Appeal from the United States District Court for the District of Colorado. D.C. No. 97-WY-240-AJ[Copyrighted Material Omitted][Copyrighted Material Omitted][Copyrighted Material Omitted][Copyrighted Material Omitted][Copyrighted Material Omitted][Copyrighted Material Omitted]

Barry D. Roseman of Roseman & Kazmierski, LLC, Denver, Colorado, for Plaintiff-Appellant.

Jerry N. Jones (Paul F. Lewis with him on the brief) of Moye, Giles, O'Keefe, Vermeire & Gorrell LLP, Denver, Colorado, for Defendant-Appellee.

Elizabeth Lamb Kearney and Brent Ruther of Law Offices of Elizabeth Lamb Kearney, Denver, Colorado; and Paula Brantner of National Employment Lawyers Association, San Francisco, California, filed a brief for amicus curiae National Employment Lawyers Association.

John J. Gallagher, Kenneth M. Willner and Neal D. Mollen of Paul, Hastings, Janofsky & Walker LLP, Washington, D.C.; Brian M. Mumaugh and Marcy G. Glenn of Holland & Hart LLP, Denver, Colorado; and David Berg of Air Transport Association of America, Washington, D.C., filed a brief for amicus curiae Air Transport Association of America.

Before TACHA, BARRETT and BRORBY, Circuit Judges.

BRORBY, Circuit Judge.

United Airlines, Inc. ("United") interviewed and rejected Ms. Bullington for the position of line pilot/flight officer on three separate occasions. Ms. Bullington brought this action pursuant to Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., and the Age Discrimination in Employment Act ("ADEA"), 29 U.S.C. 621 - 634, claiming United refused to hire her because of her gender, her age and in retaliation for complaining about alleged discrimination during the interview process. Ms. Bullington further claims United breached an implied contract or an otherwise enforceable promise by refusing to hire her. The district court granted United's motion for partial dismissal and United's subsequent motion for summary judgment, and Ms. Bullington appeals. We have jurisdiction pursuant to 28 U.S.C. 1291, and affirm in part and reverse in part.

I. Background

Ms. Bullington, a female over the age of forty, currently works for United as a ground school academic instructor. Over a two-year period, Ms. Bullington sought but was denied a position as line pilot with United on three occasions. United's application and selection process for flight officers involves three phases. In the initial phase, United accepts applications from individuals meeting certain minimum qualifications including 350 hours of flight experience, commercial pilot certification, a high school diploma, and other physical and medical requirements. United then ranks eligible applicants according to aeronautical experience. Those applicants ranked at the top of the list advance to the second phase of the selection process. Because female applicants typically have less aeronautical experience than male applicants, United ranks male and female applicants separately. United then selects a proportionate number of males and females to proceed to the second phase. At the second phase, applicants must complete a simulator flight and a formal interview. Based on the applicant's performance, a review board then decides whether to reject the applicant or to extend a conditional offer. If United extends a conditional offer, the candidate moves on to the third phase, which includes a medical exam and background check. Ms. Bullington objects to the formal interview portion of the selection process.

Two United employees conduct the formal interview, an employment representative and a flight operations representative. These individuals assess the applicant in seven broad categories or "dimensions" including: industry motivation, decision making/problem solving, compliance and conformity, leadership, interpersonal skills, technical evaluation, and appearance/presentation. Each dimension is broken down into a set of attributes or "anchors" United deems desirable in a flight officer. Interviewers ask applicants questions from a suggested list and, based on the applicant's response, evaluate whether the applicant meets United's set standards for each attribute.1 Based on those attribute evaluations, the interviewers give the applicant a numerical score for each dimension, ranging from a low of "1" to a high of "5." The dimension scores are then averaged to arrive at the applicant's overall score. An applicant must have an overall score of "3" or better to be recommended for a flight officer position. However, if an applicant scores a "2" or lower on any one dimension, her overall score will also be a "2," and the interviewers will not recommend her for a flight officer position.

United interviewed Ms. Bullington for a flight officer position three times January 1993, March 1995 and May 1995. Each time, Ms. Bullington received an overall score of "2," thereby disqualifying her from further consideration. After her first unsuccessful interview in January 1993, Ms. Bullington spoke with Ms. Nancy Stuke, United's Manager of Flight Officer Employment, and expressed her concerns that one of her interviewers was biased against her. Ms. Bullington claims Ms. Stuke failed to adequately address her complaints. After unsuccessfully interviewing a second and third time, Ms. Bullington filed suit alleging: (1) United failed to hire her on all three occasions because of her sex and age, (2) United failed to hire her in 1995 in retaliation for her complaints to Ms. Stuke in 1993, and (3) United's failure to hire her breached an implied contract or otherwise enforceable promise for career advancement.

United moved to dismiss Ms. Bullington's claims to the extent they were based on Ms. Bullington's January 1993 rejection because those claims were barred by the statute of limitations. The district court agreed and granted United's motion. United then moved for summary judgment on Ms. Bullington's remaining claims. The district court granted that motion as well, concluding Ms. Bullington failed to establish a prima facie case of age or sex discrimination under either a disparate impact or disparate treatment theory, failed to establish a prima facie case of retaliation, and failed to present sufficient evidence of an enforceable contract or promise. On appeal, Ms. Bullington argues: (1) the district court erred in concluding her claims based on United's 1993 hiring decision were barred by the statute of limitations; (2) genuine issues of material fact exist regarding her claims of sex and age discrimination, retaliation, and breach of contract/promissory estoppel; and (3) the district court erred in using a deferential standard of review in connection with United's hiring decisions.

II. Statute of Limitations

In Colorado, ADEA and Title VII complainants must file a charge of discrimination with the Equal Employment Opportunity Commission ("EEOC") within 300 days after the alleged unlawful discriminatory practice occurred.2 42 U.S.C. § 2000e-5(e)(1); 29 U.S.C. 626(d)(2). This filing is a prerequisite to a civil suit under either statute. Aronson v. Gressly, 961 F.2d 907, 911 (10th Cir. 1992). In this case, United first rejected Ms. Bullington for the position of flight officer in January 1993. In March 1993, Ms. Bullington complained to Ms. Stuke that she suspected one of her interviewers discriminated against her. In order for a claim based on this conduct to be timely, Ms. Bullington was required to file an EEOC charge within 300 days after the March 1993 incident. However, Ms. Bullington waited almost three years, until February 6, 1996, to file her charge.

