The Equal Employment Opportunity Commission (“EEOC”) appeals the district court’s summary judgment to Texas Instruments, Inc. (“TI”) on the merits of the EEOC’s claims that six manufacturing supervisors were discharged as part of a company reduction in force because of their age. EEOC asserts that a combination of favorable employee performance reviews and a special company RIF policy, three age-related comments and workforce statistics created an inference that TI’s legitimate, nondiserimina-tory reasons for the employment decisions were pretextual for age discrimination. Texas Instruments, Inc. cross-appeals the district court’s orders preventing it from raising a limitations defense. This court AFFIRMS.
BACKGROUND
Sweeping cutbacks in national defense spending and dramatically reduced procurements by the United States Department of Defense forced TI to reorganize its Defense Systems and Electronics Group (“DSEG”) to lay off approximately 850 out of 1700 DSEG manufacturing employees between 1988 and 1994. EEOC studied many of the layoffs but attempted to make a case for illegal discrimination only in regard to six manufacturing supervisors in the DSEG (the “Six Supervisors”), victims of TI’s reduction in force, who were protected by the Age Discrimination in Employment Act of 1967 (“ADEA”), 29 U.S.C. § 623(a). As a “manufacturing supervisor,” each of the men represented by the EEOC is a first-level supervisor who directly oversees hourly workers and reports, in turn, to various levels of management. In December 1993, EEOC filed suit on behalf of Hugh Calhoun (age 54), Walter Morrow (age 59), Allen Powell (age 56), Gaylon Rains (age 51), Jerry Owens (age 50), and Vernon Hillis (age 51).
Voluminous deposition testimony and other evidence were submitted by the parties. The evidence shows that TI’s RIF at DSEG was planned and implemented by Stephen Douth-it (“Douthit”), the head of DSEG’s manufacturing division. ; Douthit testified that he began the RIF by assessing both the current levels of staffing in relation to anticipated product demand and the relative ability of DSEG employees to satisfy and adapt to a rapidly changing technological environment. These assessments led Douthit to conclude that while TI should lay off its hourly employees in order of seniority, manufacturing supervisors should be assessed independently of their seniority. Douthit acknowledged that the decision not to consider seniority in the RIF of manufacturing supervisors was a *1176significant departure from TI’s traditional policy favoring senior employees.
Because abandoning the protections of seniority was such a departure, Douthit presented this suggestion to Jerry Junkins, president and CEO of TI, and Hank Hayes, the president of DSEG. Douthit argued that TI’s policy favoring senior employees would impose significant costs on the company, as supervisory staff was concentrated in high pay grades. Furthermore, unwavering commitment to seniority promised a reorganized workforce that would not have the contemporary skills necessary to assimilate new technologies.1 If seniority preferences were maintained, the RIF would terminate a disproportionate number of employees with college degrees, as many senior employees had acquired their skills primarily through experience. TI management concurred with Douthit and decided to assess manufacturing supervisors independently of their seniority. The goal of this decision, according to management, was to provide TI with the technical expertise and flexibility to serve its current and future customers in an era of dynamic technological progress.
Not only did TI eschew reliance on seniority in its RIF of the manufacturing supervisors, it also did not consider either performance evaluations or the company’s Key Personal Assessments (“KPAs”). The performance evaluations did not aid in deciding which employees to terminate and which to retain because the evaluations were not designed to make fine distinctions among employees or to rate them comparatively; with rare exceptions, the evaluations clustered ratings tightly around the group median. Likewise, TI rejected the KPAs, top-to-bottom rankings of salaried employees in a TI department, because these assessments normally correlated an employee’s rating with his pay schedule; in different terms, highly paid employees invariably received highly rated KPAs.2 The KPA rankings thus did not provide TI with useful information concerning which cross-section of employees it should retain.
Considering neither a supervisor’s seniority nor his performance evaluation or KPA, TI began the difficult process of conducting a RIF among its 45 supervisors. Nine were ultimately terminated and two were demoted; EEOC did not file suit on behalf of three of the employees who were over 40. Regarding the Six Supervisors represented by the EEOC, TI contends that their terminations were not motivated by their ages but, instead, can be explained by legitimate, individualized reasons. A brief summary of these reasons follows:
Powell and Morrow: TI asserts that Powell and Morrow, both manufacturing supervisors at the Precision Automation Center, were discharged because their skills were inferior when compared to Billy Wooley (“Wooley”), a third manufacturing supervisor who was retained during the RIF. Lewis Horn, head of the Precision Automation Center, decided to consolidate the responsibilities of the three supervisors in a single position. Horn testified that he concluded that Powell had only a basic knowledge of computer-based numeric control programming and was reluctant to learn new manufacturing technology. Horn also opined that Morrow lacked the requisite motivation and initiative to perform as the sole supervisor; Horn felt that although Morrow was the supervisor in the highest pay level, his performance was incommensurate with this level. Horn “would expect more out of a job grade 30 than the responsibility he had.... [and would expect him to go] outside and actively solicit[ ] other responsibilities besides just what he had with production control.”
*1177Wooley, by contrast, possessed the technical as well as management skills to function effectively as the consolidated supervisor; Horn observed that Wooley kept excellent contact with clients, formulated customer proposals efficiently, and had extensive knowledge of TI’s computer-based production planning systems. Because Horn considered Wooley the best candidate for the consolidated position, he was retained as the sole supervisor:
Billy Wooley was an outstanding employee who was quite knowledgeable regarding TI’s computer-based production planning systems. Powell and Morrow were not as proficient in these areas. Additionally, Wooley was good at keeping close contact with customers, and talented at formulating customer proposals. Thus, I determined that the work in the ... shop could be best consolidated by retaining Adrian and Wooley and laying off Powell and Morrow.
