatory explanation must cast sufficient doubt on the plaintiffs’ proof to permit the trier of fact legitimately to decline to draw an inference of discrimination from that proof. The bare articulation of a nondiscriminatory explanation, while sufficient to rebut an individual plaintiff’s low-threshold McDonnell Douglas showing, generally will not suffice as a rebuttal to a typical class-wide showing of pervasive discrimi-nation____

Second, the employer’s effort to rebut the pattern or practice claim by articulating a legitimate nondiscriminatory explanation may have the effect of putting before the court all the elements of a traditional disparate impact case. By its explanation of an observed disparity the employer will typically pinpoint an employment practice (or practices) having a disparate impact on a protected class. And to rebut plaintiffs’ case the employer will typically be required to introduce evidence showing that the employment practice in fact caused the observed disparity. See Burdine, supra, 450 U.S. at 258, 101 S.Ct. at 1096 (“defendant will normally attempt to prove the factual basis for its explanation”). In this situation, between the plaintiffs’ prima facie showing of disparity and the defendant’s rebuttal explanation of the disparity, the essential elements of a disparate impact case will have been placed before the trier of fact. Such a case is ripe for resolution using disparate impact analysis. Though the plaintiffs in a disparate treatment case bear the burden of persuasion as to the existence of a disparity, the defendant bears the burden of proving the business necessity of the practices causing the disparity. Albemarle Paper Co. [v. Moody ], supra, 422 U.S. [405] at 425, 95 S.Ct. [2362] at 2375 [45 L.Ed.2d 280 (1975) ]. Thus when an employer defends a disparate treatment challenge by claiming that a specific employment practice causes the observed disparity, and this defense sufficiently rebuts the plaintiffs’ initial case of disparate treatment, the defendant should at this point face a burden of proving the business necessity of the practice.

Id. at 1268-70 (some citations and footnotes omitted).

Judge Wright noted that some other circuits, including the Fifth Circuit in Pouncy, have expressed a reluctance to apply disparate impact analysis in this situation. The reluctance of these courts is based in part on the perceived unfairness of placing on the defendant the dual burden of articulating which of its employment practices caused the adverse impact and of proving the business necessity of the practice and in part on the risk that an employer will be forced to justify the entire range of its employment practices when a plaintiff *1528 shows only that a disparity exists. Id. at 1270.

The Segar Court found these concerns to be both unpersuasive and difficult to harmonize with the purposes of Title VII. The Court noted that the rebuttal burden will be placed upon an employer only after a plaintiff class has shown a disparity in the positions of members of the class and the majority group who appear to be compar-a y qua lie. e our pom e ou .

The defendant will m all likelihood point to a specific job qualification or performance/evaluation rating as the ex- , „ , f ,. ., planation for the observed disparity. „ j. , . , r Thus application of disparate impact m this situation will not, the fears of the Pouncy court notwithstanding, place on the employer any additional burden of articulation; to rebut the disparate treatment claim the employer will have had to articulate which employment practices cause an observed disparity. Nor will the employer be forced to justify all of te employment practices. The employer will be required to show the job relatedness of only the practice or practices identified as the cause of the disparity.

Id. at 1271. The Court also noted that the purpose of Title VII is the removal of artificial, arbitrary and unnecessary barriers to employment which operate invidiously to criminate on the basis of race or other impermissible classifications. The Court , S a ®

„ , , This purpose is not well served by a requirement that the plaintiff m every ^ . . , , ,, , , ,, , case pinpoint at the outset the employ- . .. . , , ment practices that cause an observed disparity between those who appear to be comparably qualified. Such a requirement in effect permits challenges only to readily perceptible barriers; it allows subtle barriers to continue to work their discriminatory effects, and thereby thwarts the crucial national purpose that Congress sought to effectuate in Title VII. “It is abundantly clear that Title VII tolerates no discrimination, subtle or otherwise.” McDonnell Douglas Corp., supra, 411 U.S. at 801, 93 S.Ct. at 1824. Thus when unnecessary employer-created barriers have been brought into the open through adjudication of a pattern or practice disparate treatment claim, these barriers should be evaluated under the disparate impact theory, as Congress intended them to be.

Id. at 1271-72.