Ms. Bullington attempts to avoid this apparent untimeliness by invoking the continuing violation doctrine. Under that doctrine, a plaintiff may recover for incidents which occurred outside the statutory time limit if at least one instance of the alleged discriminatory practice occurred within the limitations period and the earlier acts are part of a "continuing pattern of discrimination." Martin v. Nannie and The Newborns, Inc., 3 F.3d 1410, 1415 (10th Cir. 1993). To determine whether alleged incidents of discrimination constitute a continuing violation, a court considers three factors:

  (i) subject matter whether the violations constitute the same type of discrimination; (ii) frequency; and (iii) permanence whether the nature of the violations should trigger an employee's awareness of the need to assert her rights and whether the consequences of the act would continue even in the absence of a continuing intent to discriminate.

Id. at 1415.

Applying these factors, the district court determined that the events arising in 1993 and the later events in 1995 did not constitute a continuing violation. Instead, the court concluded the 1993 non-hire was an isolated event and, moreover, Ms. Bullington had reason to believe she was a victim of discrimination as early as 1993. As such, the court found application of the continuing violation doctrine inappropriate and Ms. Bullington's claims, to the extent they relied on the 1993 conduct, untimely. The court therefore dismissed those claims for failure to state a claim upon which relief may be granted.3 We review de novo the district court's dismissal for failure to state a claim upon which relief can be granted. Mascheroni v. Board of Regents, 28 F.3d 1554, 1560 (10th Cir. 1994). We uphold a dismissal "only when it appears that the plaintiff can prove no set of facts in support of the claims that would entitle [her] to relief, accepting the well-pleaded allegations of the complaint as true and construing them in the light most favorable to the plaintiff." Yoder v. Honeywell, Inc., 104 F.3d 1215, 1224 (10th Cir.) (internal quotation marks and citation omitted), cert. denied, 118 S. Ct. 55 (1997).

The continuing violation doctrine "is premised on the equitable notion that the statute of limitations should not begin to run until a reasonable person would be aware that his or her rights have been violated." Martin, 3 F.3d at 1415 n.6. Thus, a continuing violation claim will likely fail if the plaintiff knew, or through the exercise of reasonable diligence would have known, she was being discriminated against at the time the earlier events occurred. See id. We agree with the district court's conclusion that, although the 1993 conduct is of the same general type as the 1995 conduct, the 1993 decision was a discrete and salient event that put Ms. Bullington on notice that United violated her rights. The allegations contained in Ms. Bullington's Amended Complaint clearly indicate that after United declined to hire her in January 1993, she spoke to Ms. Stuke in March 1993 and expressed her opposition to "what she believed ... to have been sex and age discrimination by [United] in not selecting her for the position of line pilot." Because Ms. Bullington was, at the very least, on inquiry notice of the alleged discrimination as early as 1993, she had a duty to assert her rights at that time and she cannot rely on a continuing violation theory to avoid the statutory time bar.4

Ms. Bullington also argues for a continuing violation based on her statistical evidence of a pattern and practice of discrimination. It is true that a continuing violation may be based on either a series of related acts taken against a single individual or the maintenance of a company-wide policy or practice of discrimination. See Purrington, 996 F.2d at 1028. However, Ms. Bullington's argument below focused entirely on the specific acts taken against her and did not contend that a company-wide policy of discrimination existed before and after the limitations period. Moreover, she did not present her statistical evidence until well after the district court issued its order granting partial dismissal. Cf. John Hancock Mut. Life Ins. Co. v. Weisman, 27 F.3d 500, 506 (10th Cir. 1994) ("This court has held that it cannot, in reviewing a ruling on summary judgment, consider evidence not before the district court."). Because Ms. Bullington failed to properly assert this theory, we decline to review it on appeal. See Bancamerica Commercial Corp. v. Mosher Steel of Kansas, Inc., 100 F.3d 792, 798-99 (10th Cir. 1996) ("Where a litigant changes to a new theory on appeal that falls under the same general category as an argument presented at trial or presents a theory that was discussed in a vague and ambiguous way the theory will not be considered on appeal." (Internal quotation marks and citation omitted).

The district court properly dismissed Ms. Bullington's discrimination claims for events arising in 1992/1993 as time barred.5 We therefore turn to our review of the district court's grant of summary judgment on Ms. Bullington's remaining claims.

III. Disparate Impact

Ms. Bullington alleges both disparate treatment and disparate impact claims under Title VII. A disparate impact claim involves employment practices that are "fair in form, but discriminatory in operation." Griggs v. Duke Power Co., 401 U.S. 424, 431 (1971). A disparate impact claim differs from a disparate treatment claim in that it does not require a showing of discriminatory intent. See Ortega v. Safeway Stores, Inc., 943 F.2d 1230, 1242 (10th Cir. 1991). Instead, a plaintiff may establish a prima facie case of disparate impact discrimination by showing that a "specific identifiable employment practice or policy caused a significant disparate impact on a protected group." Id.; 42 U.S.C. § 2000e-2(k)(1)(A)(i). This prima facie case, in many respects, is more rigorous than in a disparate treatment case because a plaintiff must not merely show circumstances raising an inference of discriminatory impact but must demonstrate the discriminatory impact at issue. See Regner v. City of Chicago, 789 F.2d 534, 537 (7th Cir. 1986) (internal quotation marks and citation omitted). If plaintiff establishes a prima face case, the burden shifts to defendant to show that the challenged practice is job related and consistent with business necessity. 42 U.S.C. § 2000e-2(k)(1)(A)(i); Ortega, 943 F.2d at 1243-44. If defendant meets this burden, it is then up to the plaintiff to suggest an alternative employment practice that serves the employer's legitimate employment goals yet lacks the undesirable discriminatory effect. Ortega, 943 F.2d at 1244.

Ms. Bullington argues United's interview process caused a significant disparate impact on women. As is typical in disparate impact cases, Ms. Bullington relies on statistical evidence to establish her prima facie case. Her statistics compare the "pass rates" of male and female applicants who interviewed for United flight officer positions. The "pass rate," as defined by Ms. Bullington's expert, represents the number of applicants who received an overall score of "3" or better on the interview.6 For interviews conducted after 1994, the pass rate for women was 27.9% while the pass rate for men was 46.6%. As such, the women's pass rate is equal to only 60% of the pass rate for men a statistically significant disparity under EEOC guidelines. See 29 C.F.R. § 1607.4(D) (stating that a selection rate for a protected group which is less than 80% or 4/5 of the selection rate for the majority group is generally regarded as evidence of adverse impact). This disparity, Ms. Bullington argues, is significant enough to establish a prima facie case of disparate impact discrimination.