Calhoun: TI terminated Calhoun when it closed its Trinity Mills facility, where Calhoun had been employed.3 Much of the machinery and the machining work at Trinity Mills was transferred to TI’s Lemmon Avenue facility when Trinity Mills was closed in October, 1990. At Trinity Mills, Calhoun had been in charge of the tool shop and was responsible for the fabrication of certain tools for use by other TI employees. But the Lemmon Avenue facility had a much larger tool shop operation that was already adequately staffed. Because there were no vacant positions in which Calhoun could serve TI and because no “bumping” of other supervisors was permitted under TI’s RIF policy, he was fired.
Rains: TI acknowledges that Rains, the only manufacturing supervisor laid off from Lemmon Avenue, was terminated because of his poor performance. TI demanded that a supervisor such as Rains be proficient in three areas: computer programming, because the product is computer designed; the machining process; and knowledge of the machine tools that perform the work. Robert Anderson, to whom Rains reported, testified that Rains was lacking in all three areas and was seriously hampered in his ability to perform as a supervisor. Anderson reinforced these criticisms in an affidavit that
[Rains] had an inadequate knowledge of machinery, and he had little knowledge of the machine tools he was supervising technicians to repair. These deficiencies are significant, because if a part is not manufactured to specification, as happens occasionally but predictably, a supervisor such as Rains must determine whether the fault lies with the computer program, the machine, or the machine part. Rains’ skills and experience were such that I could not trust his judgment in these regards to the same degree as other manufacturing supervisors.
Because of his poor performance, Rains failed to receive a salary increase in 1990. Rains protested his salary stagnation to Douthit who, after personally investigating the issue, concluded that Rains did perform below par.4 Douthit testified that his
conclusion was and what I told Gaylon that he did, in fact, have a performance problem, and you know, there was sufficient evidence that he had actually been sleeping on the job.... So our agreement was that he was going to go forth and fix, you know, work on the problems that had been laid out to him and, you know, see if he could correct, you know, some of the deficiencies that had been pointed out to him.
Owens: Owens was employed as a supervisor at the Lewisville model shop at the time of his layoff. TI contends that he was terminated because of his lack of technical knowledge in machining and his inadequate communication skills. Mark Cooper, Owens’s former supervisor, testified:
Owens’ primary deficiency was that he had minimal technical and machining knowledge. Further, Owens was lacking basic *1178‘people skills’ and was found to be abrasive in his dealings with other personnel.... Because of Owens’ deficiencies, he was not as critical as other manufacturing supervisors in a time of business decline and was, therefore, laid off.
Hillis: TI stated that it terminated Hillis because he lacked high tech skills and interacted poorly with other employees. Hillis lost his job when TI closed the model shop in TI’s North Building and consolidated at the Lewisville and McKinney sites. Peter Loughlin, a manager in DSEG, testified that he considered Hillis to be
the weakest [supervisor] in the North Building model shop. He was employed in a rapidly changing ‘high tech’ environment. Hillis showed little motivation to learn the software packages that were available on the computer_ I considered him someone who shunned, rather than sought out, such challenges. Hillis’ chief shortcoming was the manner in which he interacted with the other shift supervisors.... Hillis typically did not arrive in advance of his shift to learn from the second shift supervisor how operations were proceeding, or if any problems were ‘hot’ and required special attention.
Hillis admits that he is not technologically sophisticated and that he regarded himself primarily as a machinist; he testified at his deposition that “I’m not all that educated. I’ve been around machining, and that doesn’t interface with computers and so forth.” Hil-lis also described his computer capabilities as “[z]ero.” Although TI concedes that Hillis was a satisfactory employee, it concluded that he was expendable in the RIF.5
In response, the EEOC argues that the particularized rationales suggested by TI are merely pretexts for age discrimination. EEOC refers to “statistical, anecdotal, and other circumstantial evidence to permit a reasonable factfinder to conclude that age was a factor in the discharge decisions.... ”
The statistical evidence on which the EEOC relies was compiled by Dr. John G. Claudy (“Claudy”). Claudy analyzed data provided by TI to determine whether there was a statistically significant relationship between age and the likelihood that a supervisor would be terminated. In part, Claudy calculated that while TI had discharged 38.9% of the manufacturing supervisors age 50 and older, the company had terminated only 7.4% of the supervisors under age 50. After conducting three alternative statistical tests Claudy concluded that there was a statistically significant relationship between age and discharge to the disadvantage of supervisors over fifty. Claudy admitted that his analysis did not consider the specific talents or duties of the manufacturing supervisors.
In addition to the statistical suggestion of age discrimination, the EEOC urges that TI’s departure from its traditional seniority protections as well as its decision to ignore performance evaluations and the KPAs facilitated age discrimination. TI’s RIF policy directed that the first employees to be discharged would be those on disciplinary probation, followed by those who had unsatisfactory evaluations, then volunteers, and finally, those with lower job criticality rankings. None of the Six Supervisors had been on disciplinary probation; none had received an unsatisfactory rating on his previous performance evaluation; and TI did not develop a “criticality ranking” of the supervisors before implementing the RIF, though its RIF policy required such a ranking. The EEOC asserts that TI’s departure from its established RIF procedures and its decision to ignore previous formal evaluations of the supervisors constitute evidence that the company intentionally discriminated against the Six Supervisors because of their ages.
