Denise Coleman v. Patrick R. Donaho

HAMILTON, Circuit Judge. In 2006, the United States Postal Service terminated plaintiff Denise Coleman’s 32 years of employment as a mail process *841 ing clerk. The Postal Service contends that it fired Coleman because she told her psychiatrist she was having thoughts of killing her supervisor, and it believed she posed a danger to her fellow employees. Coleman alleges that her termination was discriminatory (she is African American and a woman) and retaliatory (she had previously complained, both formally and informally, of discriminatory treatment). In support of her disparate treatment claims under Title VII of the Civil Rights Act of 1964, Coleman presented evidence that two white male employees at the same facility had recently threatened another employee at knife-point, yet received only one-week suspensions from the same manager who fired her. The district court found that these comparator employees were not similarly situated to Coleman because they had different direct supervisors and held different positions. Coleman therefore failed, in the district court’s view, to establish a prima facie case of discrimination under the “indirect method” of proof derived from McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). The district court also held that Coleman had not provided any evidence that the Postal Service’s stated reason for firing her — that she violated its rule prohibiting workplace violence and threats— was pretextual. The district court therefore granted the Postal Service’s motion for summary judgment on all claims. Coleman appeals. We reverse summary judgment on Coleman’s discrimination claims and her retaliation claims. This appeal raises two recurring questions concerning comparator evidence in employment discrimination cases using the indirect method of proof: First, just how alike must comparators be to the plaintiff to be considered similarly situated? Second, can evidence that a similarly situated employee received better treatment serve not only as an element of the plaintiff’s prima facie case, but also satisfy the plaintiffs burden to show that the employer’s legitimate nondiscriminatory reason for its action was pretextual? For the first question, we reiterate here that the similarly-situated inquiry is flexible, common-sense, and factual. It asks “essentially, are there enough common features between the individuals to allow a meaningful comparison?” Humphries v. CBOCS West, Inc., 474 F.3d 387, 405 (7th Cir.2007), ajfd, 553 U.S. 442, 128 S.Ct. 1951, 170 L.Ed.2d 864 (2008). There must be “sufficient commonalities on the key variables between the plaintiff and the would-be comparator to allow the type of comparison that, taken together with the other prima facie evidence, would allow a jury to reach an inference of discrimination.” Id. In other words, the proposed comparator must be similar enough to permit a reasonable juror to infer, in light of all the circumstances, that an impermissible animus motivated the employer’s decision. Here, Coleman’s two white, male coworkers were disciplined by the same decisionmaker, subject to the same code of conduct, and disciplined more leniently for violating the same rule as she. Their case is close enough to Coleman’s to provide a “meaningful comparison” and to permit a reasonable jury to infer discrimination. Id. The answer to the second question is yes. In McDonnell Douglas itself, the Supreme Court noted that comparator evidence would be “[ejspecially relevant” at the pretext stage. 411 U.S. at 804, 93 S.Ct. 1817. Under our circuit precedents, too, an employment discrimination plaintiff may demonstrate pretext by providing evidence that a similarly situated employee outside her protected class received more favorable treatment. Coleman has done *842 so. The evidence of selective application of the rule against violence and threats to Coleman — -whose confidential expressions of anger during inpatient psychotherapy were not direct threats at all, and who was discharged as stable before the Postal Service even heard about those thoughts— undercuts the Postal Service’s assertion that it was just neutrally enforcing its “no tolerance” policy. Together with other evidence calling into question the honesty of the Postal Service’s rationale, Coleman’s comparator evidence presents a jury question as to pretext. I. Factual and Procedural Background In assessing whether the Postal Service is entitled to summary judgment, we examine the record in the light most favorable to Coleman, the non-moving party, resolving all evidentiary conflicts in her favor and according her the benefit of all reasonable inferences that may be drawn from the record. O’Leary v. Accretive Health, Inc., 657 F.3d 625, 630 (7th Cir.2011). Our account of the facts therefore is not necessarily true in an objective sense, but reflects the standard that applies to motions for summary judgment. Coleman began working for the Postal Service in 1974. She had a good employment record until January 2005, when her longtime supervisor retired. William Berry was selected as the replacement by William Sove, the plant’s maintenance manager. Sove is white; Berry is black. Coleman believed Sove had passed over her for the promotion because she was female. She also felt Berry was treating her poorly in his new supervisory role. She related these complaints in an April 2005 email to Gregory Johnson, the head of the facility where she worked. The following month, Coleman emailed Sove, accusing him and Berry of discrimination and threatening to file a charge with the Equal Employment Opportunity Commission (EEOC). On June 5, 2005, Coleman learned that she would soon undergo surgery. Two days later, she submitted a request to Johnson and Sove to advance her two weeks of future paid sick leave for her convalescence. The same day, Berry directed Coleman to clean an especially dingy area behind a storeroom and to move some heavy boxes — tasks, she says, that were not among her regular duties. Coleman refused, telling Berry that she was unable to lift the boxes because of her upcoming surgery and that the storeroom’s chemicals and dust would exacerbate her chronic asthma. Berry issued Coleman a “Letter of Warning” for failing to follow instructions. On June 9, 2005, Johnson denied her request for advanced sick leave. As scheduled, Coleman had surgery on June 10, 2005. She returned to work on June 23, 2005, subject to the medical restriction that she avoid climbing stairs for two weeks. Because Coleman’s usual work station was up one flight of stairs, Berry informed her that she could work in the ground-floor storeroom, but because of her asthma this was not an attractive alternative to Coleman. When she rejected it, Berry sent her home. She returned to the mail facility a week later with revised medical restrictions permitting her to climb stairs once or twice per day. But Berry then told Coleman that all employees had to clock in using a particular time-clock — -a change that would require her taking more than the maximum stairs she was advised to ascend. She again left work. The