The district court disagreed. It determined Ms. Bullington's statistics did not establish a prima facie case because they failed to compare similarly situated individuals. Specifically, the court noted that women interviewees for flight officer positions generally have less aeronautical experience than male interviewees. Although its reasoning is somewhat unclear, the court apparently assumed that the male interviewee's advanced aeronautical experience level allowed them to perform better during interviews, thereby accounting for the higher male pass rate. Because Ms. Bullington's statistics did not factor in this difference in experience level, the district court found the analysis "not particularly meaningful," and granted United's motion for summary judgment.

We review the district court's grant of summary judgment de novo, applying the same legal standard as the district court. Simms v. Oklahoma, 165 F.3d 1321, 1326 (10th Cir. 1999), petition for cert. filed (U.S. May 24, 1999) (No. 98-1884). We "examine the record to determine if any genuine issue of material fact was in dispute; if not, we determine if the substantive law was correctly applied." Applied Genetics Int'l, Inc. v. First Affiliated Sec., Inc., 912 F.2d 1238, 1241 (10th Cir. 1990). In applying this standard, we view the factual record and inferences therefrom in the light most favorable to the nonmoving party. Simms, 165 F.3d at 1326. However, to survive summary judgment, the nonmoving party may not rest upon the allegations or denials of his or her pleadings, but must set forth specific facts showing that there is a genuine issue for trial. Id. Summary judgment is appropriate if the evidence is such that no reasonable jury could return a verdict for the nonmoving party. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248 (1986).

Ms. Bullington used a type of statistics called applicant flow data to establish her disparate impact claim.7 Applicant flow data, long recognized as an acceptable comparison model in discrimination cases, generally contrasts the racial or gender composition of persons who applied for the position and persons holding the at-issue jobs. See Wards Cove Packing Co., Inc. v Atonio, 490 U.S. 642, 650-51 (1989) (recognizing that statistics measuring "otherwise qualified applicants" may be probative in disparate impact cases); Hazelwood Schl. Dist. v. United States, 433 U.S. 299, 309 n.13 (1977) (noting that applicant flow data may be "very relevant" in proving discrimination). Such data is generally considered probative because it reflects how the employer's hiring procedure actually operated. See, e.g., Ramona L. Paetzold & Steven L. Willborn, The Statistics of Discrimination, § 4.03 at 7 (1998). Of course, applicant flow data, like all statistical proof, is susceptible to distortion.8 Accordingly, we require the data to cross a "threshold of reliability before it can establish even a prima facie case of disparate impact." Ortega, 943 F.2d at 1243 (internal quotation marks and citation omitted, and emphasis added). The "reliability" or usefulness of any particular analysis will depend on the surrounding facts and circumstances of the case. See Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 995 n.3, 997 (1988).

After examining the facts and circumstances of this case, we find Ms. Bullington's statistical data sufficiently reliable to raise a genuine issue of material fact regarding the existence of a statistical disparity. Her analysis identified a specific employment practice (the interview) and identified two relevant populations for impact comparison persons who interviewed for flight officer positions and persons who received a passing score on the interview. See Wards Cove, 490 U.S. at 650-51 (noting the proper comparison in a disparate impact case is between the racial composition of qualified persons in the labor market and persons holding the at-issue jobs). Her analysis focused on the specific position at issue, namely flight officer. See Ortega, 943 F.2d at 1245-46 (noting importance and difficulty of identifying the at-issue job where plaintiff challenges employer's hiring process in general). In addition, her applicant pool was appropriately limited to persons who sought out and were at least minimally qualified for the position of flight officer. See Wards Cove, 490 U.S. at 651, 653-54 (concluding plaintiff's applicant pool was too broad because it included persons who were not qualified for and did not seek the at-issue jobs). In fact, each member of the applicant pool not only applied for the at-issue position, as is the usual case with applicant flow data, but actually interviewed for the at-issue position. Based on United's interview eligibility requirements, we can therefore assume each member of the applicant pool was, at the very least, a certified pilot with a high school diploma and at least 350 hours of flight experience. The relative homogeneity of the applicant pool reassures us that its members are not so diverse as to render her statistics totally meaningless. See Wards Cove, 490 U.S. at 650-54 (concluding plaintiff's statistics were not probative of disparate impact because applicant pool was at once too broad and too narrow for at issue positions based of differences pool members' skill and interest level); Allen, 881 F.2d at 379 (concluding eligibility requirements for taking at-issue test resulted in applicant pool that was reasonably homogeneous, making large disparity in performance between blacks and whites suggestive of racial bias).

This is not to say Ms. Bullington's statistics are without fault. As the district court noted, her analysis fails to account for differences in male and female interviewees' aeronautical experience a potentially non-discriminatory explanation for the disparate impact. Cf. Fallis v. Kerr-McGee Corp., 944 F.2d 743, 746 (10th Cir. 1991) (concluding statistics offered by plaintiff in disparate treatment case did not support the jury's verdict for plaintiff because statistics failed to eliminate nondiscriminatory explanations for the alleged disparity). However, we do not believe that fault renders Ms. Bullington's statistics incapable of raising a genuine issue of material fact. We emphasize that Ms. Bullington's burden as the nonmovant is to set forth specific facts establishing a genuine issue for trial. Anderson, 477 U.S. at 248 (internal quotation marks and citation omitted). The issue of material fact, here the existence of a significant statistical disparity, need not be resolved conclusively in Ms. Bullington's favor. First Nat'l Bank v. Cities Serv. Co., 391 U.S. 253, 288 (1968). Rule 56 requires only that she present "sufficient evidence supporting the claimed factual dispute ... to require a jury or judge to resolve the parties' differing versions of the truth at trial." Id. at 288-89. Ms. Bullington's statistics, for the reasons discussed above, are sufficiently reliable to make this showing and her failure to include one potentially relevant factor (i.e., education level) does not undermine the probativeness of the statistics to such a degree that no reasonable jury could return a verdict in her favor.9 See Bazemore, 478 U.S. at 400 (noting a statistical analysis that includes less than all measurable variables may serve to prove a plaintiff's case of discrimination); Maitland v. University of Minnesota, 155 F.3d 1013, 1017 (8th Cir. 1998) ("[I]t is for the finder of fact to consider the variables that have been left out of an analysis, and the reasons given for the omissions, and then to determine the weight to accord the study's results."); Smith v. Virginia Commonwealth Univ., 84 F.3d 672, 676-77 (4th Cir. 1996) (concluding omission of variables from statistical analysis and evidence that variables were crucial demonstrated dispute of material fact regarding validity of statistical study).