The EEOC further contends that three age-based comments by TI managers articulate the company’s discriminatory intent. Fred Blair, a TI personnel director but not a decisionmaker, explained to Rains when Rains was informed of his termination that “it’s just that you’ve reached that age and years of service that we can bridge you to retirement.” The remaining two comments were allegedly directed to Casimir Tencza *1179(“Tencza”), another supervisor discharged in the RIF. Tencza was allegedly told that he was laid off in part because TI “had to make room for some of the younger supervisors” and that “his age got him.” Importantly, however, Tencza is not represented by the EEOC and has settled his claim against the company. From these statements, the EEOC seeks to imply discrimination by TI directed against its older employees.
Finally, the EEOC questions the particularized explanations suggested by TI for firing the Six Supervisors. A brief summary of the EEOC’s criticisms contrasts the agency’s position with that of TI:
Powell and Morrow: The EEOC stresses that both Morrow and Powell had outstanding performance evaluations. Further, Morrow enjoyed extensive technical skills, while Powell was evaluated as consistently generating a profit for TI while maintaining customer satisfaction.
Calhoun: Calhoun received superior ratings in his performance evaluations in the areas of “job knowledge” and safety. Also, Calhoun was doubtlessly technically competent and, while at Trinity Mills, had supervised some of the machinery that was later transferred to Lemmon Avenue when the Trinity Mills facility closed.
Rains: Although the EEOC concedes that Rains was the sole discharged supervisor who received a poor performance evaluation, it contends that during his exit interview, Rains was assured by Dial, one of his supervisors, that he had been a “good performer.” Further, as noted earlier, Rains was allegedly told by Blair that he had “reached that age and years of service that we can bridge you to retirement.”
Owens: Owens received a favorable performance evaluation that praised his communication skills and described him as a cooperative and respected employee. In these evaluations, Owens never fell below the “benchmark level of performance” necessary to satisfy his job requirements.
Hillis: Contrary to TI’s suggestion that Hillis lacked technical expertise, the EEOC emphasizes that his evaluations indicated that he had a thorough knowledge of his technical duties. For instance, in his most recent performance review, Hillis received superior ratings in evaluations of his job quality, knowledge, and safety.
After their terminations, four of the Six Supervisors — Morrow, Powell, Rains and Calhoun — filed their charges of discrimination with the EEOC on June 21, 1991; neither Owens nor Hillis filed charges with the agency. As previously noted, the EEOC filed suit on behalf of the Six Supervisors on December 3,1993.
. TI answered the lawsuit and later moved for leave to file its first amended answer, seeking to add as an affirmative defense that the EEOC’s claims were untimely under the relevant statute of limitations. TI soon moved for summary judgment on that ground. In January, 1995, TI filed an additional motion for summary judgment on the merits of the EEOC’s claims. The district court granted summary judgment on the merits but denied TI’s motions for leave to amend and for summary judgment on limitations. Each side has appealed the rulings unfavorable to it.
DISCUSSION
A. Standard of Review
This court reviews de novo the district court’s grant of summary judgment, employing the same criteria used in that court. Burfield v. Brown, Moore & Flint, Inc., 51 F.3d 583, 588 (5th Cir.1995). Summary judgment is proper only “if the pleadings, depositions, answers to interrogatories, and admissions on file, together with the affidavits, if any, show that there is no genuine issue as to any material fact and that the moving party is entitled to judgment as a matter of law.” Fed.R.Civ.P. 56(c). Factual questions and inferences are viewed in the light most favorable to the nonmovant. Lemelle v. Universal Mfg. Corp., 18 F.3d 1268, 1272 (5th Cir.1994).
Although Rule 56(e) requires the moving party to demonstrate the absence of a genuine issue of material fact, a dispute about a material fact is genuine only if the evidence is such that a reasonable jury could return a verdict for the nonmovant. See Celotex *1180Corp. v. Catrett, 477 U.S. 317, 323, 106 S.Ct. 2548, 2552, 91 L.Ed.2d 265 (1986); Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248, 106 S.Ct. 2505, 2510, 91 L.Ed.2d 202 (1986). If the moving party demonstrates the absence of a genuine issue of material fact, then the nonmovant is burdened with establishing the existence of a genuine issue for trial. Matsushita Elec. Indus. Co. v. Zenith Radio, 475 U.S. 574, 585-87, 106 S.Ct. 1348, 1355-56, 89 L.Ed.2d 538 (1986). This burden requires the nonmovant to do more than merely raise some metaphysical doubt as to the material facts. Matsushita, 475 U.S. at 586, 106 S.Ct. at 1355.