following week, Berry issued Coleman an absent-without-leave notice because she had not worked or announced her absence in five days. As this conflict unfolded, Coleman filed an EEO request for pre-complaint counseling on June 21, *843 2005, identifying Berry and Sove as the discriminating officials. She supplemented her request with additional information on July 1, 2005. On July 12, 2005, Coleman checked herself into the psychiatric unit of a hospital complaining of depression, anxiety, and insomnia. In her admission interview, Coleman experienced “severe crying spells, helplessness, [and] hopelessness with suicidal ideation.” The treating psychiatrist, Dr. Ofelia Ionescu, observed Coleman’s “extremely paranoid/obsessional thinking about being harassed by her supervisor, Mr. Berry,” and she described Coleman as “endorsing ... homicidal ideation ‘every time I’m talking about him [Berry].’” Coleman remained at the hospital for three weeks while she received talk therapy and various medication. The course of treatment did her good. When she was discharged on August 3, 2005, Coleman displayed “a marked reduction in depression and in particular the paranoid symptoms” and “a reasonable control for her anger and aggression.” In her final report, Dr. Ionescu described Coleman as a “model patient” in “stable” condition: “Alert, awake, ... oriented ... cooperative, [and] pleasant.... No formal thought disorder. Affect was reactive, smiling. Mood was ‘good.’ There were no reports of delusionsf,] ... hallucinations^] ... [or] suicidal or homicidal ideation.” But on the day of Coleman’s discharge, Dr. Ionescu returned a phone call from Berry, who had called to ask about Coleman’s treatment. In her final report, Dr. Ionescu wrote: “I did inform Mr. Berry that I am not discussing with him about [sic] my patient; but it was considered to be my responsibility [sic] as the patient’s physician to warn him that my patient had been expressing threats to his life in my presence.” The content and form of these “threats” remain something of a mystery: the record contains no elaboration from Dr. Ionescu beyond the vague “homicidal ideation” language in the discharge report. Coleman claims she never formed any plan to harm Berry and that a “language barrier” caused Dr. Ionescu, whom she described as “a foreigner,” to “take me literally.” Coleman Dep. 82. Berry immediately relayed the phone conversation with Dr. Ionescu to Sove and another upper-level manager, Charles Von Rhein. That same day, the day that Coleman was released, the three managers then decided to place her in “emergency off-duty status” without pay. Two weeks later, Berry notified the police. According to the police report, Berry explained that the Postal Service was “in the process of terminating Coleman,” and Berry wanted to “document the threat.” In October 2005, the Postal Service did an internal investigation. Berry told a postal inspector that “he hoped that Coleman would get better and maybe return to work one day.” Although Berry would later claim that he was “frightened, afraid and scared” by what he took to be “a very credible threat,” he did not express such fears to either the police or the Postal Service investigators. He also failed even to mention Coleman’s supposed threat in an email about his conflict with her, though he sent it just days after his phone call with Dr. Ionescu. While off-duty, Coleman filed two formal EEOC complaints. The first, lodged on August 13, 2005, alleged that Berry, Sove, and Von Rhein had discriminated against her on the basis of race and sex by refusing to accommodate her post-surgery medical restrictions and by denying her request for advanced sick leave. The second charge, filed on December 8, 2005, claimed that Berry and Von Rhein placed Coleman on off-duty status because of disability and *844 sex discrimination and to retaliate against her for her first EEOC complaint. Meanwhile, the Postal Service’s own investigation proceeded. According to the postal inspector’s investigative memorandum of October 11, 2005, Coleman admitted she had told her psychiatrist that she felt suicidal and homicidal, but said she had never hurt anyone or formed any sort of plan to harm Berry. As part of the internal investigation, Coleman also participated in a telephone interview with Von Rhein on December 13, 2005. She confirmed having had “homicidal thoughts” about Berry, but indicated that she was continuing in outpatient therapy and was ready to return to work. Von Rhein told Coleman that she had failed to provide documentation of her improved conditions. On December 20, 2005, a psychiatric resident then treating Coleman faxed Sove to confirm that she was “stable” and “able to return to her work duties,” provided it was not “under the supervision of ... Berry.” On January 13, 2006, Coleman was fired. Von Rhein and Sove both signed the “Notice of Removal,” which stated that the termination was based on “unacceptable conduct, as evidenced by your expressed homicidal ideations toward a postal manager.” The notice stated that by having voiced her threats toward Berry, Coleman had violated the Postal Service’s ban on “Violent and/or Threatening Behavior.” The rule provides: “it is the unequivocal policy of the Postal Service that there must be no tolerance of violence or threats of violence by anyone at any level of the Postal Service.” The notice also informed Coleman of her right to file a grievance challenging her removal, and she did so. The matter proceeded to arbitration a year later. In the hearing, her union challenged the Postal Service’s characterization of Coleman’s statements as “a true threat” and contended that the more appropriate action would have been to refer her for a fitness-for-duty examination. The arbitrator agreed, finding that the Postal Service had lacked “just cause” to terminate Coleman because it could not prove that she “actually had an intent to harm Mr. Berry.” Considering that “psychological illness” was the cause of Coleman’s “aberrant behavior,” and given her “length of satisfactory employment,” the arbitrator concluded that a fitness-for-duty examination “would have been a more reasonable course for the Service to follow.” The arbitrator ordered Coleman reinstated, pending successful completion of a fitness-for-duty exam. The arbitrator declined to award back pay, however, because he could not determine when Coleman first became qualified to return to work. Coleman passed her fitness-for-duty exam and resumed her duties at the Postal Service facility on September 1, 2007, roughly two years after she was suspended. During this period, Coleman pursued her two EEOC charges against the Postal Service. An administrative law judge denied both complaints, and the EEOC rejected Coleman’s consolidated appeals on April 28, 2009. Coleman then filed this suit alleging that the Postal Service had discriminated against her on the basis of race, sex, and disability by placing her on off-duty status and terminating her, and had retaliated against her for reporting discrimination. Coleman also alleged that the Postal Service violated the