This appeal represents the final phase of what has been protracted and difficult litigation for both sides. Plaintiffs allege that the civilian employee promotion system used at Kelly Air Force Base has operated to discriminate against Hispanic males. After a prolonged discovery and motions period, the district court granted the defendant's motion for summary judgment and plaintiffs appealed. Upon initial submission, we ordered a limited remand so that the district court could explain its reasoning with respect to its summary judgment. The district court prepared an order explaining its decision and the parties filed supplemental briefs. We must now decide the issue of whether the district court erred in granting summary judgment for the defendant, given the limited evidence before it.

Plaintiffs Jesus G. Munoz and Manuel Munoz, Jr. are Hispanic males, brothers, employed as part of the civilian workforce at Kelly Air Force Base near San Antonio, Texas [Kelly]. They brought this suit on behalf of themselves and all Hispanic male civilian employees at Kelly under Federal Rule of Civil Procedure 23(b)(2). Plaintiffs allege that the promotion system used by the Air Force for civilian employees has a disparate impact on Hispanic males, i.e. that the system results in fewer Hispanic males receiving promotions than would be expected based on the proportion of the civilian workforce at Kelly that they comprise.

An explanation of plaintiffs' claims first requires an understanding of the general context in which hiring and promotion takes place at Kelly. Civilian employment at Kelly is organized on the "GS" (General Service) scale, a salary and promotions grid in common use throughout civilian federal employment. Each GS level, or grade, represents a salary range. The GS level to which an employee is assigned depends upon such factors as education level, skill level, time in service, and degree of authority of the position he occupies. Over the course of a career in federal civilian employment, an employee may occupy several different GS levels or steps within a GS level. Certain GS levels are not open to employees without particular qualifications (e.g. a college degree or its equivalent). As a general matter, each job opening is allocated to a particular GS level or range of levels, thus setting the maximum salary that position could accrue. A federal civilian job also has skills requirements and responsibilities attached to it that in part define its GS range.

At Kelly, civilian promotions are handled in part by a Merit Promotion Plan that includes an automated system called the "Personnel Placement and Referral System," or PPRS. Under PPRS, employees need not submit applications for promotions. Rather, as a position becomes available, PPRS considers all eligible employees within the defined area of consideration for the position (e.g. Kelly Air Force Base or the entire Air Force). PPRS recursively eliminates employees under increasingly specific job requirements until the desired number of candidates is reached. PPRS thus works like a funnel, at first considering all nominally eligible employees for a promotion and then narrowing the field based on successively more detailed requirements until a short, ranked list is generated. Each stage of this narrowing is known as a "progression level factor," or PLF. Ties between employees are broken by reference to appraisal scores, awards, and service computation date, in order. The list is hand-checked and then forwarded to the selecting official, who chooses one of the employees for the promotion.

The automated program is not without subjective elements. At the beginning of the promotion process, three-person teams establish and rank the job skills relevant to the position. This ranked list is called a "Promotion Evaluation Pattern," or PEP. The PLFs used by the automated programto narrow the field for a given promotion are derived from these PEPs. Furthermore, within the automated PPRS program, ties between eligible employees are broken in part by appraisal scores and awards and service computation dates. An employee's appraisal scores and receipt of any awards depend, to a large degree, on the discretion of his supervisors. Lastly, after a finite list of names for a promotion has been prepared by the PPRS, a selecting officer chooses one employee from the group. Though the officer's range of choice is limited to the list derived from the PPRS, the actual selection from within the group is left to the selecting officer's discretion. Thus, promotions at Kelly comprise both subjective and objective components that are significantly intertwined.