B. Demonstrating Pretext
TI acknowledges and the district court assumed that the EEOC established a prima facie case of discrimination against the Six Supervisors simply by showing that they were over 40, and were discharged, while younger supervisors remained on the payroll. Meinecke v. H & R Block of Houston, 66 F.3d 77, 83 (5th Cir.1995). TI utilized its opportunity through affidavits and depositions to explain the legitimate, nondiscriminatory reasons for each adverse employment action. The burden then fell on EEOC to raise a genuine, material fact issue that those reasons were not the real reasons and were pretexts for age discrimination. Id. As the district court concluded that EEOC’s evidence did not discharge its burden, summary judgment was granted. The court held:
[t]he EEOC has failed to demonstrate a genuine issue for trial about the legitimacy of the nondiscriminatory reasons given by TI for the discharge of the [S]ix [Supervisors. The three theories set forth by the EEOC which, it argues, show a pretext for age discrimination are either not supported by the evidence or are not sound legally. The court concludes that no genuine issue of material fact exists on the question of whether TI intended to discriminate against the supervisors.6
Sitting en banc, this court recently discussed the burden confronting an ADEA plaintiff who seeks to demonstrate that an employer’s proffered rationales for adverse employment acts were merely pretextual and that discrimination based on age was an actual reason for these adverse acts. In Rhodes v. Guiberson Oil Tools, 75 F.3d 989 (5th Cir.1996) (en banc), this court explained that the plaintiff in an ADEA disparate treatment case must offer evidence to rebut each of the employer’s articulated legitimate, nondiscriminatory reasons. The court reasoned that
[I]n tandem with a prima facie case, the evidence allowing rejection of the employer’s proffered reasons will often, perhaps usually, permit a finding of discrimination without additional evidence. Thus, a jury issue will be presented and a plaintiff can avoid summary judgment and judgment as a matter of law if the evidence taken as a whole (1) creates a fact issue as to whether each of the employer’s stated reasons was what actually motivated the employer and (2) creates a reasonable inference that age was a determinative factor in the actions of which the plaintiff complains. The employer, of course, will be entitled to summary judgment if the evidence taken as a whole would not allow a jury to infer that the actual reason for the discharge was discriminatory.
Rhodes, 75 F.3d at 994 (emphasis added). Whatever evidence a plaintiff tenders must rebut each of the employer’s proffered rationales:
[I]n some cases, for instance, the fact that one of the nondiscriminatory reasons in the record has proved to be highly questionable may not be sufficient to cast doubt on the remaining reasons. Likewise, an employer’s explanation for its proffer of a pretextual reason may preclude a finding of discrimination.
Id. at 994 (citing Woods v. Friction Materials, 30 F.3d 255, 261 n. 3 (1st Cir.1994)) (concluding that a jury could not infer age *1181discrimination if the proffered reason was in fact a pretext, though not for discrimination).
There is another wrinkle on the standards for evaluating discrimination. In the context of a reduction in force, which is itself a legitimate, nondiseriminatory reason for discharge, the fact that an employee is qualified for his job is less relevant — some employees may have to be let go despite competent performance. Walther v. Lone Star Gas Co., 952 F.2d at 124. If, however, the older employee shows that he was terminated in favor of younger, clearly less qualified individuals, a genuine, material fact issue exists. Walther, 952 F.2d at 12B.
C. The EEOC’s Evidence of Pretext
(1) TI’s Age-Based Comments
EEOC first relies on three age-related comments allegedly made by TI employees as proof of TI’s discriminatory motivation.7 Plaintiff Rains testified that Blair, a TI personnel director, explained the decision to terminate Rains by suggesting that “it’s just that you’ve reached that age and years of service that we can bridge you to retirement.” The two remaining statements were allegedly made by Horn to Tencza, a supervisor discharged in the RIF who is not represented by the EEOC. Horn allegedly said of Tencza, “his age hot him,” and that TI had to make room for younger supervisors.
This court has repeatedly held that “stray remarks” do not demonstrate age discrimination. See, e.g., Waggoner v. City of Garland, 987 F.2d 1160, 1166 (5th Cir.1993) (a statement by a decisionmaker that an employee was an “old_” and that a younger person could complete his work faster was a stray remark insufficient to establish age discrimination); Guthrie v. Tifco Industries, 941 F.2d 374, 378-79 (5th Cir.1991) (holding that such “statements are too vague to be accepted as direct evidence of age-bias.”); Turner v. North American Rubber, Inc., 979 F.2d 55, 59 (5th Cir.1992). (vague and remote remarks cannot establish age discrimination). In order for an age-based comment to be probative of an employer’s discriminatory intent, it must be direct and unambiguous, allowing a reasonable jury to conclude without any inferences or presumptions that age was an impermissible factor in the decision to terminate the employee. Bodenheimer v. PPG Industries Inc., 5 F.3d 955, 958 (5th Cir.1993).
The statement allegedly made by Blair to Rains is a stray remark that does not demonstrate age bias. Blair’s statement simply recognized a fact concerning Rains’s seniority, an observation which did not imply seniority was the reason for discharge. This interpretation of the statement is consistent with the context in which it was allegedly made, as Rains’s duty was to explain termination benefit packages to employees. See, e.g., Guthrie, 941 F.2d at 378-79 (statement that founder of company suggested to his son that he needed “to surround himself with people his age” is not direct evidence of age discrimination); Turner, 979 F.2d at 59 (comment that an employee needed “three young tigers” to assist with operations is not probative of discrimination). TI also notes that Blair was not a decisionmaker at TI with respect to the RIF and that he had no input into the decision to terminate any of the Six Supervisors. Blair’s stray remark is thus not probative of whether TI’s decision to terminate Rains was motivated by age discrimination. See, e.g., Rhodes, 75 F.3d at 994; Cone v. Longmont & United Hosp., 14 F.3d 526, 529 (10th Cir.1994) (age-related comment by non-decisionmakers are not material to showing employer’s age discrimination.)