Rehabilitation Act by failing to accommodate her disability. Following discovery, the district court granted the Postal Service’s motion for summary judgment in its entirety. Its judgment on Coleman’s discrimination and retaliation claims rested on three grounds: First, the court held that Coleman had failed to establish a prima facie case under *845 the McDonnell Douglas “indirect” method of proof because she had not identified any similarly situated employees outside of her protected classes who were treated more favorably. Second, the district court determined that Coleman had offered no evidence of pretext. Third, the district court held that Coleman had not presented sufficient direct or circumstantial evidence of discriminatory or retaliatory animus under Title VII’s “direct” method of proof. Coleman appeals from summary judgment on her Title VII claims of race and sex discrimination and retaliation. She does not seek review of summary judgment on her disability claims. We consider first the race and sex discrimination claims, and then the retaliation claims. II. Discrimination Claims Title VII makes it unlawful for an employer to discharge or discipline an employee because of that person’s race or sex, among other grounds. 42 U.S.C. § 2000e. In a disparate treatment case such as this one, a plaintiff may prove discrimination either directly or indirectly. See Silverman v. Board of Educ. of the City of Chicago, 637 F.3d 729, 733 (7th Cir.2011). Under the “direct method,” the plaintiff may avoid summary judgment by presenting sufficient evidence, either direct or circumstantial, that the employer’s discriminatory animus motivated an adverse employment action. Of course, “smoking gun” evidence of discriminatory intent is hard to come by. See United States Postal Service Board of Governors v. Aikens, 460 U.S. 711, 716, 103 S.Ct. 1478, 75 L.Ed.2d 403 (1983) (“There will seldom be ‘eyewitness’ testimony as to the employer’s mental processes.”). So in a line of cases beginning with McDonnell Douglas, the Supreme Court developed a burden-shifting framework known as the “indirect method” of proof, designed to “sharpen the inquiry into the elusive factual question of intentional discrimination.” Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 255, n. 8, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981). Coleman has attempted to establish discrimination through both the direct and indirect methods of proof. Because she presented sufficient evidence to survive summary judgment under the indirect method, there is no need to evaluate her discrimination claims under the direct method. A. The McDonnell Douglas Framework Under the indirect method, the plaintiff carries “the initial burden under the statute of establishing a prima facie case of ... discrimination.” McDonnell Douglas, 411 U.S. at 802, 93 S.Ct. 1817. To establish a prima facie case of discrimination a plaintiff must offer evidence that: “(1) she is a member of a protected class, (2) her job performance met [the employer’s] legitimate expectations, (3) she suffered an adverse employment action, and (4) another similarly situated individual who was not in the protected class was treated more favorably than the plaintiff.” Burks v. Wisconsin Dep’t of Transportation, 464 F.3d 744, 750-51 (7th Cir.2006). Once a prima facie case is established, a presumption of discrimination is triggered. “The burden then must shift to the employer to articulate some legitimate, nondiscriminatory reason” for its action. McDonnell Douglas, 411 U.S. at 802, 93 S.Ct. 1817; see Burks, 464 F.3d at 751. When the employer does so, the burden shifts back to the plaintiff, who must present evidence that the stated reason is a “pretext,” which in turn permits an inference of unlawful discrimination. McDonnell Douglas, 411 U.S. at 804, 93 S.Ct. 1817; see Burks, 464 F.3d at 751. *846 The Postal Service concedes for purposes of summary judgment that Coleman has satisfied the first three elements of her prima facie case: (1) she is a member of two protected classes (race and sex); (2) her job performance was satisfactory; and (3) the Postal Service subjected her to two adverse employment actions (placement on emergency off-duty status and then termination). The Postal Service disputes the fourth element, arguing that the white, male co-workers Coleman identified as receiving more favorable treatment were not similarly situated as a matter of law. The Postal Service has also offered a non-discriminatory reason for terminating Coleman — it claims she violated its code of conduct — but Coleman contends that this reason is pretextual. B. Similarly Situated Co-workers The similarly-situated analysis calls for a “flexible, common-sense” examination of all relevant factors. Henry v. Jones, 507 F.3d 558, 564 (7th Cir.2007). “All things being equal, if an employer takes an action against one employee in a protected class but not another outside that class, one can infer discrimination. The ‘similarly situated’ prong establishes whether all things are in fact equal.” Filar v. Board of Educ. of City of Chicago, 526 F.3d 1054, 1061 (7th Cir.2008) (internal citation omitted). Its purpose is to eliminate other possible explanatory variables, “such as differing roles, performance histories, or decision-making personnel, which helps isolate the critical independent variable”— discriminatory animus. Humphries, 474 F.3d at 405. Similarly situated employees “must be ‘directly comparable’ to the plaintiff ‘in all material respects,’ ” but they need not be identical in every conceivable way. Patterson v. Indiana Newspapers, Inc., 589 F.3d 357, 365-66 (7th Cir.2009), quoting Raymond v. Ameritech Corp., 442 F.3d 600, 610-11 (7th Cir.2006). We are looking for comparators, not “clone[s].” Chaney v. Plainfield Healthcare Center, 612 F.3d 908, 916 (7th Cir.2010). So long as the distinctions between the plaintiff and the proposed comparators are not “so significant that they render the comparison effectively useless,” the similarly-situated requirement is satisfied. Humphries, 474 F.3d at 405; see also Crawford v. Indiana Harbor Belt R.R. Co., 461 F.3d 844, 846 (7th Cir.2006) (the question is whether “members of the comparison group are sufficiently comparable to [the plaintiff] to suggest that [the plaintiff] was singled out for worse treatment”). This flexible standard reflects the Supreme Court’s approach to Title VII in McDonnell Douglas and its progeny. To offer a prima facie case of discrimination under the indirect method, the plaintiffs burden is “not onerous.” Burdine, 450 U.S. at 253, 101 S.Ct. 1089. The Supreme Court “never intended” the requirements “to be rigid, mechanized, or ritualistic ... [but] merely a sensible, orderly way to evaluate the evidence in light of common experience as it bears on the critical question of discrimination.” Furnco Construction Corp. v. Waters, 438 U.S. 567, 577, 98 S.Ct. 2943, 57 L.Ed.2d 957 (1978). The Court has cautioned that “precise equivalence ... between employees is not the ultimate question.” McDonald v. Santa Fe Trail Transportation Co., 427 U.S. 273, 283 n. 11, 96 S.Ct. 2574, 