Nor do we suggest United cannot rebut Ms. Bullington's statistics. We simply conclude the district court's basis for granting summary judgment was insufficient, and United has not shown an absence of issues of material fact with respect to Ms. Bullington's prima facie case. United did attempt to demonstrate the incompleteness of Ms. Bullington's statistics by listing the relative aeronautical experience of male and female flight officers actually hired during the relevant time frame. However, United did not correlate this data with the interview pass rates in any meaningful way and we are left to speculate about the actual impact of experience level on interview performance. Thus, while United's evidence does show a potential weakness in Ms. Bullington's statistics, we are unable to conclude that the evidence is so one-sided that United must prevail as a matter of law. See Anderson, 477 U.S. at 251-52 (stating that relevant inquiry is whether the evidence presents a sufficient disagreement to require submission to a jury or whether it is so one-sided that one party must prevail as a matter of law); Celotex Corp. v. Catrett, 477 U.S. 317, 325 (1986) (stating that movant's burden is to show that there is an absence of evidence to support the nonmoving party's case); Cities Serv., 391 U.S. at 288-90 (concluding that summary judgment was appropriate in the absence of any significant probative evidence tending to support the complaint). Accordingly we conclude the district court's grant of summary judgment in favor of United on this issue was premature.10

IV. Disparate Treatment

Ms. Bullington next asserts that the district court erred in granting summary judgment on her disparate treatment claims. Disparate treatment claims involve "the most easily understood type of discrimination" in which an employer treats an individual less favorably than others because of her protected status. International Bhd. of Teamsters, 431 U.S. at 335 n. 15. Because disparate treatment is a form of intentional discrimination, the plaintiff must prove that her employer acted with a discriminatory intent or motive. See Faulkner v. Super Valu Stores, Inc., 3 F.3d 1419, 1424 (10th Cir. 1993). The basic allocation of burdens for a disparate treatment claim is set forth in McDonnell Douglas Corp. v. Green, 411 U.S. 792 (1973). Under the McDonnell Douglas framework, Ms. Bullington has the initial burden of establishing a prima facie case of discrimination, which in an ADEA or Title VII case requires her to show: "1) she is a member of the class protected by the statute; 2) she suffered an adverse employment action; 3) she was qualified for the position at issue; and 4) she was treated less favorably than others not in the protected class." Sanchez v. Denver Pub. Schls., 164 F.3d 527, 531 (10th Cir. 1998). If she establishes a prima facie case, the burden shifts to United to articulate a legitimate, nondiscriminatory reason for the adverse employment decision. McDonnell Douglas, 411 U.S. at 802-03. If United offers a legitimate, nondiscriminatory reason for its actions, the burden reverts to Ms. Bullington to show United's proffered reason was a pretext for discrimination. Id. at 804-05.

After applying this framework, the district court discerned two bases for granting summary judgment. First, the court concluded that Ms. Bullington failed to establish a prima facie case of disparate treatment. Second, the court determined Ms. Bullington failed to present any evidence that United's proffered reasons for not selecting her were a pretext for discrimination. Of these two bases, the district court appears to have focused more closely on the second, scrutinizing United's proffered reasons for not hiring Ms. Bullington i.e., her poor performance during the interviews and Ms. Bullington's attacks on those proffered reasons. We therefore find it appropriate to assume, for the purposes of this opinion, that Ms. Bullington established a prima facie case of discrimination and proceed directly to the second and third steps of the McDonnell Douglas analysis.11 See V-1 Oil Co. v. Utah State Dep't of Public Safety, 131 F.3d 1415, 1422 (10th Cir. 1997) ("[W]e may affirm for any grounds supported in the record.").

The district court concluded that United chose not to hire Ms. Bullington because she did not meet criteria United believed necessary for a position as flight officer. The court emphasized that all four United representatives who interviewed Ms. Bullington concluded that she did not possess "attributes" United considers essential for flight officers and gave her unsatisfactory ratings in several of the seven "dimensions" United uses to assess flight officer suitability.12 Thus, the district court essentially found that United articulated legitimate, nondiscriminatory reasons for not hiring Ms. Bullington. The record supports this conclusion. The only remaining issue, then, is whether Ms. Bullington has shown "that there is a genuine dispute of material fact as to whether the employer's proffered reason for the challenged action is pretextual." Randle v. City of Aurora, 69 F.3d 441, 451 (10th Cir. 1995).

To establish pretext a plaintiff must show either that "a discriminatory reason more likely motivated the employer or ... that the employer's proffered explanation is unworthy of credence." Texas Dep't of Community Affairs v. Burdine, 450 U.S. 248, 256 (1981). Plaintiff may accomplish this by demonstrating "'such weaknesses, implausibilities, inconsistencies, incoherencies, or contradictions in the employer's proffered legitimate reasons for its action that a reasonable factfinder could rationally find them unworthy of credence.'" Morgan v. Hilti, Inc., 108 F.3d 1319, 1323 (10th Cir. 1997) (quoting Olson v. General Elec. Astrospace, 101 F.3d 947, 951-52 (3d Cir. 1996)). However, the plaintiff's "mere conjecture that [her] employer's explanation is a pretext for intentional discrimination is an insufficient basis for denial of summary judgment." Branson v. Price River Coal Co., 853 F.2d 768, 772 (10th Cir. 1988).

In this case, Ms. Bullington bases her pretext argument on the following evidence: disputes regarding things she said and did during the interview, the interviewers' use of gender and age stereotypes, a comparison of her qualifications with those of successful flight officer interviewees, and statistical evidence. We conclude that, even viewing this evidence in the light most favorable to Ms. Bullington, it fails to demonstrate a genuine issue of fact as to whether United's reasons for not hiring her were pretextual.