Further, the statements allegedly made concerning Tencza’s termination are not probative in themselves or relevant to TI’s decision to terminate the Six Supervisors. Horn, who allegedly make the remarks, recommended replacing Tencza, age 55, with Garner, age 51, an action which speaks louder about Horn’s motivation than the words attributed to him. Moreover, the *1182statement about TI’s need to make room for younger supervisors reflects the kind of truism this court and others have .held does not evidence discrimination. See, e.g., Birkbeck v. Marvel Lighting Corp., 30 F.3d 507, 512 (4th Cir.1994), cert. denied — U.S. -, 115 S.Ct. 666, 130 L.Ed.2d 600 (1994) (statement that “there comes a time when we have to make room for younger people” creates no inference of age discrimination). But even if Horn’s alleged remarks were probative of discrimination against Tencza, they cannot carry the heavier burden of condemning the motivation of TI toward the Six Supervisors: Horn was a first-tier evaluator of layoffs whose recommendations had to be approved by Dial and Douthit, and Horn was partly responsible for recommendations on only two of the Six Supervisors. In short, the alleged statements about Tencza’s termination are no more probative of disparate treatment of the Six Supervisors than whatever was said to the numerous supervisors over 40 years old whom TI did not discharge.8
(2) TPs Conscious Departure from its oum Procedures
The EEOC urges that TI’s conscious decision to ignore its usual policies and procedures when conducting the RIF demonstrates that its expressed reasons for terminating the six supervisors are post hoc pretexts for age discrimination. The district court rejected this inference, concluding that the EEOC had not shown a discriminatory connection between TI’s failure to follow its policies and its selection of supervisors subject to the RIF.
This court has observed that an employer’s “disregard of its own hiring system does not of itself conclusively establish that improper discrimination occurred or that a nondiscriminatory explanation for an action is pretextual.” Risher v. Aldridge, 889 F.2d 592, 597 (5th Cir.1989); see also, Moore v. Eli Lilly & Co., 990 F.2d 812, 819 (5th Cir.), cert. denied, 510 U.S. 976, 114 S.Ct. 467, 126 L.Ed.2d 419 (1993). This court elaborated in Moore that
[pjroof that an employer did not follow correct or standard procedures in the termination or demotion of an employee may well serve as the basis for a wrongful discharge action under state law. As we have stated, however, the ADEA was not created to redress wrongful discharge simply because the terminated worker was over the age of forty. A discharge may well be unfair or even unlawful and yet not be evidence of age bias under the ADEA. To make out an ADEA claim, the plaintiff must establish some nexus between the employment actions taken by the employer and the employee’s age. [A] bald assertion that one exists ... simply will not suffice.
Id. 990 F.2d at 819 (citing Bienkowski, 851 F.2d at 1508 n. 6).
The EEOC has not demonstrated such a nexus; rather, as in Moore, its argument rests on a bald assertion that one exists. What happened in this case, without dispute, is that TI created a new layoff policy regarding supervisors that disregarded seniority; the company did not simply fail to follow the usual seniority-protective policy. As previously summarized, TI carefully outlined why it disregarded the seniority protections typically afforded to its employees in a RIF and why it did not consider performance evaluations or KPAs when determining which supervisors to retain. The EEOC failed to undermine this decision. Specifically, Douth-it testified at his deposition that under preexisting TI policy he was “not permitted to lay off individuals with more than 15 years of seniority without gaining the approval of the president of the company.” Douthit lamented that the economic costs of the seniority policy were extreme since “the mix of [TI’s] job grades was becoming more and more concentrated towards the higher end, which was driving our average pay up, which was *1183driving our cost up, which was making us less competitive in the marketplace.” Additionally, TI’s seniority rule would force TI to displace a disproportionate number of employees with college degrees since “the majority of [TI’s] degree people had less than 15 years of service while ... a high percentage of people [had] greater than 15 years of service [but no] degrees.” Instead -of this expensive, inflexible protection of seniority, Douthit favored a RIF policy which would make “those decisions on a merit basis ... as opposed to, you know, a fixed set of rules that did not allow [TI] the flexibility to make decisions that were dictated by business conditions.”
Douthit advocated a merit-based retention policy to the CEO of TI and the president of the DSEG organization. Douthit documented the present concentration of supervisors in higher pay grades, the skills that TI most needed to retain in its RIF, and- the senior employees who would be discharged if seniority protections were waived. The company agreed to modify its traditional RIF policy in part “[b]ecause technology had changed. [A] lot of high seniority people had skill sets that were ... grounded in the 1960s and 1970s technology. Products of the future were going to require 1980s and 1990s technology. [Some senior employees] had not or could not ... make the transition to those new technologies and required skills.”
Before TI undertook the difficult task of selecting employees to terminate during its RIF, the company recognized that its traditional protection of seniority would hamper future performance. As a result, after identifying the skills most vital to the company’s ability to adapt to a rapidly changing technological environment, TI waived its seniority protections and laid1 off supervisors who might have otherwise been retained because of their seniority. TI’s decision to replace a seniority policy that would impede its ability to reduce its workforce while maximizing the efficiency and expertise of the remaining employees does not, without a clear nexus to discrimination, create an inference of age discrimination.
(3) Criticisms of TI’s Particularized Reasons
This court has already summarized the criticisms advanced by the EEOC of TI’s reasons for discharging each of the Six Supervisors. Rather than demonstrate the likely falsehood of those reasons, EEOC responds that each of the supervisors had competent performance evaluations and KPAs and was considered an effective employee by TI. Critically, the EEOC’s criticisms rest on information found in TI’s annual performance evaluations and its KPAs, measures which are misleading guides to the RIF decisions TI had to make.