49 L.Ed.2d 493 (1976). The touchstone of the similarly-situated inquiry is simply whether the employees are “comparable.” Id., quoting McDonnell Douglas, 411 U.S. at 804, 93 S.Ct. 1817. Whether a comparator is similarly situated is “usually a question for the fact-finder,” and summary judgment is appropriate only when “no reasonable fact- *847 finder could find that plaintiffs have met their burden on the issue.” Srail v. Village of Lisle, 588 F.3d 940, 945 (7th Cir.2009). There must be “enough common factors ... to allow for a meaningful comparison in order to divine whether intentional discrimination was at play.” Bar-ricks v. Eli Lilly and Co., 481 F.3d 556, 560 (7th Cir.2007). The “number [of relevant factors] depends on the context of the case.” Radue v. Kimberly-Clark Corp., 219 F.3d 612, 617 (7th Cir.2000). In the usual case a plaintiff must at least show that the comparators (1) “dealt with the same supervisor,” (2) “were subject to the same standards,” and (3) “engaged in similar conduct without such differentiating or mitigating circumstances as would distinguish their conduct or the employer’s treatment of them.” Gates v. Caterpillar, Inc., 513 F.3d 680, 690 (7th Cir.2008), quoting Snipes v. Illinois Dep’t of Corrections, 291 F.3d 460, 463 (7th Cir.2002). This is not a “magic formula,” however, and the similarly-situated inquiry should not devolve into a mechanical, “one-to-one mapping between employees.” Humphries, 474 F.3d at 405. With this legal standard in mind, we turn to Coleman’s proposed comparators. According to Coleman’s evidence, two white male employees, Frank Arient and Robert Pelletier, “held a knife to the throat of a black male co-worker” “while holding down his legs.” 1 Arient’s and Pelletier’s direct supervisor, Brian Turkovich, learned of the incident a few days later and conducted an investigation. Von Rhein, who supervised all three men, participated in the investigation, personally interviewing the two attackers and several witnesses. Von Rhein and Turkovich concluded that the incident was just “horseplay,” and Von Rhein suspended Arient and Pelletier without pay for fourteen days. Von Rhein and Turkovich later reduced these suspensions to seven days after objections by the union. According to Von Rhein, Sove approved their suspensions. Von Rhein Dep. 175. In his own deposition, Sove described Arient’s and Pelletier’s actions as “some stupid prank that they were playing with each other.” Sove Dep. 121. The district court concluded that Arient and Pelletier could not serve as comparators because they “reported to a different supervisor” and “held a substantially different job than Coleman.” Although the court acknowledged there was “at least some similarity in terms of the seriousness of the incident,” it was “not enough” to overcome the other dissimilarities. We think that this analysis focused too much on minor differences and was too demanding for purposes of summary judgment. 1. Same Supervisor The similarly-situated requirement “normally entails” the existence of a common supervisor. Badue, 219 F.3d at 617. When the same supervisor treats an otherwise equivalent employee better, one can often reasonably infer that an unlawful animus was at play. The inference of discrimination is weaker when there are different decision-makers, since they “may rely on different factors when deciding whether, and how severely, to discipline an employee.” Ellis v. United Parcel Service, 523 F.3d 823, 826 (7th Cir.2008); see also Little v. Illinois Dep’t of Revenue, 369 F.3d 1007, 1012 (7th Cir.2004) (discipline from a different supervisor “sheds no *848 light” on the disciplinary decision). For this reason, this court generally requires a plaintiff to demonstrate at a minimum that a comparator was treated more favorably by the same decision-maker who fired the plaintiff. See Ellis, 523 F.3d at 826. In this case, there was a common decision-maker for Coleman, Arient, and Pelletier: the facility’s maintenance operations manager, Charles Von Rhein. Von Rhein approved Coleman’s termination and the men’s suspensions. The district court relied on the fact that Arient’s and Pelletier’s direct supervisor (Turkovich) was not the same as Coleman’s (Berry). But this misses the point of the common supervisor factor. While we have sometimes phrased the question ambiguously as whether the comparators “dealt with the same supervisor,” e.g., Gates, 513 F.3d at 690 (emphasis added), the real question is whether they were “treated more favorably by the same decisionmaker.” Ellis, 523 F.3d at 826 (emphasis added); see Little, 369 F.3d at 1012 (“A similarly-situated employee must have been disciplined, or not, by the same decisionmaker who imposed an adverse employment action on the plaintiff.”). This point follows logically from the cause of action itself, which requires proof “that the decisionmaker has acted for a prohibited reason.” Schandelmeier-Bartels v. Chicago Park Dist., 634 F.3d 372, 379 (7th Cir.2011), quoting Rogers v. City of Chicago, 320 F.3d 748, 754 (7th Cir.2003) (emphasis in original). Under Title VII, a “decisionmaker is the person ‘responsible for the contested decision.’ ” Id., quoting Rogers, 320 F.3d at 754. For both Coleman’s termination and Arient’s and Pelletier’s suspensions, that person was Von Rhein. He signed the letters placing Coleman on off-duty status and terminating her, and he conducted the internal investigation of her in the interim period. Von Rhein also personally investigated Arient’s and Pelletier’s actions and testified that he made the decision to suspend them. The district court downplayed Von Rhein’s supervisory role in the response to the knife incident, asserting he merely “sign[ed] off on Turkovich’s decision” to suspend them. But, again, the issue is not only who proposed the suspension but who was “responsible” for the decision. Schandelmeier-Bartels, 634 F.3d at 379, quoting Rogers, 320 F.3d at 754. Only Von Rhein, and not Turkovich, had the authority to discipline Arient and Pelletier. For purposes of Title VII, he was the decision-maker. 