Plaintiffs filed this action in 1985, alleging sex and national origin discrimination in violation of Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. 2000 et seq. Plaintiffs filed this case as a class action under Federal Rule of Civil Procedure 23(b)(2) on behalf of themselves and all other Hispanic male civilian employees, contending that the promotion system at Kelly had an adverse disparate impact on Hispanic males.1 In February of 1989, the district court granted summary judgment for defendant, finding that plaintiffs had failed to exhaust their administrative remedies. This Court reversed that initial summary judgment and remanded for further proceedings. Munoz v. Aldridge, 894 F.2d 1489 (5th Cir. 1990).

On remand, the case was referred to a magistrate judge who held a class certification hearing and then recommended that the case be certified as a class action. The class was defined as "all Hispanic male employees, GS-09 to GS-14 at the SA-ALC, who on or after 2/26/80 were eligible for promotion to positions at grade GS-11 and above at SA-ALC which are covered by Merit Promotion Certificates produced by the Central Civilian Personnel Office at SA-ALC, up to and including the date of 09/26/91."2 The class was certified in accordance with the magistrate's report.

During discovery, plaintiffs sought certain information regarding the Air Force's promotion procedures, including access to the algorithm used in the automated PPRS process. After an in camera review of the algorithm, the district court denied plaintiffs' request. No objections were filed to that denial. Experts for both sides filed reports. Discovery was re-opened in order to allow plaintiffs' expert, Dr. Benz, to file two additional reports on which he was afterwards deposed by the defense. Plaintiffs also submitted affidavits from certain class members explaining their personal experiences with the promotion system at Kelly. Two plaintiffs, Manuel Munoz, Jr. and Michael Galvan, submitted additional affidavits containing partial analyses of Kelly's promotion data. After discovery was closed for the second time, upon defendant's motion, the district court ordered summary judgment against plaintiffs.

Plaintiffs appealed the summary judgment to this court. We ordered a limited remand so that the district court could explain its reasoning. We now have before us the original appeal, the district court's order explaining its reasons for granting summary judgment, and supplemental briefs filed by the parties in response to the district court's explanation.

The primary issues raised in this case are whether the district court properlyexcluded plaintiffs' expert's affidavits as being unreliable, whether plaintiffs' case should survive summary judgment even without their expert's statistical evidence, and whether any individual claims brought by plaintiffs survive the summary judgment order. Plaintiffs also argue that various rulings during the discovery period improperly deprived them of the opportunity to develop their case.

The litigation in the trial court has concentrated on plaintiffs' disparate impact class claims. As discussed below, issues have been raised in this appeal regarding a disparate treatment class claim and both disparate impact and disparate treatment individual claims.

In the context of Title VII litigation, we recognize two types of discrimination claims: disparate treatment and disparate impact. Disparate treatment refers to deliberate discrimination in the terms or conditions of employment, in this case, promotion, on account of race, national origin, or gender. The prima facie elements of a claim for disparate treatment are: (1) that the plaintiff is a member of a protected class under the statute; (2) that he applied and was qualified for a job or promotion for which his employer was seeking applicants; (3) that, despite his qualifications, he was rejected; (4) that afterwards the position remained open and the employer continued to look for candidates with plaintiff's qualifications. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 (1973).

Once that showing has been made, the burden of production shifts to the employer to articulate a legitimate, non-discriminatory reason for the employment action. See id. at 802-03. The burden of persuasion, however, remains at all times with the plaintiff. See Texas Dept. of Community Affairs v. Burdine, 405 U.S. 248, 253 (1981). In order to win his disparate treatment claim, the plaintiff must then demonstrate that the reason articulated by the employer was pretextual, see McDonnell Douglas, 411 U.S. at 805, meaning a pretext for discrimination, per se. See St. Mary's Honor Ctr. v. Hicks, 509 U.S. 502, 510-12 (1993). The plaintiff still bears the burden of showing intent on the employer's part. See id.