TI rejected use of the performance evaluations and KPAs after determining that neither report provided worthwhile information about the comparative worth of a particular employee. Stephen Leven (“Leven”), Vice President of Human Resources at TI, attested that
[t]he performance evaluations completed annually for TI’s employees have tended to be quite uniform. It was intended by TI that the vast majority of employees would receive similar performance ratings, with only exceptionally good or poor performers deviating from this norm. The purpose of this common rating is to avoid making fine distinctions among employees who are performing adequately when no pending business decisions turn directly on such ratings.
For instance, of the 135 evaluations performed on a group of 45 manufacturing supervisors during the three years preceding the RIF, only two evaluations gave a summary rating below the group median.9 *1184Another drawback of the performance evaluations is that they provide no useful information regarding the ability of an employee to adapt to changing technology. Leven explained that
the written performance evaluations typically rate an individual with respect to the job he or she currently performs, but are silent with respect to the value to TI of the job they are performing, or the usefulness of their skills as applied to future operations and technology.
Given the tendency of these evaluations to cluster employee performance ratings and their lack of focus on the company’s future technological needs, TI had a legitimate business basis not to rely on the evaluations in conducting its RIF.
The KPAs were also deficient for TI’s RIF decisions. Leven explained that
[t]he KPA is performed annually and provides a “top to bottom” ranking of exempt employees in each organization, including manufacturing operations, according to their perceived value. The KPAs, however, are closely tied to salary determinations and reflect considerations in addition to performance.
Indeed, the KPA rankings were often simply reiterations of the pay gradations at TI; the greater the pay, the higher the KPA. Also, because they are merely top-to-bottom rankings, the KPAs do not provide information about the proper mix of employees and specialties that should be retained in a RIF. As Douthit explained in his deposition, it is “entirely conceivable that you could have somebody that would be high on the KPA, you know, good performer, but may reside in a skill category that was no longer or was not critical to the success of the business.”
In sum, the EEOC’s proof fails to undermine TI’s legitimate, non-discriminatory, and individualized reasons for terminating the Six Supervisors for several reasons. First, the agency makes general assertions that the supervisors were competent employees, but it does not directly challenge TI’s individualized explanations for not retaining the Six Supervisors. The agency showed only that the supervisors were qualified, not that they were clearly better qualified than the supervisors retained in the RIF. See Walther, supra;10 see also EEOC v. Maryville Sales Corp., 27 F.3d 1089, 1096 n. 5 (5th Cir.1994), cert. denied, — U.S. -, 115 S.Ct. 1252, 131 L.Ed.2d 133 (1995). Second, the EEOC’s showing is premised on the performance evaluations and KPAs, data irrelevant to the RIF decisions. Thus, while some of TI’s criticisms of individual supervisors’ performances appear to be contradicted by the ratings and comments on their annual evaluations, the record demonstrates that the annual evaluations were not designed to compare or differentiate among employees, the critical task for management conducting a RIF. Third, the EEOC has not offered proof to indict TI’s logic for abandoning either the KPAs of the performance evaluations.
(4) Statistical Evidence
The EEOC finally contends that its statistical tests indicating that age was a significant predictor of the likelihood of layoff permit a genuine inference of pretext. The district court summarily rejected this contention, reasoning that statistics will only rarely rebut an- employer’s particularized, legitimate, and nondiscriminatory reasons for the adverse employment decision.
The district court relied extensively on this court’s opinion in Walther v. Lone Star Gas Co., 977 F.2d 161, 162 (5th Cir.1992). In Walther, this court explained that statistical evidence usually cannot rebut the employer’s articulated nondiscriminatory reasons. The court observed that
*1185gross statistical disparities resulting from a reduction in force or similar evidence may be probative of discriminatory intent, motive or purpose. Such statistics might in an unusual case provide adequate circumstantial evidence that an individual employee was discharged as part of a larger pattern of layoffs targeting older employees. This is not to say that such statistics are enough to rebut a valid, nondiscriminatory reason for discharging a ‘particular employee. Generally, they are not_ [PJroof of pretext, hence of discriminatory intent, by statistics alone would be a challenging endeavor.
Walther, 977 F.2d at 162 (emphasis added) (citations omitted). Other circuits have expressed similar skepticism about the ability of statistics to rebut the employer’s nondiscriminatory reasons. See, e.g., LeBlanc v. Great American Ins. Co., 6 F.3d 836, 848 (1st Cir.1993), cert. denied, — U.S. -, 114 S.Ct. 1398, 128 L.Ed.2d 72 (1994) (noting that “a company’s overall employment statistics will have little direct bearing on the specific intentions of the employer when dismissing a particular individual.”); Barnes v. GenCorp., Inc., 896 F.2d 1467, 1469 (6th Cir.), cert. denied, 498 U.S. 878, 111 S.Ct. 211, 112 L.Ed.2d 171 (1990) (when the defendant offers particularized reasons, this “cannot be rebutted by reference to the statistics already presented since the statistics here do not tend to establish that age played a factor in any particular decision.”) (citations omitted) (emphasis in original).