2. Same Standards of Conduct The Postal Service contends that because Arient and Pelletier had different job titles and duties, they cannot be considered situated similarly to Coleman. That is not correct. In the context of this case of differential discipline, it is irrelevant to the comparison that Arient and Pelletier are maintenance mechanics and Coleman is a maintenance support clerk. We have repeatedly made clear that a “difference in job title alone is not dispositive.” Boumehdi v. Plastag Holdings, LLC, 489 F.3d 781, 791 (7th Cir.2007); see Rodgers v. White, 657 F.3d 511, 518 (7th Cir.2011) (“Formal job titles and rank are not dispositive.... ”). The question is not whether the employer classified the comparators in the same way, “but whether the employer subjected them to different employment policies.” Lathem v. Dep’t of Children & Youth Services, 172 F.3d 786, 793 (11th Cir.1999). Comparators need only be similar enough to enable “a meaningful comparison.” Humphries, 474 F.3d at 405. Arient and Pelletier worked at the same job site as Coleman, were subject to the same standards of conduct, violated the same rule, *849 and were disciplined by the same supervisor. Their different titles and duties do not defeat, as a matter of law, the probative value of their different disciplinary treatment. The application of this “same standards” factor also depends on the specific facts of the case. In cases involving the quality of job performance, for example, a would-be comparator’s professional role may be so different from the plaintiffs as to “render the comparison effectively useless.” Humphries, 474 F.3d at 405; accord, e.g., Senske v. Sybase, Inc., 588 F.3d 501, 510 (7th Cir.2009) (salesmen with “lower-ranking sales positions” were not similarly situated to the plaintiff, who was fired for performance reasons); Burks, 464 F.3d at 751 (a receptionist and a supervisor were not similarly situated to the plaintiff, a program manager who was fired for performance reasons); Keri v. Board of Trustees of Purdue University, 458 F.3d 620, 626 (7th Cir.2006) (tenured university professors were not similarly situated to untenured plaintiff professor who was not reappointed after “widespread complaints from both students and supervisors”). Where the issue is the quality of a plaintiffs work, a difference between the plaintiffs and comparators’ positions can be important because this difference will often by itself account for the less favorable treatment of the plaintiff. Cf. Senske, 588 F.3d at 510 (“the comparators must be similar enough that differences in their treatment cannot be explained by other variables, such as distinctions in their roles or performance histories”). In contrast, Arient’s and Pelletier’s different positions provide no such self-evident explanation for their more lenient punishment. The reason is obvious. Coleman and her comparators were disciplined not for bad performance but for violating a general workplace rule that applied to employees in all departments and of all ranks. In such misconduct cases (as opposed to performance cases), comparisons between employees with different positions are more likely to be useful. See, e.g., Rodgers, 657 F.3d at 513 (where plaintiff was punished more harshly than his supervisor for the same misconduct, the “general rule” that “employees of differing ranks usually make poor comparators ... does not apply”). “[W]hen uneven discipline is the basis for a claim of discrimination, the most-relevant similarities are those between the employees’ alleged misconduct, performance standards, and disciplining supervisor,” rather than job description and duties. Id. at 518. The real issue is whether Arient and Pelletier were subject to the same standards of conduct as Coleman, and of course they were. The Postal Service rules against workplace violence and threats apply equally to mechanics and clerks. The employee handbook frames the prohibition in all-encompassing terms: “it is the unequivocal policy of the Postal Service that there must be no tolerance of violence or threats of violence by anyone at any level of the Postal Service. Similarly, there must be no tolerance of harassment, intimidation, threats, or bullying by anyone at any level.” (Emphases added.) Since the purpose of the rule is to ensure a “safe and humane working environment,” there is no objective reason for it to apply with greater or lesser force to employees of certain positions. Even if there might have been some theoretical basis for enforcing the rule differently based on job position, there is no evidence that the Postal Service actually took Arient’s, Pelletier’s, or Coleman’s roles into account when it disciplined them. A proposed comparator’s position or rank may be important, but only *850 “provided that the employer took these factors into account when making the personnel decision in question.” Eaton v. Indiana Dep’t of Corrections, 657 F.3d 551, 559 (7th Cir.2011) (emphasis in original), quoting Patterson v. Avery Dennison Corp., 281 F.3d 676, 680 (7th Cir.2002); see also Peirick v. Indiana University-Purdue University Indianapolis, 510 F.3d 681, 689 (7th Cir.2007) (“we doubt that the [employer] took heed of employee classifications when doling out sanctions”). “A characteristic that distinguishes two employees, regardless of its significance when objectively considered, does not render the employees non-comparable if the employer never considered that characteristic ... [because it] cannot provide any insight as to whether the employer’s decision as motivated by discriminatory intent.” Eaton, 657 F.3d at 559. Here, the record provides no indication that the Postal Service considered job titles at all significant when deciding on discipline for Arient, Pelletier, and Coleman. There are a number of potential explanations for why Arient and Pelletier got off with such lighter punishments than Coleman. Perhaps it was because managers honestly perceived them as less culpable or dangerous. Perhaps it was because they were white or male. But it was surely not because they were mechanics. 3. Conduct of Comparable Seriousness In a disparate discipline case, the similarly-situated inquiry often hinges on whether co-workers “engaged in comparable rule or policy violations” and received more lenient discipline. Naik v. Boehring er Ingelheim Pharms., Inc., 627 F.3d 596, 600 (7th Cir.2010), quoting Patterson, 589 F.3d at 365-66. The Supreme Court has made clear that “precise equivalence in culpability between employees is not the ultimate question: as we indicated in McDonnell Douglas, an allegation that other ‘employees involved in acts against [the employer] of comparable seriousness ’ ” received more favorable treatment “is adequate to plead an inferential case” of discrimination. McDonald, 427 U.S. at 283 n. 11, 96 S.Ct. 2574, quoting McDonnell Douglas, 411 U.S. at 804, 93 S.Ct. 1817. Following this language, our circuit, like many others, has adopted this “comparable seriousness” standard. E.g., Peirick, 510 F.3d at 689; Davis v. Wisconsin Dep’t of Corrections, 445 F.3d 971, 978 (7th Cir.2006); Johnson v. Artim Transportation System, Inc., 826 F.2d 538, 543 (7th Cir.1987). 