First, Ms. Bullington lists numerous disputes she has with the notes and summaries prepared by her interviewers. The disputed "notes" are essentially the United representatives' informal, handwritten notations of Ms. Bullington's responses to various questions during the interview. The "summaries" are typed memos in which the interviewer summarizes and assesses Ms. Bullington's performance and makes a final recommendation regarding hiring. Ms. Bullington claims the notes and summaries are not accurate reflections of her interview performance. Further, she contends that those inaccuracies amount to fact issues regarding the legitimacy of United's proffered reason for not hiring her and that the district court improperly ignored them in granting summary judgment. We disagree.

A review of the record shows that the vast majority of the "factual disputes" alleged by Ms. Bullington are in reality her opinion that the interviewers were wrong in their assessment of her qualifications.13

However, her own opinions about her qualifications do not give rise to a material fact dispute. See Simms, 165 F.3d at 1329. Moreover, even if we were to assume that United misjudged Ms. Bullington's qualifications, such evidence would not preclude summary judgment in this case. The relevant inquiry is not whether United's proffered reasons were wise, fair or correct, but whether United honestly believed those reasons and acted in good faith upon those beliefs.14 See Sanchez v. Philip Morris Inc., 992 F.2d 244, 247 (10th Cir. 1993) ("Title VII is not violated by the exercise of erroneous or even illogical business judgment."); Kariotis, 131 F.3d at 677 ("[A]rguing about the accuracy of the employer's assessment [of plaintiff's performance] is a distraction, because the question is not whether the employer's reasons for a decision are right but whether the employer's description of its reasons is honest." (Internal quotation marks and citations omitted.)); Fischbach v. District of Columbia Dep't of Corrections, 86 F.3d 1180, 1183 (D.C. Cir. 1996) (concluding relevant issue was whether the employer honestly believed in the reasons it offered for not promoting plaintiff and not the correctness or desirability of those reasons). Ms. Bullington presents no evidence indicating United did not believe the interviewers' assessment of her qualifications, and our review of the record reveals no evidence that United failed to act in good faith in reliance on those assessments. We therefore discern no showing of pretext.15

Likewise, we find little merit in Ms. Bullington's argument that United based its employment decisions on gender and age stereotypes. Our review of the record reveals little that could be construed as stereotyped assumptions about the qualifications, work habits, or personality traits of female or older interviewees. See, e.g., Price Waterhouse v. Hopkins, 490 U.S. 228, 251 (1989). Rather, Ms. Bullington's argument merely takes issue with what she believes is an incorrect assessment of her communication skills, goals, and motivation level. As discussed above, Ms. Bullington's opinion about the fairness or accuracy of the interviewers' evaluation is not evidence of pretext. See Simms, 165 F.3d at 1329.

Next, Ms. Bullington offers a comparison of her qualifications with those of seven other male and/or younger individuals that United interviewed and hired as flight officers. She claims United hired these individuals despite the fact many were less qualified than her or, in many cases, had the same deficiencies identified during her interview. As such, she argues, there is reason to disbelieve United's proffered nondiscriminatory reasons for not hiring her. It is true that a comparative analysis of job applicants' qualifications may be relevant in proving pretext where, as here, the employer claims lack of qualification as reason for an employment decision. See Sanchez v. Philip Morris, 992 F.2d at 247. However, we emphasize that an employer does not violate Title VII by choosing between equally qualified candidates, so long as the decision is not based on unlawful criteria. See Burdine, 450 U.S. at 259. Therefore, pretext cannot be shown simply by identifying minor differences between plaintiff's qualifications and those of successful applicants. See Sanchez v. Philip Morris, 992 F.2d at 247-48; Chock v. Northwest Airlines, Inc., 113 F.3d 861, 864 (8th Cir. 1997) ("[A] comparison that reveals that the plaintiff was only similarly qualified or not as qualified as the selected candidate would not raise an inference of racial discrimination."). The disparity in qualifications must be "overwhelming" to be evidence of pretext. Sanchez v. Philip Morris, 992 F.2d at 247; see also Odom v. Frank, 3 F.3d 839, 847 (5th Cir. 1993) (difference in qualifications of applicants must be so apparent as "to jump off the page and slap us in the face" to support a finding of pretext).

A comparison of Ms. Bullington's qualifications with those of the other interviewees in this case gives us no reason to question United's explanation for its hiring decision. Ms. Bullington evidence does not show that she was overwhelmingly better qualified than the other candidates. At most, the seven other candidates were similarly qualified and the fact that United chose between them is not evidence of pretext. See Sanchez v. Philip Morris, 992 F.2d at 247 (concluding no evidence of pretext where plaintiff failed to demonstrate that he was overwhelmingly better qualified that the other applicants).

Last, Ms. Bullington argues that the same statistical analysis of male and female pass rates she offered as evidence of disparate impact is also persuasive evidence of pretext. We disagree. The probative value of statistical evidence varies greatly according to the type of discrimination alleged. In a disparate impact case, statistical evidence plays a central role because the plaintiff is attempting to show a particular practice had a disproportionate impact on a particular group, and not the employer's discriminatory intent. However, in an individual disparate treatment case, the focus is on how and why an employer treated a particular individual the way it did. As such, statistical evidence of the employer's general hiring patterns is considerably less probative. See LeBlanc v. Great American Ins. Co., 6 F.3d 836, 848 (1st Cir. 1993), cert. denied, 511 U.S. 1018 (1994). Moreover, because overall employment statistics have little bearing on the specific intentions of the employer in making particular hiring decisions, such statistical evidence will rarely suffice to rebut an employer's legitimate, nondiscriminatory reasons for a particular adverse employment action. See id. at 848 (statistical evidence did not provide sufficient basis for a reasonable jury to find that defendant terminated plaintiff because of his age); Equal Employment Opportunity Comm'n v. Texas Instruments, Inc., 100 F.3d 1173, 1185 (5th Cir. 1996) (statistical evidence did not support an inference that defendant employer's reasons for terminating plaintiffs was merely pretextual).