While Walther explains that generalized statistical evidence will rarely rebut a particularized nondiscriminatory rationale, statistical evidence may be probative of pretext in limited circumstances, however. See Deloach v. Delchamps, Inc., 897 F.2d 815, 820 (5th Cir.1990) (“Delchamps further claims that while statistical data may be used to establish a prima facie case, it cannot be used to establish pretext. While that may be true when that is the only evidence a plaintiff has to establish pretext ... we do not believe the same rule applies when a plaintiff offers additional evidence.”). The probative value of statistical evidence ultimately depends on all the surrounding facts, circumstances, and other evidence of discrimination. International Bd. of Teamsters v. United States, 431 U.S. 324, 340, 97 S.Ct. 1843, 1856-57, 52 L.Ed.2d 396 (1977).
In the instant case, the statistical evidence offered by the EEOC does not support an inference that TI’s reasons for terminating the Six Supervisors were merely pretextual. Importantly, TI’s expert, Dr. Daniel S. Hamermesh (“Hamermesh”), showed that the EEOC’s results are only statistically significant for a subgroup of supervisors over 50, although the protected class includes all employees over 40 years old. Hamermesh found that the layoffs were not statistically significant for ages between 40 and 45 and those above 52; the layoffs had a statistically significant impact only for ages 46 through 51. Of the Six Supervisors, only Rains, Owens, and Hillis fall in the subgroup for which a statistically significant result was deduced. EEOC did not sue on behalf of the 47 or 49-year old supervisors who were laid off.
As this court has recognized, “particularly in age discrimination cases where innumerable groupings of employees are possible according to ages and divisions within the corporate structure, statistics are easily manipulated and may be deceptive.” Walther v. Lone Star Gas Co., 952 F.2d 119, 124 (5th Cir.), on rehearing, 977 F.2d 161 (5th Cir.1992). EEOC’s expert had no explanation for his selection of the arbitrary age 50 cutoff other than that was the group EEOC asked him to consider. EEOC’s choice of age groups for statistical review is not probative of age discrimination.
Further, the EEOC’s own expert admitted during his deposition that his statistical analysis did not consider the specific talents or duties of the manufacturing supervisors at TI. Because the EEOC’s statistics do not even purport to analyze the facts concerning individual supervisors, the statistics are impotent, without more, to rebut TI’s particularized reasons for the termination of the Six Supervisors. The statistics cannot satisfy the requirement in Rhodes that the EEOC demonstrate a fact issue as to whether each of the TI’s stated reasons actually motivated its RIF. Rhodes, 75 F.3d at 994.
*1186Moreover, the authorities on which the EEOC relies are inapposite. Several cases considered the proffered statistical evidence probative of pretext only because the plaintiff had offered particularized evidence directly challenging the defendant’s announced rationale. See, e.g., Deloach, 897 F.2d at 818-20 (only when coupled with other evidence contradicting employer’s reasons was statistical evidence probative of pretext); Freeman v. Package Machinery Co., 865 F.2d 1331, 1342-42 (1st Cir.1988) (statistical evidence is probative when it demonstrated a pattern consistent only with discrimination and it was combined with independent and conflicting testimony of job performance); Krodel v. Young, 748 F.2d 701, 710 (D.C.Cir.1984), cert. denied, 474 U.S. 817, 106 S.Ct. 62, 88 L.Ed.2d 51 (1985) (without more, statistical evidence is “less significant” and “ordinarily not dispositive” in disparate treatment cases); Reeves v. General Foods Corp., 682 F.2d 515, 523-25 (5th Cir.1982) (Reeves presented evidence that “his performance was such that the jury could have decided that General Foods’ admittedly articulated explanation of poor performance was unworthy of credence.”). Likewise, in Bienkowski v. American Airlines, Inc., 851 F.2d 1503 (5th Cir.1988), this court recognized that statistical evidence was only probative of intent when the plaintiff had disputed with other evidence the defendant’s performance evaluation; in Bienkowski, the plaintiff had produced various affidavits from which “a jury could conclude that his supervisors’ evaluation of his performance lacked veracity and that their true motivation resided in their age-based comments.” Id. at 1508. In other cases cited by the EEOC, the court merely explained in dicta that statistical evidencé can be evidence of pretext. See, e.g., EEOC v. Manville Sales Corp., 27 F.3d 1089, 1096 n. 5 (5th Cir.1994); Gusman v. Unisys Corp., 986 F.2d 1146, 1147-48 (7th Cir.1993) (suggesting that “[cjlaims of discrimination are hard to prove one case at a time,” but emphasizing that “[ajmple evidence warranted a conclusion [by the jury] that Gusman’s immediate supervisor ... believed that older workers are inferior.”). Finally, contrary to the EEOC’s suggestion, the court in Mac-Dissi v. Valmont Indus., Inc., 856 F.2d 1054 (8th Cir.1988), relied on statistical evidence merely to conclude that the plaintiff had established a prima facie case of age discrimination; in fact, the court distinguished Mac-Dissi from other eases where the plaintiffs attempted to use statistics to prove pretext. Id. at 1058. None of these cases suggest that statistics which rely on the arbitrarily selected cutoff used here can be probative of pretext in these disparate treatment cases.