2 Comparators must have “engaged in similar — not identical — conduct to qualify as similarly situated.” Peirick, 510 F.3d at 691, 689 (reversing summary judgment in relevant part; university tennis coach “accused of using abusive language, unsafe driving, leaving students behind during a road trip, and pitting the students against the administration” was similarly situated to coaches who “did not engage in the exact same misconduct” but who “violated the very same rules”), quoting Ezell v. Potter, 400 F.3d 1041, 1050 (7th Cir.2005) (reversing summary judgment in relevant part; mail carrier accused of taking too long a lunch was similarly situated to another carrier who had lost a piece of *851 certified mail). To determine “whether two employees have engaged in similar misconduct, the critical question is whether they have engaged in conduct of comparable seriousness.” Peirick, 510 F.3d at 689. Again, the analysis is straightforward here. Arient and Pelletier violated the Postal Service rule that prohibits “Violent and/or Threatening Behavior” — the same rule Coleman was accused of breaking. That they did not break the rule in precisely the same manner does not mean that summary judgment was appropriate. By directly threatening another employee with a knife in the workplace, Arient and Pelletier engaged in conduct that appears, at least for purposes of summary judgment, at least as serious as Coleman’s indirect “threat” against Berry — and arguably even more so. 3 Where a proposed comparator violated the same rule as the plaintiff in an equivalent or more serious manner, courts should not demand strict factual parallels. See Lynn v. Deaconess Medical Center-West Campus, 160 F.3d 484, 488 (8th Cir.1998) (“To require that employees always have to engage in the exact same offense as a prerequisite for finding them similarly situated would result in a scenario where evidence of favorable treatment of an employee who has committed a different but more serious, perhaps even criminal offense, could never be relevant to prove discrimination.”), abrogated on other grounds, Torgerson v. City of Rochester, 643 F.3d 1031 (8th Cir.2011) (en banc). The Postal Service argues that Arient and Pelletier are not appropriate comparators because the Postal Service viewed their behavior “as an ‘isolated instance’ where ‘no particular threats were involved,’ ” while Coleman had made a “credible threat.” The Postal Service may make that argument at trial, but it is not a winner on summary judgment. When two grown men hold a person down while brandishing a knife (whether at his throat or not), not only is a “particular threat[ ] ... involved” — a jury could reasonably conclude that it was a far more immediate one than an employee confiding to her psychiatrist in a private therapy session that she was having thoughts about killing her boss. To be sure, the Postal Service is right to take seriously all threats made by, and against, its employees. But at the summary judgment stage, the employer cannot defeat a plaintiffs prima facie case of discrimination on the theory that it applied its “no tolerance” policy on threats to some workers while dismissing dangerous acts of others as mere “horseplay.” See Gordon v. United Airlines, Inc., 246 F.3d 878, 891 (7th Cir.2001) (“It is not the province of this court to question an employer’s decision to punish some conduct more harshly than other conduct. Nevertheless, we are not bound by the labels that an employer uses and must scrutinize the conduct behind those labels to determine if they are applied to similar conduct.”). Such fact issues are the province of the jury. 4 We have noted with some concern the tendency of judges in employment dis *852 crimination cases “to require closer and closer comparability between the plaintiff and the members of the comparison group.” Crawford, 461 F.3d at 846. 5 The purpose of the similarly-situated requirement is to “provide plaintiffs the ‘boost’ that the McDonnell Douglas framework intended.” Humphries, 474 F.3d at 406, citing Stone v. City of Indianapolis Public Utilities Div., 281 F.3d 640, 643 (7th Cir. 2002). Demanding nearly identical comparators can transform this evidentiary “boost” into an insurmountable hurdle. Coleman’s proposed comparators (1) “dealt with the same supervisor,” (2) “were subject to the same standards,” and (3) “engaged in similar conduct” of comparable seriousness. Gates, 513 F.3d at 690. They are similar enough to permit a reasonable inference of discrimination, and that is all McDonnell Douglas requires. C. Pretext The Postal Service has offered a legitimate, nondiscriminatory reason for terminating Coleman — it claims she “posed a threat to kill her supervisor.” To show this reason is pretextual, Coleman “must present evidence suggesting that the employer is dissembling.” O’Leary, 657 F.3d at 635. “The question is not whether the employer’s stated reason was inaccurate or unfair, but whether the employer honestly believed the reasons it has offered to explain the discharge.” Id. “It is not the court’s concern that an employer may be wrong about its employee’s performance, or may be too hard on its employee. Rather, the only question is whether the employer’s proffered reason was pretextual, meaning that it was a lie.” Naik, 627 F.3d at 601, quoting Ineichen v. Ameritech, 410 F.3d 956, 961 (7th Cir.2005). To meet this burden, Coleman must “identify such weaknesses, implausibilities, inconsistencies, or contradictions” in the Postal Service’s asserted reason “that a reasonable person could find [it] unworthy of credence.” Boumehdi, 489 *853 F.3d at 792. If the Postal Service terminated Coleman because it “honestly believed” she posed a threat to other employees — even if this reason was “foolish, trivial, or baseless” — Coleman loses. Id. On the other hand, “if the stated reason, even if actually present to the mind of the employer, wasn’t what induced him to take the challenged employment action, it was a pretext.” Forrester v. Rauland-Borg Corp., 453 F.3d 416, 418 (7th Cir.2006). To show pretext, Coleman argues that the labor arbitrator who ordered her reinstated found that the Postal Service did not honestly believe she was a threat, and that the district court should have given his decision preclusive effect. We disagree on both points. On the merits, however, we agree that Coleman has presented enough evidence of pretext to avoid summary judgment. First, like the arbitrator, we question whether Coleman’s statements about Berry rose to the level of a “true threat,” and thus whether Coleman can fairly be said to have violated any workplace rule at all. Second, a number of background facts cast doubt on the assertion that Coleman was dangerous: her statements came in a private therapy session, the Postal Service learned of them the same day the psychiatrist discharged Coleman as stable, and it had options short of termination available to gauge her propensity for violence. Third, Coleman’s comparator evidence tends to show that her Postal Service managers did not enforce this rule evenhandedly. This evidence of similarly situated co-workers is also relevant to the pretext inquiry. It suggests that the Postal Service decision-makers here did not take the rule against threats as seriously as they claimed. As the Supreme Court, this court, and other circuits have held, a discrimination plaintiff may employ such comparator evidence to discharge her burden at the pretext stage as well as to satisfy the fourth element of her prima facie case. Based on this evidence here, a reasonable jury could conclude that the Postal Service’s stated reason for firing Coleman was pretextual. 