Disparate treatment claims can be brought as class actions as well. Plaintiffs in a class action disparate treatment case must show a "pattern or practice" of discrimination by the employer, i.e. that "racial discrimination was the company's standard operating procedure-the regular rather than the unusual practice." Teamsters v. United States, 431 U.S. 324, 336 (1977). Proving a pattern or practice is necessary to establishing a prima facie case in a disparate treatment class action: "Proving isolated or sporadic discriminatory acts by the employer is insufficient to establish a prima facie case . . . ." Cooper v. Federal Reserve Bank of Richmond, 467 U.S. 867, 875-76 (1984).

Disparate impact claims, recognized in Griggs v. Duke Power Co., 401 U.S. 424 (1971), do not require proof of intent to discriminate. Instead, they focus on facially neutral employment practices that create such statistical disparities disadvantaging members of a protected group that they are "functionally equivalent to intentional discrimination." Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 987 (1988). Plaintiffs must identify specific practices as being responsible for any observed disparities, see Johnson v. Uncle Ben's, Inc., 965 F.2d 1363, 1367 (5th Cir. 1992), cert. denied, 511 U.S. 1068 (1994), and must conduct a systemic analysis of those employment practices in order to establish their case. See Black Fire Fighters Ass'n v. City ofDallas, 905 F.2d 63, 63 (5th Cir. 1990). Disparate impact claims may be brought by either individual plaintiffs or a class. In either case, the evidence will focus on the degree of statistical disparity between protected and non-protected workers in regards to employment or promotion.

We review a grant of summary judgment de novo. See Frazier v. Garrison Indep. Sch. Dist., 980 F.2d 1514,1520 (5th Cir. 1993); E.E.O.C. v. Southern Publishing Co., 894 F.2d 785, 789 (5th Cir. 1990). Exclusion of expert testimony under Federal Rules of Evidence Rule 702 is within the traditional discretion of the trial court, however, see Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 592-93 (1993), and we review it only for an abuse of discretion which amounts to manifest error. See Boyd v. State Farm Ins. Cos., 158 F.3d 326, 331 (5th Cir. 1998) ("With respect to expert testimony offered in the summary judgment context, the trial court has broad discretion to rule on the admissibility of the expert's evidence and its ruling must be sustained unless manifestly erroneous"). We also review orders involving discovery under a deferential abuse of discretion standard. See Geiserman v. MacDonald, 893 F.2d 787, 789 (5th Cir. 1990); see also Hodges v. United States, 597 F.2d 1014, 1018 (5th Cir. 1979).

Because this case involves the exclusion of expert testimony for the purposes of a summary judgment determination, the applicable standards of review overlap somewhat. We will therefore review the district court's exclusion of plaintiffs' expert's evidence and all discovery-related rulings for abuse of discretion, and then review de novo the grant of summary judgment based on the evidence properly before the district court. See General Elec. Co. v. Joiner, 522 U.S. 136, 118 S. Ct. 512, 517 (1997) ("The question of admissibility of expert testimony is not . . . an issue of fact, and is reviewable under the abuse of discretion standard"); Curtis v. M & S Petroleum, Inc., 174 F.3d 661, 667-68 (5th Cir. 1999) ("We must first review the trial court's evidentiary rulings under an abuse of discretion standard. . . . Then, with the record defined, we must review de novo the order granting summary judgment as a matter of law").

Claims of disparate impact under Title VII must, of necessity, rely heavily on statistical proof. See Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 987 (1988). Plaintiffs in this case have relied almost exclusively on the statistical analysis contained in the affidavits and testimony of their sole expert, Dr. Benz. On limited remand, the district court explained that it found Dr. Benz' analysis to be unreliable and so did not consider it in reaching the summary judgment ruling.3

The district court acted well within its discretion in evaluating the reliability of Dr. Benz' evidence at the summary judgment stage: "In considering a Rule 56(c) motion opposed by expert testimony, the trial court has broad discretion to rule on the admissibility of the expert's evidence . . . [and] may inquire into the reliability and foundation of any expert's opinion . . . ." Washington v. Armstrong World Indus., Inc., 839 F.2d 1121, 1123 (5th Cir. 1988).