(5) Cumulative Impact of EEOC’s Evidence
EEOC strenuously criticizes the district court for having discounted each of its types of evidence separately and ignored that, taken together, all of the agency’s evidence bespoke pretext sufficiently to warrant a jury trial. The district court was obliged, however, to consider the admissibility of evidence offered to support the parties’ summary judgment positions .pursuant to Fed. Rule Civ.P. 56(e). His discharge of that duty led to a separate look at EEOC’s categories of evidence: alleged “ageist” comments; TI’s departure from prior procedures and employee evaluation standards; and statistical data. Like the district court, we have undertaken de novo review of the record, including EEOC’s attempted refutation of the individualized reasons for terminating each supervisor and the other evidence offered by the agency. We agree with the district court that portions of the agency’s proffered evidence, e.g. the statistics and stray remarks, were not probative of age discrimination. The agency failed to join issue on TI’s business reasons for using a non-seniority-proteetive method of selecting which supervisors to lay off. The agency’s attempted refutation of TI’s individualized, nondiscriminatory reasons for discharging the Six Supervisors misses the mark because it does not undermine the comparative rating of the Six Supervisors among all the supervisors exposed to the RIF. EEOC also fails to cast doubt on TI’s explanation that the satisfactory performance of a supervisor during a period of economic stability does not necessarily establish the employee’s essentiality to the company and its future in an era of downsizing. “Evidence” that does not imply pretext taken *1187alone does not do so when cumulated. See Holmberg v. Baxter Healthcare Corp., 901 F.2d 1387, 1391 (7th Cir.1990) (“the sum of four nondiseriminatory episodes does not support [a] case anymore than viewing the four episodes separately”). The agency’s case, in sum, confused a quarrel with the merits of the company’s business decision — a quarrel in which the ADEA plays no role— with a case of illegal age discrimination.
STATUTE OF LIMITATIONS
In July 1994, TI sought leave to amend its answer and expand its plea of limitations as an affirmative defense. The original answer asserted only that non-willful ADEA violations were time-barred. Later that month but before the district court had ruled on the motion to amend, TI moved for summary judgment on the grounds that the EEOC’s claims were untimely. On the same day that the district court granted TI summary judgment on the merits of the EEOC’s claims, the district court denied TI’s motion for leave to amend as moot and denied summary judgment on limitations.
TI has attempted to cross-appeal this denial of summary judgment and the court’s refusal to rule on its motion to amend. Even if we have appellate jurisdiction over these interlocutory orders, it is unnecessary to discuss the timeliness of EEOC’s complaint because, following the district court’s lead, we have alternatively ruled on the merits.
CONCLUSION
For the foregoing reasons, the district court’s award of summary judgment to TI on the merits of the claims brought by the EEOC is AFFIRMED.
. TI argues that older employees at DSEG tended not to have state-of-the-art technological training, observing that the older "supervisors who had learned their skill on-the-job through their former positions as machinists, found themselves with dated skills unless they kept abreast of the new computer-based production methods.” Joe Dial, site manager at a TI manufacturing facility, concurred and concluded in a two-page assessment to Douthit that TI’s future operations demanded supervisors with enhanced capabilities in statistics and experimental design.
. While this result is hardly counter-intuitive, TI suggests that many of the employees who received high marks on their KPAs actually per-N formed deficiently.
. During his deposition, Joe Dial explained that TI intended to close down or significantly downsize operations in at least three TI facilities: Trinity Mills, Colorado Springs, and the North Building.
. Rains in fact received the lowest performance rating of any of the 45 manufacturing supervisors on the performance evaluation conducted prior to the layoff.
. Loughlin concluded that "[w]hile Hillis’ performance would have permitted him to remain employed had business remained at appropriate levels, the downturn in business required me to designate the weakest performer in my group, which was Hillis."
. The "three theories" to which the district court referred involved the efforts of the EEOC to demonstrate pretext through the use of (1) age-related comments; (2) TI's conscious failure to follow its own RIF procedures; and (3) statistical evidence of age discrimination. EEOC also contested the veracity of TI's reasons for each individual termination.
. The district court found and analyzed only two statements, one allegedly made by Blair to Rains, and the other directed at Tencza. On appeal, the EEOC points out that two separate age-based comments were made to Tencza. Although the making of these comments is disputed, we must assume that they were made for purposes of summary judgment.
. Moreover, the cases on which the EEOC relies are distinguishable because the remarks in those cases demonstrated discriminatory intent by the final decisionmaker in both the plaintiff's and the third parly’s cases. See Shattuck v. Kinetic Concepts, Inc., 49 F.3d 1106, 1109 (5th Cir.1995). Courts that allow statements about non-parties like Tencza to be used do so because the discriminatory animus that motivated those comments provides circumstantial evidence of discriminatory intent in the particular case; any possible inference of discrimination in this case is far weaker and more remote from the decisionmak-ing process.
. TI’s management did not attempt to make fine distinctions between employees in the performance evaluations. Whit Smith, an expert witness for TI, opined that
[a]n analysis of the 1988 and 1989 performance reviews for the forty-five manufacturing supervisors reveals that there is not enough variance in the ratings of these individuals to enable management to utilize these as a tool in making a decision regarding the retention or layoff of any individual supervisor. With only one exception, each manufacturing supervisor was rated by as his/her manager as fully meeting or, in some cases exceeding, job require*1184ments and performance standards. This rating was consistently given to each manufacturing supervisor, even in those few cases in which the need for improvement ... was identified through the manager's comments. Managers appear to be reluctant to rate any supervisor as not meeting expectations....
(emphasis added).
. "... a plaintiff can take his case to a juiy with evidence that he was clearly better qualified than younger employees who were retained." Walther, 952 F.2d at 123, citing Thornbrough v. Columbus & Greenville R. Co., 760 F.2d 633, 647 (5th Cir.1985).