We agree with the analysis by the Court in s Thus on remand) if the plaintiffs are able to establish a prima facie case of disparate treatment) the defendants may not rebut the presumption of discrimination b reliance on tbe supervisory registers or ^ written tegtg unlegg thoge procedures , , , , have been validated as required under a , . , , . disparate impact analysis. 1 ” J

Although the focus of plaintiffs’ appeal concerning discrimination in promo-tions is on promotions to initial supervisory positions, plaintiffs also challenge defendant’s system for promotions to higher level supervisory positions and to non-superviSory positions. The district court found that plaintiffs’ statistics on these promotions showed a mix of positive and negative deviations and that such results are typical 0f a non-discriminatory environment. We cannot gay that thege findings by the dis-trict court are clearly erroneous,

B. Details

Plaintiffs alieged that blacks were discriminated against in the awardi of de. . ., . . . , , , . , tails, i.e., temporary assignments to higher level positions to fill in for absent employ-eeg and to in ience in thoge jobs. m, , . , , , . . , The plaintiffs contend that being assigned ~ to a detail increases ones chances for per, , , , , , It, “afen4t Pr0f°tl0I\ a a latef date'. The distr‘ct court found Pontiffs statistics on details unconvincing for a number of reasons> but a primary reason was that plain-statistics ^ compared the number of employees detailed to the number of em-Payees in the craft work force. The Court thus found that plaintiffs failed to control for other variables which were important in determining who was assigned to details, including the nature of the detail, and the experience, availability, and location of the employee. The Court found defendant’s statistics on details more convincing. De *1529 fendant’s statistics compared the number of employees detailed to a combination of the total number of employees in the craft and the number of employees on the supervisory register. Apparently, preference in details is given to employees on the supervisory register, but employees who are not on the register also often receive details. The Court’s conclusions with respect to details are infected by the same error that infected its conclusions as to promotions to initial level supervisory positions. The Court erroneously relied upon statistics by the defendant comparing those employees detailed to those employees on the supervisory register. As discussed above, this is not the appropriate comparison. Therefore, the Court’s conclusions with respect to details are reversed and remanded for further proceedings.

C. Awards

The plaintiffs alleged that blacks were discriminated against in the giving of awards. The district court found that the studies by both plaintiff and defendant showed that blacks were sometimes awarded fewer awards than whites and sometimes more. The court also found that other characteristics, including performance levels, may account for any unequal distribution of awards. We cannot say that the court’s findings as to awards are clearly erroneous.

D. Discipline

Both parties and the court agree that black employees at the Jacksonville Post Office receive a higher percentage of discipline as a group than white employees. The court nevertheless found that race was not a statistically significant factor in the imposition of discipline at the Jacksonville Post Office. In so finding, the court relied both on the fact that there are internal and external review mechanisms for disciplinary actions at the Post Office and on data produced by the defendants showing that there are significant differences in the characteristics of the black and white postal employees. The court stated:

The system by which discipline is imposed at the Jacksonville Post Office, however, has innumerable checks and balances. Both internal reviews and several methods of appeal are available to all employees. Disciplinary actions are subjected to such scrutiny as to substantially reduce the likelihood of unwarranted discipline. Thus, there is the clear likelihood that the disciplinary actions are reflective of the employee’s conduct, not race.

Consistent with that likelihood are data which showed that there are significant differences in the characteristics of the black and white postal employees. The data shows, and the court finds, that black employees at the Jacksonville Post Office are younger and take more leave than their white counterparts ... It is likely that there are other different characteristics, patterns of conduct, or reactions to circumstances which exist between groups of employees____ [The defendant’s statistics], thus, permit the Court to infer that race is not a statistically significant factor in the imposition of discipline at the Jacksonville Post Office.

We cannot say that the court’s findings with respect to discipline are clearly erroneous.

IV. INDIVIDUAL CLAIMS

The district court ruled that “as a matter of law ... no class representative or class member who testified established a claim of racial discrimination.” We vacate this judgment for reconsideration in light of our findings on the class claims, discussed above. If the district court on remand should find a pattern or practice of discrimination against the class, this should be taken into consideration in the court’s evaluation of the individual claims. See Donaldson v. Pillsbury Co., 554 F.2d 825, 833 (8th Cir.), cert. denied, 434 U.S. 856, 98 S.Ct. 177, 54 L.Ed.2d 128 (1977).