1. The Effect of the Arbitration The arbitration does not support issue preclusion on the issue of pretext for two independent reasons. First, the arbitrator did not decide the same issue of pretext. Second, Coleman’s case is subject to the general rule under Title VII that arbitration decisions do not bind either side regarding statutory discrimination claims. We consider first just what the arbitrator decided. Issue preclusion requires an identity of issues. Issue preclusion, also known as collateral estoppel, “bars ‘successive litigation of an issue of fact or law actually litigated and resolved in a valid court determination essential to the prior judgment,’ even if the issue recurs in the context of a different claim.” Taylor v. Sturgell, 553 U.S. 880, 892, 128 S.Ct. 2161, 171 L.Ed.2d 155 (2008), quoting New Hampshire v. Maine, 532 U.S. 742, 748, 121 S.Ct. 1808, 149 L.Ed.2d 968 (2001). In some cases, administrative adjudications may have preclusive effect. See, e.g., University of Tennessee v. Elliott, 478 U.S. 788, 799, 106 S.Ct. 3220, 92 L.Ed.2d 635 (1986) (“when a state agency acting in a judicial capacity resolves disputed issues of fact properly before it which the parties had an adequate opportunity to litigate, federal courts must give the agency’s factfinding the same preclusive effect to which it would be entitled in the State’s courts”) (internal citation and quotation marks omitted). Whatever the original forum, however, the doctrine “applies only when (among other things) the same issue is involved in the two proceedings and the determination of that ques *854 tion is ‘essential’ to the prior judgment.” King v. Burlington Northern & Santa Fe Ry. Co., 538 F.3d 814, 818 (7th Cir.2008). Here the arbitrator did not examine whether the Postal Service honestly believed Coleman was a danger, but only whether Coleman really was a danger. Finding that Coleman’s statements to her psychiatrist did not constitute a “true threat,” the arbitrator ruled that the Postal Service lacked just cause to terminate her. This finding is not the same as a finding that the Postal Service decision-makers were lying about their motives. The most that Coleman can say is that the arbitrator was skeptical that Berry genuinely feared Coleman, suspecting he had “embellished” his story. Even if Berry exaggerated his reaction to news of the threat, that would not prove that Von Rhein and Sove, the supervisors who decided to terminate Coleman, were also disingenuous. The arbitrator acknowledged that Coleman’s behavior raised “serious concerns about her fitness for duty, and under what conditions she might be able to work,” and he ordered Coleman to undergo a psychiatric examination to ascertain whether she was ready to return. He did not determine that the Postal Service’s concerns about Coleman were lies, but only that it had failed to meet its “burden of proving that [she] engaged in conduct warranting her removal.” Issue preclusion therefore could not apply. Second, whatever the arbitrator’s findings, his decision could not trigger collateral estoppel in this action. In Alexander v. Gardner-Denver Co., 415 U.S. 36, 59-60, 94 S.Ct. 1011, 39 L.Ed.2d 147 (1974), the Supreme Court held that arbitration decisions do not have preclusive effect in later litigation under Title VII. The Court explained that “Congress intended federal courts to exercise final responsibility for enforcement of Title VII; deferral to arbitral decisions would be inconsistent with that goal.” Id. at 56, 94 S.Ct. 1011. The only exception to this rule is where a clause in a collective bargaining agreement has explicitly mandated that “employment-related discrimination claims would be resolved in arbitration.” 14 Penn Plaza v. Pyett, 556 U.S. 247, 129 S.Ct. 1456, 1464, 173 L.Ed.2d 398 (2009). In 14 Penn Plaza, the Supreme Court held that such clauses are enforceable, distinguishing Gardner-Denver on the grounds that, in that case, the “employee’s collective-bargaining agreement did not mandate arbitration of statutory antidiscrimination claims.” Id. at 1467. Yet the Court recognized the continuing vitality of Gardner-Denver in cases like this one, where the CBA did not “clearly and unmistakably require[] union members to arbitrate claims arising under” federal anti-discrimination laws: where the “collective-bargaining agreement [gives] the arbitrator ‘authority to resolve only questions of contractual rights,’ his decision could not prevent the employee from bringing the Title VII claim in federal court ‘regardless of whether certain contractual rights are similar to, or duplicative of, the substantive rights secured by Title VII.’ ” Id. at 1461, 1467, quoting Gardner-Denver, 415 U.S. at 53-54, 94 S.Ct. 1011. Here, the collective bargaining agreement did not require submission of Title VII claims to labor arbitration. Under Gardner-Denver, then, even if the arbitrator had reached the pretext issue, his findings would not have preclusive effect here. 2. Evidence of Pretext Without giving preclusive effect to the arbitral decision, however, we find that Coleman has offered evidence of pretext in the form of context. “[A]n evaluation of context is essential to determine whether an employer’s explanation is fishy *855 enough to support an inference that the real reason must be discriminatory.” Loudermilk v. Best Pallet Co., 636 F.3d 312, 315 (7th Cir.2011). Much of Coleman’s context evidence is recounted in the arbitrator’s findings. She is not barred from relying on the these findings as evidence that the Postal Service’s stated reason for terminating her was a pretext. As the Gardner-Denver Court stated: “The arbitral decision may be admitted as evidence and accorded such weight as the court deems appropriate.” 415 U.S. at 60, 94 S.Ct. 1011. In this case, several of the arbitrator’s findings provide support for Coleman’s argument that the Postal Service’s purported reasons for terminating her were pretextual. First, the arbitrator concluded that Coleman’s statements to Dr. Ionescu did not constitute a “true threat.” We think this is a reasonable inference that is tantamount to a finding that Coleman did not actually violate the Postal Service’s rule against threats of violence. Granted, even if Coleman broke no rule, the Postal Service may still have mistakenly believed she did — and that’s what counts in the pretext analysis. See Forrester, 453 F.3d at 418. Nevertheless, the Postal Service can be presumed to understand its own code of conduct. The incongruity between Coleman’s non-violation and her termination casts at least some doubt on the Postal Service’s motives. See, e.g., Loudermilk, 636 F.3d at 315 (“The Civil Rights Act of 1964 does not require employers to have ‘just cause’ for sacking a worker, but an employer who advances a fishy reason takes the risk that disbelief of the reason will support an inference that it is a pretext for discrimination.”). And there is inherent “fishiness” in an employer’s proffered reason when it rests on a policy that does not legitimately apply to the employee who was terminated. See, e.g., Gordon, 246 F.3d at 889 (“Here, an employer applied a rarely used label to sanction conduct that does not clearly fall within the chosen category.... [W]hen considered together with the inconsistency [in the employer’s definition of the rule], it is sufficient evidence of pretext and, therefore, precludes summary judgment.”); Stalter v. Wal-Mart Stores, Inc., 195 F.3d 285, 290 (7th Cir.1999) (reversing summary judgment for employer where plaintiff had been fired supposedly for theft; eating a few corn