In the instant case, we find the statistics insufficient to allow a reasonable trier of fact to infer that United's proffered reasons were a pretext for discrimination. First, the statistics do not compare the ages of successful and unsuccessful interviewees and are thus totally irrelevant to Ms. Bullington's disparate treatment claim under the ADEA. Second, the statistics were designed to reflect the generalized impact of United's interview process on women applicants. They do not address United's specific reasons for not hiring Ms. Bullington, namely her poor interview performance and lack of qualifications. The statistics thus shed little light on the central issue of the pretext analysis the motive behind United's decision not to hire Ms. Bullington. See id., 100 F.3d at 1185 (plaintiffs' statistics were not evidence of pretext because they did not even purport to analyze the pertinent facts surrounding plaintiffs' terminations); Gadson v. Concord Hosp., 966 F.2d 32, 35 (1st Cir. 1992) (statistics did not meet plaintiff's pretext burden because there was no indication of a connection between the statistics and the employer's treatment of plaintiff). This is not to say that the evidence is totally irrelevant to the pretext issue. See McDonnell Douglas, 411 U.S. at 804-05 (including statistics in list of evidence available to plaintiff to show pretext); Bruno v. W.B. Saunders Co., 882 F.2d 760, 767 (3d Cir. 1989) (plaintiff's statistical analysis was relevant and thus admissible on issue of pretext), cert. denied, 493 U.S. 1062 (1990). However, considering the limited insight it gives us into United's motive in not hiring Ms. Bullington and the lack of other probative evidence specifically rebutting United's proffered reasons, we conclude that the statistics are insufficient to demonstrate a genuine fact issue in this case. See Texas Instruments, 100 F.3d at 1186 (statistical evidence is only probative of intent when combined with other evidence specifically rebutting the defendant's legitimate, nondiscriminatory reasons).

In sum, we hold that Ms. Bullington has adduced insufficient evidence for a reasonable trier of fact to infer that United's decision not to hire her was motivated by sex or age animus. The district court's decision to grant summary judgment in United's favor on Ms. Bullington's disparate treatment claims was thus proper.

V. Retaliation

Ms. Bullington claims United retaliated against her based on a conversation she had in March 1993 with Ms. Stuke, United's Manager of Flight Officer Employment. During this conversation, Ms. Bullington allegedly informed Ms. Stuke of her "strong concerns" that one of her interviewers in the 1993 interview was biased against her. Because of this complaint, Ms. Bullington claims Ms. Stuke retaliated against her by influencing the interviewers' hiring decisions in her 1995 interviews. As proof of Ms. Stuke's animus towards her, Ms. Bullington points to a conversation which occurred shortly before her March 1995 interview between Ms. Stuke and Mr. H. Jeffery Bartels about a recommendation Mr. Bartels submitted in support of Ms. Bullington's flight officer application. In his affidavit, Mr. Bartels states that Ms. Stuke asked him if he was sure he wanted to submit a recommendation for Ms. Bullington and told him that Ms. Bullington acted like a "real airhead" and "held a troll doll for good luck" during her previous interview. Ms. Stuke also allegedly suggested that if Mr. Bartels wanted to assist Ms. Bullington, he would help her be "more professional" in her next interview.

To establish a prima facie case of retaliation, Ms. Bullington must show: "1) she was engaged in protected opposition to Title VII or ADEA discrimination; 2) she was subjected to adverse employment action; and 3) a causal connection existed between the protected activity and the adverse employment action." See Sanchez v. Denver Pub. Schls., 164 F.3d at 533. The causal connection may be shown by producing "evidence of circumstances that justify an inference of retaliatory motive, such as protected conduct closely followed by adverse action." Burrus v. United Tel. Co. of Kansas, Inc., 683 F.2d 339, 343 (10th Cir.), cert. denied, 459 U.S. 1071 (1982). In other words, Ms. Bullington must present some evidence that her employer undertook the adverse employment action for the purpose of retaliation. See Randlett v. Shalala, 118 F.3d 857, 862 (1st Cir. 1997) ("[T]he adverse action must have been taken for the purpose of retaliating."); see also Willis v. Marion County Auditor's Office, 118 F.3d 542, 546-47 (7th Cir. 1997) (concluding plaintiff failed to demonstrate that relevant decision maker harbored retaliatory animus towards plaintiff); Kneibert v. Thomson Newspapers, Michigan Inc., 129 F.3d 444, 455-56 (8th Cir. 1997) (concluding plaintiff presented genuine issue of material fact where statements by employee involved in decision making process revealed retaliatory motive).

The district court concluded summary judgment in favor of United was appropriate because Ms. Bullington failed to establish a nexus or causal connection between her 1993 complaint to Ms. Stuke and the interviewers' hiring selections in 1995. We agree. The interviewers' decisions in 1995 were remote in time from Ms. Bullington's 1993 complaint, thus undercutting an inference of retaliatory motive. See Burrus, 683 F.2d at 343. In addition, none of Ms. Stuke's comments to Mr. Bartels, on their face, indicate a discriminatory animus towards Ms. Bullington and we do not believe a reasonable jury could infer such a motive based on that conversation alone. See Little v. Cox's Supermarkets, 71 F.3d 637, 643 (7th Cir. 1995) ("[C]ourts are not required to evaluate every conceivable inference which can be drawn from evidentiary matter, but only reasonable ones." (Internal quotation marks and citation omitted.)). Moreover, even if we assume that Ms. Stuke harbored some retaliatory animus, that evidence would not establish the requisite causal link because we find no evidence that Ms. Stuke played any part in the adverse employment decisions. Instead, the record reveals that the four interviewers were solely responsible for making the hiring recommendations, and we find no evidence that a retaliatory animus motivated their decisions or that they were even aware of Ms. Bullington's 1993 complaint. Ms. Bullington claims Ms. Stuke "was in a position to influence the interviewers' recommendations" because she selected them to perform the interviews and supervised them. However, evidence of an opportunity to influence does not amount to evidence of actual influence and Ms. Bullington's mere speculation about Ms. Stuke's influence is not enough to demonstrate a genuine issue of material fact. See Branson, 853 F.2d at 772 ("[T]he plaintiffs' mere conjecture that their employer's explanation is a pretext for intentional discrimination is an insufficient basis for denial of summary judgment."); see also Willis, 118 F.3d at 547 (affirming summary judgment of plaintiff's retaliation claim where no reasonable juror could conclude that relevant decision maker acted with retaliatory motive or was a conduit for other employees' racial animosity). Because Ms. Burlington has failed to provide any evidence demonstrating a causal connection between the adverse employment decisions and her protected activity, she has failed to establish a prima facie case of retaliation.