V. EXHAUSTION OF ADMINISTRATIVE REMEDIES

A prerequisite to the filing of a Title VII lawsuit is the exhaustion of administrative remedies. Brown v. General *1530 Services Administration, 425 U.S. 820, 832, 96 S.Ct. 1961, 1967, 48 L.Ed.2d 402 (1976); 42 U.S.C. § 2000e-16(c). Griffin filed an administrative complaint in 1971 under the then-existing third-party complaint procedure. 8 Civil Service Commission regulations required agencies to set up two different administrative procedures to deal with discrimination complaints. See 5 C.F.R. §§ 713.211-222, .251 (1971). The first procedure was to cover administrative review and determination of complaints of discrimination by individual employees. Id., § 713.211-.222. The second was to cover review of personnel policies at the request of “third parties.” 9 In its cross-appeal, the government argues that the third-party complaint procedure does not provide a proper administrative foundation for civil suit. Here, the trial court rejected this argument.

Two courts that have considered this issue previously have held that the third-party complaint procedure cannot satisfy the requirement of exhaustion of remedies. James v. Rumsfeld, 580 F.2d 224, 227-28 (6th Cir.1978); Caton v. Canal Zone Government, 522 F.Supp. 1, 8 (D.C.Z.1981), aff'd, 669 F.2d 218 (5th Cir.1982). Neither of these decisions, however, addressed the confusing state of the Civil Service Commission regulations governing federal employee complaints at the time Griffin filed his third-party complaint.

Prior to 1977, there was no clear means by which a federal employee could raise class claims. The Civil Service Commission would not allow an individual complaint to include class claims. Barrett v. United States Civil Service Commission, 69 F.R.D. 544, 549, 552 (D.D.C.1975). If one filed an individual complaint with the agency and tried to raise class claims in court, the government would argue that there had been no exhaustion because the class claims were not raised in the agency. See McLaughlin v. Callaway, 382 F.Supp. 885, 891 (S.D.Ala.1974). If one filed a third party complaint, as here, the government would argue that such a complaint was not an appropriate basis for a civil action. See Swain v. Hoffman, 547 F.2d 921, 923 (5th Cir.1977). When Swain and McLaughlin reached the Fifth Circuit simultaneously, the Court held that class action suits could be maintained based on individual complaints. Swain, 547 F.2d at 924; McLaughlin v. Hoffman, 547 F.2d 918, 921 (5th Cir.1977). The problem presented by the regulatory scheme was resolved when the court in Barrett, 69 F.R.D. at 552, ordered the Civil Service Commission to accept class claims in the future. This led to revocation of the third party complaint regulations and issuance of comprehensive *1531 class action regulations, now found at 29 C.F.R. §§ 1613.601 et seq.

Griffin sought in good faith to raise class claims under the obscure regulatory scheme existing prior to 1977. The government should not be allowed to take advantage of ambiguities created by its own regulations. The purpose of the administrative exhaustion requirement is to put the agency on notice of all issues in contention and to allow the agency an opportunity to investigate those issues. The complaint in this case did put the agency on notice of the issues raised and the agency conducted a thorough investigation. Thus the purpose of the exhaustion requirement has been fulfilled. Because the confusing Civil Service Commission regulations in existence in 1971 provided no clear means by which class action claims could be raised at the administrative level, Barrett, 69 F.R.D. at 553, and because the administrative complaint filed by Griffin in this case satisfied the purpose of the administrative exhaustion requirement, we hold that, in this case, the district court correctly found that Griffin’s third-party complaint satisfied the requirement of exhaustion of administrative remedies.

VI. CERTIFICATION OF CLASS ACTION

On cross-appeal the government contends that the district court erred in certifying this class action and in failing to decertify the class prior to judgment. The government argues that the commonality, typicality, and adequacy of representation requirements of Fed.R.Civ.P. 23 were not met. 10

Questions concerning class certification are left to the sound discretion of the district court. Assuming the district court’s determination is made within the parameters of Rule 23, its decision on class certification will be upheld absent an abuse of discretion. Freeman v. Motor Convoy, Inc., 700 F.2d 1339, 1347 (11th Cir.1983). This class was originally certified by the district court on January 9, 1973. The class consists of all past, present, and future black employees of the Jacksonville Post Office. 11 In response to a defense motion to dismiss the class claims in 1982, the district court gave a detailed opinion dated September 8,1982, explaining its reasons for allowing the class to proceed as certified in 1973. We hold that the district court did not abuse its discretion in allowing this suit to proceed as a class action.

“The commonality and typicality requirements of Rule 23(a) tend to merge.” General Telephone Co. of Southwest v. Falcon, 457 U.S. 147,157 n. 13, 102 S.Ct. 2364, 2371 n. 13, 72 L.Ed.2d 740 (1982). In this case, both requirements turn on a two-step inquiry. First, we must determine whether the 22 named plaintiffs raise claims within the periphery of the claims raised by Griffin’s administrative complaint and hence are proper named plaintiffs. Next, if the first inquiry is answered in the affirmative, we must determine whether the claims raised by these 22 plaintiffs satisfy the commonality and typicality requirements.