If the basis for an expert's opinion is clearly unreliable, the district court may disregard that opinion in deciding whether a party has created a genuine issue of material fact. See Berry v. Armstrong Rubber Co., 989 F.2d 822, 824 (5th Cir. 1993), cert. denied sub nom, Cooper v. Armstrong Rubber Co., 510 U.S. 1117 (1994); see also Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 596 (1993) (if "the trial court concludes that the scintilla of [expert] evidence presented supporting a position is insufficient to allow a reasonable juror to conclude that the position more likely than not is true, the court remains free to . . . grant summary judgment"). Both the determination of reliability itself and the factors taken into account are left to the discretion of the district court consistent with its gatekeeping function under Fed. R. Evid. 702. See Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 119 S. Ct. 1167, 1176 (1999).

In its order following remand, the district court explains at length its reasons for rejecting Dr. Benz' reports and testimony for purposes of its summary judgment determination. After a careful review both of the order and the record, we cannot conclude that the district court abused its discretion in excluding plaintiffs' expert reports and testimony.

The facts showing the insufficiency of Dr. Benz' evidence range from particular miscalculations to his general approach to the analysis. For example, in one table Dr. Benz relied on in reaching his conclusions, columns representing Hispanic males and all other employees should add up to 100%, but in seven out of twenty cases they do not. Some of those errors could be attributable to rounding, but no satisfactory explanation was given for totals as low as 25%, 16.67%, and 8.33%. Any reliance on that table was necessarily misplaced.

On a broader level, Dr. Benz' analysis fails to meet the standards for admissibility of scientific evidence in that the district court found that his methods were not in accord with those of experts in his field. See Kumho Tire Co., 119 S. Ct. at 1176 ("The objective of [the gatekeeper requirement] . . . is to make certain that an expert . . . employs in the courtroom the same level of intellectual rigor that characterizes the practice of an expert in the relevant field"). Dr. Benz began his analysis with the assumption that Kelly's promotion system discriminated against Hispanic males, an indicator that he lacked the necessary objectivity to make his analyses credible. See Viterbo v. Dow Chemical Co., 826 F.2d 420, 423 n.2 (5th Cir. 1987). Dr. Benz also stated that discrimination was the "cause" of the disparities he had observed, a statement which he later recanted as "overzealous" since statistics can show only correlation and not causation. See, e.g., Tagatz v. Marquette Univ., 861 F.2d 1040, 1044(7th Cir. 1988).

In his depositions, he admitted to failing to consider other variables such as education and experience as explanations for any observed discrepancy between promotion rates and to not performing a multiple regression analysis. See, e.g., Tagatz v. Marquette Univ., 861 F.2d 1040, 1045 (7th Cir. 1988) (holding that failure to control for other explanatory variables makes an expert's table "essentially worthless"). Finally, Dr. Benz relied on the plaintiffs' compilations of data, which gives rise to a "common-sense skepticism" regarding the expert's evaluation, see Christophersen v. Allied-Signal Corp., 939 F.2d 1106, 1111(5th Cir. 1991) (en banc), cert. denied, 503 U.S. 912 (1992), overruled on other grounds by Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 587 n.5 (1993), and did not seek to verify the information presented to him. See Sheats v. Bowen, 318 F. Supp. 640, 644 (D. Del. 1970).

Taken cumulatively, the problems with Dr. Benz' expert evidence indicate that his expert testimony could be unreliable. Consistent with the role of the district court as "gatekeeper" for scientific evidence, see General Elec. Co. v. Joiner, 522 U.S. 136, 118 S. Ct. 512, 517 (1997), it was not an abuse of discretion on the part of the district court to exclude Dr. Benz' testimony and reports from the summary judgment evidence.4

C. Sufficiency of plaintiffs' other evidence

Having determined that it was not an abuse of discretion for the district court to disregard Dr. Benz' expert reports and deposition testimony, we must now determine whether plaintiffs had other sufficient evidence before the court to survive defendant's summary judgment motion.