VI. Breach of Contract/Promissory Estoppel

Lastly, Ms. Bullington claims the district court erred in granting summary judgment on her breach of contract or promissory estoppel claims. Ms. Bullington contends the United representative who initially interviewed her for the position of academic instructor told her United loved to hire from within and after one year's work, she would be "almost guaranteed a job." In addition, Ms. Bullington claims the second United representative who interviewed her for the flight instructor position told her "it will take you a year and they'll hire you as a pilot," and the United Vice President for Flight Standards and Training told all academic instructors on many occasions United employees could and usually did work their way up. Ms. Bullington argues these statements amount to a "promise[] of preferential treatment and a virtual guarantee of employment as a flight officer."Under Colorado law, an implied contract can arise out of a company's personnel policies and procedures. See Vasey v. Martin Marietta Corp., 29 F.3d 1460, 1464 (10th Cir. 1994). To establish such a contract, an employee must show that the employer's actions manifested an intent to be bound. Id. To establish an employer's binding intent, the terms of the offer must be sufficiently definite and detailed to enable the court to determine whether a contract has been performed. Id. at 1465. Terms which are nothing more than "vague assurances" by the employer will not suffice. Id. If the employee fails to show an implied contract, she may nevertheless attempt to enforce her employer's policies under a theory of promissory estoppel. Id. at 1466. To succeed under that theory, the employee must demonstrate that the employer should have reasonably expected the employee to consider the policy as a commitment from the employer, that the employee reasonably relied on the statements to her detriment, and that injustice can be avoided only by enforcement of the policy. Orback v. Hewlett-Packard Co., 97 F.3d 429, 433 (10th Cir. 1996), cert. denied, 520 U.S. 1241 (1997).

After a thorough review of the record, we conclude that no reasonable juror could find that the statements Ms. Bullington allegedly relied on amount to an enforceable contract or promise. The statements, at most, constitute vague assurances about career advancement opportunities and a general preference for promoting from within. See Vasey, 29 F.3d at 1465. They do not manifest an intent to contract and any reliance upon them as creating a contract or promise was unwarranted. See id. at 1466; Orback, 97 F.3d at 433; see also Dobbs v. Chevron U.S.A., Inc., 39 F.3d 1064, 1069 (10th Cir. 1994) (applying similar state law to conclude vague oral statements by employer did not amount to contract for continued employment). Accordingly, we agree with the district court that, as a matter of law, the statements are not sufficiently definite to be legally enforceable representations. Summary judgment was therefore appropriate.

The district court's partial dismissal of Ms. Bullington's claims on statute of limitations grounds and its grant of summary judgment in favor of United on the disparate treatment, retaliation, and breach of contract/promissory estoppel claims is AFFIRMED. The district court's grant of summary judgment on the disparate impact claim is REVERSED and that claim is REMANDED for further proceedings.

Notes:

1 For example, under the dimension "leadership," United lists "responsibility" as a desired attribute or anchor. To evaluate whether an applicant meets United standards, United suggests the following questions, "What has been your most disappointing leadership experience? What did you learn from that situation?" The interviewer then evaluates the applicant's answer according to United's guidelines for that anchor. An applicant meets United's "responsibility" standards if she accepts responsibility for her own actions. The applicant exceeds United standards if she accepts responsibility for her own and her subordinate's actions and initiates corrective action. The applicant falls below United standards if she blames others for non-performance or failure.

2 The 300-day filing period applies to "deferral states" in which the EEOC defers to the enforcement efforts of a state agency empowered to undertake employment discrimination investigations. 42 U.S.C. § 2000e-5. Otherwise, the filing period is 180 days. Id. 3 We note that Rule 12(b)(6) is a proper vehicle for dismissing a complaint that, on its face, indicates the existence of an affirmative defense such as noncompliance with the limitations period. See Robinson v. Dalton, 107 F.3d 1018, 1021 (3d Cir. 1997) ("A complaint does not state a claim upon which relief may be granted unless it asserts the satisfaction of the precondition to suit specified by Title VII."); 5A Charles Alan Wright & Arthur R. Miller, Federal Practice and Procedure § 1357 at 349-52 (2d ed. 1990) ("The complaint also is subject to dismissal under Rule 12(b)(6) when its allegations indicate the existence of an affirmative defense.").

4 Our decision in Martin does not require a different result. In Martin, we concluded that plaintiff's allegations of a consistent and frequent pattern of sexual harassment were sufficient to raise a triable issue on her continuing violation claim even though she failed to demonstrate all three factors in the continuing violation analysis. 3 F.3d at 1415-16. This conclusion was possible because the three-factor analysis we adopted in Purrington v. University of Utah is not a bright line test. 996 F.2d 1025, 1028 (10th Cir. 1993). Rather, the factors are merely tools to assist courts in analyzing whether incidents of discrimination "constitute a continuing course of discrimination or whether they are discrete unrelated acts." Martin, 5 F.3d at 1415. A particular court's use of the factors will vary according to facts and procedural history of the case. Here, the district court focused on what it considered to be the greatest weakness in Ms. Bullington's continuing violation argument her apparent awareness of the alleged discrimination in 1993. The court's emphasis on the third, "permanence" factor was not inappropriate considering the nature of Ms. Bullington's allegations and the underlying purpose of the continuing violation doctrine discussed above. See Martin, 3 F.3d at 1415 n.6; see also Selan v. Kiley, 969 F.2d 560, 565 (7th Cir. 1992) (stressing the significance of the third, "permanence" factor in a continuing violation analysis).

5 Ms. Miller does not appeal nor do we consider the district court's dismissal of that portion of the implied contract claim based on events arising in 1992/1993.

6 As such, Ms. Bullington's statistics do not reflect the number of men and women United actually hired as pilots, since not all applicants who received a "passing" score were hired or even made an offer.

7 We note Ms. Bullington's data differ somewhat from traditional applicant flow data in that it compares persons who interviewed for the at issue position with persons who received a passing score, rather than comparing persons who applied for the at issue position with persons who were hired.

8 For example, an employer's hidden hiring preferences may indirectly affect the overall make-up of the applicant pool. See, e.g., International Bhd. of Teamsters v. United States, 431 U.S. 324, 369 (1977) (noting employer's reputation for not hiring blacks may have deterred employees from applying thereby distorting the applicant pool); Allen v. Seidman, 881 F.2d 375, 379 (7th Cir. 1989) (noting employer's affirmative action plan may cause applicant pool to overassess the availability of qualified minority applicants). In addition, applicant flow data may fail to account for important variations within the pool itself such as differences in education level and experience. Failure to account for these variables may result in statistical anomalies that are not really probative of discrimination. See, e.g., Coward v. ADT Sec. Sys. Inc., 140 F.3d 271, 312 (D.C. Cir. 1998) (finding plaintiff's regression analysis so incomplete as to be irrelevant because it failed to account for key variables in the applicant pool); see also Bazemore v. United States, 478 U.S. 385, 400 (1986) (noting that the omission of variables from a regression analysis may render the analysis less probative).