As discussed above, this suit began with the filing of a third-party complaint by Ernest Griffin under the then-existing Civil Service Commission procedures. The investigation of the third-party complaint considered both issues of discrimination against Griffin personally and issues of discrimination against other named plaintiffs and class members. The investigative report indicated that while Griffin was on the supervisory register, he was not within the “zone of consideration” and therefore was not eligible for promotion. In addition, the investigation included subsequent allegations by Griffin that he had received *1532 a suspension and that he had been threatened with termination. Thus, Griffin was directly affected by the discrimination alleged in his third-party complaint. In addition, an affidavit prepared by Griffin at the request of the Postal Service investigator identified 13 specific areas of discrimination against blacks in the Jacksonville Postal Service. 12

The only issues that may be raised in a class action claim are those issues that were raised by representative parties in their administrative complaints, together with those issues that may reasonably be expected to grow out of the administrative investigation of their claims. Eastland v. Tennessee Valley Authority, 553 F.2d 364, 372 (5th Cir.1977), cert. denied, 434 U.S. 985, 98 S.Ct. 611, 54 L.Ed.2d 479 (1977). It is not necessary that members of the class bring an administrative charge as a prerequisite to joining as co-plaintiffs in the litigation. It is sufficient if they are in a class and assert the same or some of the same issues. As co-plaintiffs, however, they must proceed within the periphery of the issues which Griffin could assert. Oatis v. Crown Zellerbach Corp., 398 F.2d 496, 499 (5th Cir.1968); see Jackson v. Seaboard Coast Line R. Co., 678 F.2d 992, 1005 (11th Cir.1982). In this case, we find that the issues asserted by co-plaintiffs are within the periphery of the issues asserted by Griffin in his administrative complaint and fully investigated by the Postal Service. Thus, the 22 named plaintiffs are appropriately named as co-plaintiffs.

We turn, therefore, to the question of whether the claims of these plaintiffs satisfy the commonality and typicality requirements of Rule 23(a). The Supreme Court has recently reaffirmed that litigants seeking to maintain class actions under Title VII must meet the requirements of Rule 23. Falcon, 457 U.S. at 156, 102 S.Ct. at 2370 (1982). The class claims must be limited to those “fairly encompassed” by the named plaintiffs’ claims. Id. at 158, 102 S.Ct. at 2371. In this case, information in the record by the time the district court issued its final pre-trial order approving class certification reflected that the 22 named plaintiffs had claims relating to tests and other procedures used to get on the supervisory registers; promotion from the supervisory registers; promotions to other high level positions; discipline; training, particularly for higher level positions; awards; and details.

The district court’s opinion of September 8, 1982, gave careful consideration to the requirements of Falcon. The court found that the 22 named plaintiffs had alleged sufficiently diverse employment practices that the court might infer that discriminatory treatment was typical of defendant’s promotion practices and that defendant’s promotion practices were motivated by a pervasive policy of racial discrimination. The court found that these allegations were sufficient to bridge the gap between the named plaintiffs’ individual claims and those made on behalf of a class of all black employees. The district court did not *1533 abuse its discretion in making this determination.

The Postal Service also contends that plaintiffs failed to meet the Fed.R. Civ.P. 23(a) requirement that they fairly and adequately represent the interests of the class. The Postal Service asserts several alleged errors and lack of preparedness by plaintiffs’ counsel.' The adequate representation requirement involves questions of whether plaintiffs’ counsel are qualified, experienced, and generally able to conduct the proposed litigation, and of whether plaintiffs have interests antagonistic to those of the rest of the class. Johnson v. Georgia Highway Express, Inc., 417 F.2d 1122, 1125 (5th Cir.1969). The district court did not abuse its discretion in determining that plaintiffs satisfied the adequate representation requirement.

VII. CONCLUSION

In conclusion, we hold that the district court erred in dismissing plaintiffs’ challenge to the written examinations and in dismissing plaintiffs’ disparate impact claims. We further hold that the district court erred as a matter of law in its evaluation of plaintiffs’ disparate treatment claims regarding promotions to initial level supervisory positions and selection for details. We affirm the district court’s findings of no discrimination under the disparate treatment theory with respect to discipline, awards, and promotions above the initial supervisory level and to non-supervisory positions. We affirm the holdings of the district court that plaintiffs had exhausted their administrative remedies and that this case was appropriately certified as a class action. We vacate the district court’s determination on the individual claims of discrimination for reconsideration in light of its subsequent determinations of the class claims. 13

REVERSED IN PART, AFFIRMED IN PART, AND REMANDED.