A party is entitled to summary judgment if, when the evidence is viewed in the light most favorable to the nonmovant, there are no genuine issues of material fact in dispute. See Frazier v. Garrison Indep. Sch. Dist., 980 F.2d 1514 (5th Cir. 1993); Fed.R.Civ.P. 56(c). To withstand a motion for summary judgment, the nonmoving party must come forward with evidence to support each essential element of its claims on which it will bear the burden of proof at trial. See Celotex Corp. v. Catrett, 477 U.S. 317 (1986); National Ass'n of Gov't Employees v. City Pub. Serv. Bd. of San Antonio, 40 F.3d 698, 712 (5th Cir. 1994). A dispute about a material fact (i.e. one which might affect the outcome of the trial) is "genuine" only "if the evidence is such that a reasonable jury could return a verdict for the nonmoving party." Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248 (1986).

With Dr. Benz' evidence excluded, plaintiffs have little else to rely on in attempting to overcome summary judgment on either disparate impact or disparate treatment claims. The primary evidence they cite to consists of indicators of disparate impact gleaned from the defense experts' reports and two declarations filed by individual plaintiffs Manuel Munoz, Jr. and Michael Galvan. In its order following remand, the district court rejected these latter declarations as untimely filed and endorsed defendant's objections to them.

Plaintiffs rely on reports by defense experts Dr. Wayne Cascio and William Ruch. Plaintiffs point out that according to Dr. Cascio's data, Hispanic males suffered a disparate impact arising from defendant's use of appraisal scores to break ties between employees in the PPRS system that was statistically significant for the year 1987. According to plaintiffs, Dr. Cascio's analysis also shows that Hispanic males received fewer awards than would be expected at statistically significant levels for GS-11 in 1991 and GS-13 in 1987. However, such results do not create a genuine issue of material fact regarding disparate impact or disparate treatment either on their own or taken in combination with all other evidence before the district court. The data points which could suggest disparate impact are isolated in the record and do not support the plaintiffs' allegations of systemic discrimination. See Johnson v. Uncle Ben's, Inc., 965 F.2d 1363, 1367 (5th Cir. 1992) (holding that prima facie case of disparate impact requiresa "systemic analysis of the [disparate] effects of all promotional criteria for each rank").

Plaintiffs also rely on two declarations filed by class members. The magistrate struck Manuel Munoz, Jr.'s affidavit filed in October of 1995 in support of plaintiffs' second motion for partial summary judgment. The affidavit was struck on three grounds. First, it was untimely, seeking to introduce new statistical evidence after discovery had already been reopened and closed once to allow plaintiffs to submit additional expert evidence. Second, it used a methodology (the "adverse impact barometer") which the magistrate and plaintiffs' own expert found to be unreliable. Third, since he is not an expert in statistics, Mr. Munoz was unqualified to advance the statistical evidence contained in the affidavit.

We review a district court's striking of an affidavit for failure to comply with its scheduling order under a deferential abuse of discretion standard. See Geiserman v. MacDonald, 893 F.2d 787, 790 (5th Cir. 1990)("a trial court's decision to exclude evidence as a means of enforcing a pretrial order 'must not be disturbed' absent a clear abuse of discretion") [citation omitted]. It was not an abuse of discretion to strike an affidavit seeking to introduce new statistical evidence after an extension for the filing of such reports had already passed. Because the Munoz affidavit was properly struck, it was not before the district court and we do not consider it now as part of plaintiffs' summary judgment evidence. See Barrett v. Atlantic Richfield Co., 95 F.3d 375, 382 (5th Cir. 1996).

The other affidavit in issue was filed by Michael Galvan, also a member of the plaintiff class. The affidavit was filed well after the deadline for statistical reports, after even the Munoz affidavit. In his affidavit, Mr. Galvan states that certain job series had a higher concentration of Hispanic employees and that certain other job series had a higher concentration of non-Hispanic employees. He further states that there were fewer promotion opportunities above the GS-13 level in the job series with the higher concentration of Hispanic employees. Two charts containing bar graphs depicting promotion by grade for various job series are attached.