9 We recognize that some statistical analyses may be so incomplete as to be irrelevant. See Bazemore, 478 U.S. at 400 n.10. Such an analysis would be insufficient to defeat summary judgment. See Anderson, 477 U.S. at 249-50 (noting summary judgment may be granted if evidence offered by nonmovant is merely colorable or not significantly probative); Coward, 140 F.3d at 312 (granting summary judgment for defendant employer where plaintiffs' statistical analysis was flawed as a matter of law). However, for the reasons discussed above, we do not find that to be the case here.

10 Because the district court never reached the second step of the disparate impact analysis, we decline to address any arguments relating to business necessity/job-relatedness of the interview process or the proper level of deference the court should give to United's showing under Spurlock v. United Airlines, Inc., 475 F.2d 216 (10th Cir. 1972).

11 United argues we should consider these arguments at the prima facie stage of the McDonnell Douglas analysis because they relate to Ms. Bullington's qualifications for the at-issue position. Our decision in Kenworthy v. Conoco, Inc. precludes such an approach. 979 F.2d 1462 (10th Cir. 1992). In Kenworthy, the district court entered judgment in favor of the defendant on plaintiff's discrimination claims because it determined plaintiff failed to establish a prima facie case. Id. at 1469. However, the court's reasoning focused on the legitimate non-discriminatory reasons offered by defendant for not promoting the plaintiff, specifically her failure to meet qualifications for the at-issue position. Id. Such a legal analysis, we concluded, was in error because it inappropriately short-circuited the McDonnell Douglas framework at the prima facie stage and frustrated the plaintiff's ability to establish that the defendant's proffered reasons were pretextual. Id. at 1469-70.

To avoid this result, we held that the defendant's reasons for not promoting the plaintiff i.e., her failure to meet subjective criteria for the at-issue position should be considered in addressing whether those articulated reasons are legitimate or pretextual and not as a challenge to the sufficiency of plaintiff's prima facie case. Id. at 1470; see also Thomas v. Denny's, Inc., 111 F.3d 1506, 1510-11 (10th Cir.) (noting the employer's subjective reasons for not promoting plaintiff are not properly considered at the prima facie stage), cert. denied, 118 S. Ct. 626 (1997). At the prima facie stage, the court need only conclude that the plaintiff has shown through credible evidence, including her own testimony, that she was minimally qualified for the position she sought, even if the defendant disputes that evidence. Id. In this way, the court could fully consider the defendant's evidence of legitimate nondiscriminatory reasons for the adverse action as well as the plaintiff's evidence that those proffered reasons were pretextual. Id. This reasoning applies with equal force to the instant case. We will not allow United to "short circuit" the McDonnell Douglas analysis by challenging Ms. Bullington's qualifications at the prima facie stage. Such an approach would unduly limit our ability to consider Ms. Bullington's evidence that United's assessment of her qualifications was a pretext for discrimination.

12 Specifically, United interviewers gave Ms. Bullington unacceptable scores in six of the seven dimensions. Among other alleged deficiencies, United claims Ms. Bullington's presentation and interpersonal skills were very poor, that her responses indicated a significant lack of confidence, focus and motivation, and that each interviewer had serious doubts about her leadership abilities, problem-solving skills, dependability and ability to focus on goals. Ms. Bullington also received failing scores in her technical evaluations.

13 By and large, Ms. Bullington claims the interviewers relied on misinterpretations, inaccurate assumptions, and highly subjective evaluations, and twisted her responses out of context. For example, one United interviewer indicated that Ms. Bullington demonstrated an inability to focus on her career and education goals and that her background raised serious concerns about her dependability and long term commitment to United. In response, Ms. Bullington "disagrees" with the interviewer's comments and states that the interviewer failed to giver her credit for the fact that she put her job duties at United ahead of her educational goals or that she had completed her G.E.D. and flight school while raising seven children and working several jobs. Ms. Bullington obviously has her own opinions about her qualifications for the flight officer position. Those subjective opinions, however, do not demonstrate a fact dispute about the genuineness of United's assessment of her qualifications. See Kariotis v. Navistar Int'l Transp. Corp., 131 F.3d 672, 677 (7th Cir. 1997) ("[A]n opportunity for rebuttal is not an invitation to criticize the employer's evaluation process or simply to question its conclusion about the quality of an employee's performance.").

14 Ms. Bullington argues the district court, relying on our decision in Spurlock, 475 F.2d 216, improperly granted special deference to United's employment decisions based on the economic and human risks involved in hiring competent pilots. We disagree. First, we note that Spurlock is a disparate impact case and that the discussion of risk factors contained therein relates to the second prong of the disparate impact case business necessity an issue the district court never reached in this case. See Spurlock, 475 F.2d at 218-20. Second, unlike Ms. Bullington, we find no error in the district court's pronouncement that it "is not particularly well-suited to evaluating the qualifications that go into making a successful line pilot at United Airlines" and that it "is peculiarly ill-suited to determine the best candidates." We previously have recognized that district courts, when analyzing the pretext issue, do not sit as "super-personnel departments" free to second-guess the business judgment of an employer. Simms, 165 F.3d at 1330. Thus, the court's role in this case was not to determine if United's hiring decisions were wise or fair, but rather, in the context of Ms. Bullington's disparate treatment claim, if those decisions were motivated by discriminatory animus. The court gave no improper, "special deference" to United's business decisions in determining whether they were a pretext for discrimination.

15 Ms. Bullington also contends the interviewers "did not report her interview responses accurately, misinterpreted what she had said and twisted her responses out of context" and claims that some of the interviewers' summaries are inconsistent with their notes or with their testimony in this case. We have reviewed these alleged inconsistencies and find them to be insignificant at best and too minor to give rise to an inference of pretext. Cf. Lucas v. Dover Corp., 857 F.2d 1397, 1402 (10th Cir.1988) (concluding instances of alleged contradictions and inconsistencies were too insubstantial to allow a reasonable jury to infer pretext).