Notes

. (1) D’Alver L. Wilson, Distribution Clerk; (2) Charles C. McRae, Clerk; (3) Richard Deloney, Mailhandler; (4) Samuel George, Clerk; (5) Alphonso West, Clerk; (6) Erno Denefield, Mail-handler; (7) Thaddeus E. Raysor, Clerk; (8) Margie L. Raysor, Distribution Clerk; (9) Joe Bailey, Jr., Clerk; (10) Andrew Edwards, Carrier; (11) Claude L. Smith, Clerk Technician; (12) Smith M. Morgan, Clerk Technician; (13) Jesse L. Wilcox, Clerk; (14) Harvey J. Harper, Clerk; (15) Joyce A. Scales, Clerk; (16) Albert Jackson, Jr., Mailhandler; (17) Kenneth A. Rosier, Distribution Clerk; (18) Andrew D. Martin, Jr., Clerk; (19) James Williams, Clerk; (20) John H. Fowler, Mechanic; (21) Doris D. Galvin, Relief Window Clerk.

. Year Race Levels 1-6 Levels 7 & Above % at 7+

1969 Black 720 10 1%

Other 1181 171 13%

1970 Black 711 12 2%

Other 1133 171 13%

1971 Black 666 11 2%

Other 1134 166 13%

1972 Black 647 • 11 2%

Other 1068 166 13%

1973 Black 756 17 2%

Other 1230 162 12%

1974 Black 753 23 3%

,07= 1975 Black 744 27 4%

Other 1213 185 13%

1976 Black 725 28 4%

0ther 1159 186 140/0

1977 Black 712 28 40/0

0ther 1122 181 140/0

1978 Black 702 43 60/0

0ther 1103 183 140/0

1979 BIack 680 68 90/0

0ther 1106 227 170/0

1980 BIack 734 31 6%

0ther 1189 195 140/0

1981 BIack 728 46 6°/o

0ther 1225 183 130/0

Source: Table 12, Plaintiffs' Exhibit #1.

1968 1084 583 86 3 .1481 D-121

1970 1070 580 123 16 .4160 D-10

1973 1053 617 216 49 .1206 D-ll

1974 1055 652 155 71 .1399 D-01

1975 982 627 230 84 .3246 D-06

May 77 949 641 81 26 .2436 D-03

Dec. 77 949 641 89 32 .6700 D-03

1978 PM 933 611 14 5 .1708

Vab 933 611 16 9 .4413

CS 933 611 105 59 .1811

MP 933 611 116 93 .9492

*D-12 indicates that in the number to the left the decimal point should be followed by 12 zeroes, and so on.

Source: Table 2, Plaintiffs’ Exhibit #1.

Incidences of Disciplinary Action by Race, Percentage of Discipline and Year.

Year Black Other Total Probability

# % # %

1969 66 54% 57 46% 123 .001

1970 77 66% 39 34% 116 .001

1971 - 63 65% 34 35% 97 .001

1972 51 65% 28 35% 79 .001

1973 38 55% 31 45% 69 .001

1974 100 51% 98 49% 198 .001

1975 324 64% 182 36% 36 .001

1976 161 56% 124 44% 285 .001

1977 22 67% 11 33% 33 .001

1978 204 58% 144 41% 348 .001

1979 154 55% 124 45% 278 .001

1980 178 52% 165 48% 343 .001

1981 305 52% 285 48% 590 .001

Source: Table .1, Plaintiffs' Exhibit #1.

. Fifth Circuit decisions prior to October 1, 1981, are binding in this Circuit and cannot be overruled except by the court acting en banc. Bonner v. City of Prichard, 661 F.2d 1206, 1207 (11th Cir.1981).