Regardless of whether Mr. Galvan could be determined competent to offer statistical evidence and whether the lateness of the affidavit would bar its consideration, the affidavit has deficiencies on its face that render it inadequate as summary judgment evidence for the plaintiffs. First, Mr. Galvan's bar charts reflect the number of Hispanic employees who received certain promotions, not the number of Hispanic males in accordance with the class definition in this case. See Lopez v. Laborers Int'l Union Local 18, 987 F.2d 1210, 1213-15 (5th Cir. 1993) (describing the importance of using the correct comparison pool in evaluating statistical significance of apparent disparate impact). Second, Mr. Galvan does not state whether the series depicted in his bar graphs represent the entire universe of promotions during the relevant period, making statistical comparisons impossible. Third, allegations regarding concentration of Hispanic employees in certain job series relate more to hiring and initial placementthan to promotion decisions, the subject of this lawsuit.

Mr. Galvan does not purport to show, nor is there evidence elsewhere, that any concentration of Hispanic employees in particular job series was due to discrimination rather than differences in which jobs Hispanic employees applied for or other race- and gender-neutral explanations for any perceived differences. See Watson v. Fort Worth Bank and Trust, 487 U.S. 977, 992 (1988) ("It is completely unrealistic to assume that unlawful discrimination is the sole cause of people failing to gravitate to jobs and employers in accord with the laws of chance"). Finally, Mr. Galvan does not(and indeed could not, if he is not an expert) draw any statistical inferences from the data he presents. The court is unable to perform the necessary analyses for the plaintiffs, particularly without adequate information regarding the universe of promotions included in the affidavit, the selection criteria used in compiling the data, or other such vital information. Upon reviewing the affidavit and other filings in the case, we find that Mr. Galvan's affidavit is not sufficient to raise a genuine issue of material fact regarding plaintiffs' claims.

Taken as a whole, plaintiffs' evidence properly before the district court does not raise a genuine issue of material fact as to either disparate impact or disparate treatment class-wide claims.

D. Failure to identify alleged discriminatory practices with particularity

Defendant alleges that plaintiffs have failed to identify the particular employment practices they challenge and that instead, this case represents an attack on civilian employment generally at Kelly. Plaintiffs counter that their attack is adequately focused.

It is a matter of settled law that a disparate impact class action is not the proper mechanism with which to attack the cumulative effects of an employer's practices. See Pouncy v. Prudential Ins. Co. of America, 668 F.2d 795, 800 (5th Cir. 1982); see also Anderson v. Douglas & Lomason Co., Inc., 26 F.3d 1277, 1284 (5th Cir. 1994), cert. denied, 513 U.S. 1149 (1995). Plaintiffs must provide evidence "isolating and identifying the specific employment practices that are allegedly responsible for any observed statistical disparities." Wards Cove Packing Co., Inc. v. Atonio, 490 U.S. 642, 109 S. Ct. 2115, 2124 (1989). In the present case, plaintiffs have focused their challenge on a number of different aspects of the promotion system at Kelly over the course of this litigation.5

While defendant objects with justification to the shifting focus of plaintiffs' claims, part of the difficulty in this case arises from the nature of civilian promotions at Kelly. Plaintiffs cannot bring a suit challenging an entire employment system; yet where a promotion system uses tightly integrated and overlapping criteria, it may be difficult as a practical matter for plaintiffs to isolate the particular step responsible for observed discrimination.

In this case, it may be that the overlap between the secret algorithm used by PPRS with individual subjective decisions such as awards during service and selection among a list of candidates would make it difficult to separate out the statistical impact of each portion of the promotion decision. However, plaintiffs have not provided an adequate statistical study of civilian promotions at Kelly from which a court could draw such conclusions. Because we find that plaintiffs have failed to adduce sufficient evidence to create a genuine issue of material fact as to disparate impact or treatment, we need not rule on whether they failed to plead with sufficient specificity the particular aspects of the promotion system they contend are responsible for the discrimination they allege.

E. Discovery limitations and their impact on plaintiffs' ability to develop their case

Plaintiffs object to a number of discovery limitations imposed by the magistrate judge and district court over thecourse of this litigation. As a preliminary matter, we note plaintiffs' obligation to object contemporaneously to adverse discovery orders. Without such objections, we can only review for plain error. See Douglass v. United Svcs. Automobile Ass'n, 79 F.3d 1415 (5th Cir. 1996) (en banc). District courts have considerable discretion in managing discovery. See Geiserman v. MacDonald, 893 F.2d 787, 789 (5th Cir. 1990); see also Fed.R.Civ.P. 16(b). That broad discretion is particularly important in complex litigation such as this where fairness to all litigants as well as judicial economy require that the length and burdensomeness of discovery and motions be limited. However, that discretion is not unlimited, of course, and a reviewing court may reverse.

The primary discovery ruling plaintiffs object to is the district court's decision to deny access to the Air Force algorithm used in the PPRS process. The district court ordered the algorithm sealed following an in camera review of the computer program in which it determined that the algorithm did not contain any evidence of discrimination on the part of the Air Force. Plaintiffs did not object at the time the algorithm was sealed, and it was eventually returned to the Air Force. Due to the lack of contemporaneous objection, we can only review for plain error. Under that standard, we find that the district court was within its discretion in refusing access to the algorithm.

It is unlikely that denial of access to the algorithm unduly prejudiced plaintiffs' claims. Defendant had already supplied detailed information on the overall promotion system and the inputs used by the PPRS automated system. Plaintiffs also already knew what factors were used to break ties between employees and what weight was assigned to each. Denial of access to the algorithm arguably could make it more difficult to identify with specificity the aspects of Kelly's promotion system responsible for any observed disparate impact. However, plaintiffs' claim does not fail on those grounds and therefore a remand to the district court solely because of its denial of access to the algorithm would be inappropriate.

Plaintiffs also object to several other discovery orders by the magistrate and district court. As noted above, we must allow for the district court's discretion in discovery matters, especially in complex litigation. After a thorough and careful review of the record and plaintiffs' objections, we do not find reversible error in any of the district court's discovery orders.

Plaintiffs contend that even if this court upholds the summary judgment as against the class action, the individual plaintiffs' claims should survive and should be remanded for trial. The defense counters that this action was brought solely as a class action and that no individual claims were ever pled. On the face of the complaint, no individual claims seem to be alleged, and the bulk of litigation below seems to have been premised on the existence of only a class claim.

However, this case has appeared in this court once before and on that hearing the panel decision makes note of individual claims in addition to the class claim on the face of the complaint. See Munoz v. Aldridge, 894 F.2d 1489, 1491-95 (5th Cir. 1990) (remanding for further consideration class and individual claims). Without full briefing from the parties on this issue, our careful review of the record has not revealed whether the individual claims referred to in the previous opinion from this court were officially dropped from the litigation.6 In order to give due considerationto plaintiffs' claims given this uncertainty, we examine the summary judgment assuming that some form of individual claims have been brought. We find, however, that any individual claims would also fail to survive summary judgment.

In spite of the inclusion of several individual affidavits in support of plaintiffs' second motion for partial summary judgment (advanced as evidence of class discrimination according to the plaintiffs' characterization), the only possible individual claims would be those appearing in the complaint, i.e. those of Jesus and Manuel Munoz, Jr. See, e.g., Coon v. Georgia-Pac. Corp., 829 F.2d 1563 (11th Cir. 1987) (limiting claims pursuant to Fed.R.Civ.P. 8(a) to those stated in the complaint). It is well established that a class representative cannot have interests adverse to the class, see Fed.R.Civ.P. 23(a)(4), nor can individual liability be adjudged before class liability is determined. See Bernard v. Gulf Oil Corp., 841 F.2d 547, 569 (5th Cir. 1988).