. We note that while many subsequent Fifth Circuit cases have followed Pouncy, see e.g., Carroll v. Sears, Roebuck & Co., 708 F.2d 183, 188 (5th Cir.1983) ("The use of subjective criteria to evaluate employees in hiring and job placement decisions is not within the category of facially neutral procedures to which the disparate impact model is applied."), at least one post-Pouncy Fifth Circuit decision has applied disparate impact analysis to a subjective promotional system. Page v. U.S. Industries, Inc., 726 F.2d 1038 (5th Cir.1984). In Page, the Court noted the Pouncy decision but stated:

It is clear that a promotional system which is based upon subjective selection criteria is not discriminatory per se. Consequently, such a system can be facially neutral but yet be discriminatorily applied so that it impacts adversely on one group. In using a disparate impact analysis, the district court pointed out that many of this Court’s decisions examine the classwide impact of a subjective promotional system. See, e.g., James v. Stockham Valves & Fittings Co., 559 F.2d 310 (5th Cir. 1977), cert. denied, 434 U.S. 1034, 98 S.Ct. 767, 54 L.Ed.2d 781 (1978); Rowe v. General Motors Corp., 457 F.2d 348 (5th Cir. 1972). We agree with the district court’s assessment that the subjective promotional system in this case indeed may have had a classwide impact. Thus, we examine the defendant’s promotional system under the disparate impact model as well.

Id. at 1046. We note also the recent decision of the D.C. Circuit in Segar v. Smith, 738 F.2d 1249, 1288 n. 34 (D.C.Cir.1984) ("Though these practices arguably encompass some subjective judgments as to agents’ performance, we find that disparate impact appropriately applies to them.”)

. E.g., 401 U.S. at 430, 91 S.Ct. at 853 ("practices, procedures, or tests”); id. at 431, 91 S.Ct. at 853 ("practices”); id. at 432, 91 S.Ct. at 854 (“employment procedures or testing mechanisms"); id. ("any given requirement”).

. Griffin appealed to the Civil Service Commission following his discharge in 1976. This properly exhausted administrative remedies as to his discharge, but could not support all of this suit which challenges both discharge and other personnel practices dating back to 1969.

. 5 C.F.R. § 713.251: Third party allegations of discrimination.

(a) COVERAGE. This section applies to general allegations by organizations or other third parties of discrimination in personnel matters within the agency which are unrelated to an individual complaint of discrimination subject to §§ 713.211 through 713.222.

(b) AGENCY PROCEDURE. The organization or third party shall state the allegation with sufficient specificity so that the agency may investigate the allegation. The agency may require additional specificity as necessary to proceed with its investigation. The agency shall establish a file on each general allegation, and this file shall contain copies of all material used in making the decision on the allegation. The agency shall furnish a copy of this file to the party submitting the allegation and shall make it available to the Commission for review on request. The agency shall notify the party submitting the allegation of its decision, including any corrective action taken on the general allegations, and shall furnish to the Commission on request a copy of its decision.

(c) COMMISSION PROCEDURES. If the third party disagrees with the agency decision, it may, within 30 days after receipt of the decision, request the Commission to review it. The request should be in writing and shall set forth with particularity the basis for the request. When the Commission receives such a request, it shall make, or require the agency to make, any additional investigations the Commission deems necessary. The Commission shall issue a decision on the allegation ordering such corrective action, with or without back pay, as it deems appropriate.

. Fed.R.Civ.P. 23(a) Prerequisites to a Class Action. One or more members of a class may sue or be sued as representative parties on behalf of all only if (1) the class is so numerous that joinder of all members is impracticable, (2) there are questions of law or fact common to the class, (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class, and (4) the representative parties will fairly and adequately protect the interests of the class.

. As originally certified, the class included applicants for employment, but applicants were dropped from the class prior to trial.

. The 13 identified areas were:

(1) Lack of training and development in supervisory positions for blacks.

(2) No black employees are in Customer Relations.

(3) No blacks are assigned as Postal Inspection Aides.

(4) Poor representation of blacks on Postmaster's Promotional Advisory Board.

(5) Little or no recognition given blacks with college degrees. Qualified black employees with college degrees have not been trained for higher positions.

(6) Lack of advancements have caused blacks to leave Postal Service.

(7) Only token blacks appointed to Level 7 and above.

(8) No blacks in Finance Office, Transfer Clerks, Administration, Training Section and Safety Division. Only one black in Personnel.

(9) Caucasians not on supervisor’s roster are being used on 204B [details].

(10) Qualified blacks in Custodian Section not promoted. Caucasian brought from outside for foreman position with no experience.

(11) Postmaster in direct violation of Executive Order No. 11478 C(2), 3 a, d, 5, 8, 9 and 13.

(12) No blacks appointed to Postmaster's Administrative Staff.

(13) No black supervisors assigned to first floor at West Bay Annex.

. The plaintiffs objected to the district court order requiring them to pay the government’s costs of $62,513.16. This award of costs is, of course, vacated along with the decision on the merits.