The preliminary question regarding any individual claims is whether we have jurisdiction at all, since summary judgment only as to the class claim would not dispose of the entire case and therefore would not be an appealable final order. Such is not the case, however. In his last motion for summary judgment, defendant requested dismissal of plaintiffs' complaint as a whole, thus putting any individual plaintiffs on notice that summary judgment could be pending against them. The district court's order dismissed "all plaintiffs' claims" with prejudice, disposing of any individual claims along with the class claim. The summary judgment is thus a final and appealable order and its propriety is legitimately before us.

Any individual claims based on disparate impact would fail both because they would be duplicative of class claims in the same action and also because the same statistical proof deemed inadequate for the class would likewise fail to forestall summary judgment on individual claims. See Matsushita Elec. Indus. Co., Ltd. v. Epstein, 516 U.S. 367, 379 (1996) (defeat of class claim is binding on class members' individual claims based on same allegations and facts).

As noted earlier, a prima facie case of individual disparate treatment in promotion must show: that the plaintiff is a member of a protected class, that he was qualified for and applied for a promotion, and that the position remained open and the employer continued to look for candidates with similar qualifications. See McDonnell Douglas Corp. v. Green, 411 U.S. 792, 802 (1973). The burden of persuasion remains with the plaintiff, see Texas Dept. of Community Affairs v. Burdine, 405 U.S. 248, 253 (1981), and a reviewing court may look to all the evidence submitted to determine whether defendant met its burden of producing a legitimate reason for the employment action. See In re Royale Airlines, Inc., 98 F.3d 852, 856 (5th Cir. 1996) (judgment may be affirmed on any grounds in the record).

The named plaintiffs arguably made out a prima facie case on their individual claims, since the complaint and later-filed affidavits allege the required elements for disparate treatment: that the named plaintiffs are members of a protected class and that they were denied promotions. However, all of the promotion denials challenged by the class, including those of the named plaintiffs, arose from the Merit Promotion Plan at Kelly. The promotion plan itself, including the PPRS algorithm which the district court determined was not probative of any claims of discrimination, is the "reason" for the denial of promotions in issue. Plaintiffs would have to show, therefore, either that there is a genuine issue as to whether the promotion plan indicates intentional discriminationor that the promotions denied them were somehow capable of being isolated from the promotion system and that purposeful discrimination was behind those employment actions.

Nothing in the record adequately alleges or shows purpose on the part of defendant. A complete failure of proof as to one element requires summary judgment against the entirety of the claim. See Celotex Corp. v. Catrett, 477 U.S. 317 (1986). Because plaintiffs have not shown any evidence going to purpose, any individual claims based on disparate treatment fail to withstand summary judgment.

We note that the failure of proof on the class claim does not bar all individual class members from bringing their own suits, provided that they do not base their claims solely on issues already adjudicated in this action and that they can show individualized proof of discrimination. See Cooper v. Federal Reserve Bank of Richmond, 467 U.S. 867, 880 (1984).7

This has been protracted and difficult litigation for both sides. We fully appreciate the importance of this case, both in terms of the resources dedicated to it and its impact on the plaintiff class-members' lives. After fourteen years in the federal courts, the case comes before us now on a narrow range of issues for which the scope of our review is narrowly circumscribed. The discovery rulings and the exclusion of plaintiffs' key expert evidence were not an abuse of discretion by the district court. Without statistical proof, dismissal of the class claims was appropriate. Any individual claims have not produced sufficient proof of disparate impact or of purpose necessary to a disparate treatment claim to survive summary judgment.

The summary judgment against all plaintiffs' claims is hereby AFFIRMED.