chips from an open bag in a break room did not “fit within a reasonable understanding of the term ‘theft’ ” and a “jury could certainly infer ... that [the employer’s] claim of theft was a pretext for [the plaintiffs] termination”). As the arbitrator also identified, there are real questions as to whether the Postal Service could have honestly considered Coleman dangerous. For one, he emphasized that Coleman made her statements in a private, confidential therapy session: [W]e have an employee who, after determining she could not deal with the stress and frustration of being unable to work following her surgery, voluntarily admits herself for psychiatric treatment. During this treatment, her psychiatrist probes the depth of [Coleman’s] anger and finds that she is experiencing suicidal and homicidal ideations. The special context in which Coleman expressed her anger cannot possibly have been lost on the Postal Service. The psychotherapeutic environment is one in which such extreme feelings would understandably arise — and indeed, the one in which they should be most encouraged. As the Supreme Court has noted: “Effective psychotherapy ... depends upon an atmosphere of confidence and trust in which the patient is willing to make a frank and complete disclosure of facts, emotions, memories, and fears.” Jaffee v. *856 Redmond, 518 U.S. 1, 10, 116 S.Ct. 1923, 135 L.Ed.2d 337 (1996); see also Tarasoff v. Regents of University of California, 17 Cal.3d 425, 131 Cal.Rptr. 14, 551 P.2d 334, 347 (1976) (“We realize that the open and confidential character of psychotherapeutic dialogue encourages patients to express threats of violence, few of which are ever executed. Certainly a therapist should not be encouraged routinely to reveal such threats; such disclosures could seriously disrupt the patient’s relationship with his therapist and with the persons threatened.”). It would therefore be troubling to think that anyone who confides to her psychiatrist that she has fantasized about killing her boss could automatically be subject to termination for cause. To be sure, the situation changes when a patient expresses a genuine and ongoing intent to harm another person. That was the allegation in the canonical Tarasoff case. See 131 Cal. Rptr. 14, 551 P.2d at 341 (“Poddar informed Moore, his therapist, that he was going to kill an unnamed girl, readily identifiable as Tatiana, when she returned home from spending the summer in Brazil.”). In this case, however, the Postal Service had little reason to believe that Coleman posed a continuing threat — and even more to the point, it appears to have made no effort to ascertain whether she did or not. On the contrary, the Postal Service had good reason to believe that whatever danger Coleman ever posed had subsided by the time she sought to return to work, well after she expressed this thought to her therapist. Dr. Ionescu informed Berry of Coleman’s statements the very same day she discharged Coleman in “stable” condition, describing her as a “cooperative, pleasant,” “reactive,” “smiling,” “model patient.” 6 As the arbitrator noted: It is obvious that any homicidal ideation [Coleman] may have had toward Mr. Berry was part and parcel of her psychiatric condition for which she sought treatment. At the time she expressed this ideation, she was hospitalized and, therefore, incapable of acting upon it. She was not released from the hospital until it had abated. On summary judgment, Coleman is entitled to the reasonable inferences (a) that Dr. Ionescu would not have released her from treatment if she believed Coleman posed a danger to herself or others, and (b) that supervisors considering the matter should and would have realized as much before firing her. Finally, if the Postal Service’s real concern was Coleman’s potential danger, why did it not simply order her to undergo a psychological evaluation? As the arbitrator noted, “Both Mr. Berry and Mr. Von Rhein ... acknowledged that they could *857 have referred [Coleman] for a fitness-for-duty examination.” He concluded that, “[u]nder the unique circumstances attendant to this case, that would have been a more reasonable course for the Service to follow. With her length of satisfactory employment with the Postal Service, she deserved as much.” The Postal Service’s failure to take this seemingly natural step is further evidence suggesting that Coleman’s mental stability was not its real motivation for firing her. In short, while the arbitral decision is not binding, its factual predicates and analysis give some boost to Coleman’s claim that the Postal Service’s asserted reasons for terminating her were pretextual. 3. Comparator Evidence to Show Pretext Coleman has also presented additional evidence of pretext: her evidence that similarly situated employees outside her protected classes received more favorable treatment from the same decision-maker. As detailed above, Arient and Pelletier broke the same rule that Coleman allegedly did and did so, a jury could reasonably conclude, in a much more egregious manner. Such evidence of selective enforcement of a rule “calls into question the veracity of the employer’s explanation.” Olsen v. Marshall & Ilsley Corp., 267 F.3d 597, 601 (7th Cir.2001); accord, e.g., Belli Santi v. CNA Ins. Cos., 88 F.3d 192, 202 (3d Cir.1996) (The plaintiffs “showing that the company did not enforce such a policy” is evidence from which the “jury ... could rationally conclude that the legitimate non-retaliatory reason offered by [the employer] was a pretext for discharging [the plaintiff].”); Williams v. City of Valdosta, 689 F.2d 964, 975 (11th Cir.1982) (“It is undisputed, however, that the City’s adherence to its formal promotional policy was inconsistent and arbitrary at best. This inconsistency supports the conclusion that resort to the examination requirement was a pretext for singling out Williams for unfavorable treatment.”). Combined with the additional circumstances discussed above, Coleman’s evidence is sufficient to defeat summary judgment on the pretext issue. The Supreme Court holds that comparator evidence is relevant at the pretext stage. In McDonnell Douglas itself, the Supreme Court taught that “evidence that white employees involved in acts ... of comparable seriousness” received more favorable treatment would be “[especially relevant” to a showing that the employer’s “stated reason for [the plaintiffs] rejection was in fact pretext.” 411 U.S. at 804, 93 S.Ct. 1817. In Burdine, too, the Court made clear that in the pretext inquiry, “it is the plaintiffs task to demonstrate that similarly situated employees were not treated equally.” 450 U.S. at 258, 101 S.Ct. 1089, citing McDonnell Douglas, 411 U.S. at 804, 93 S.Ct. 1817. And in a closely related context, the Supreme Court has affirmed the value of qualifications evidence (that is, evidence that the employer hired a less qualified person outside the plaintiffs protected class) in the pretext inquiry. Ash v. Tyson Foods, Inc., 546 U.S. 454, 457, 126 S.Ct. 1195, 163 L.Ed.2d 1053 (2006) (“qualifications evidence may suffice, at least in some circumstances, to show pretext”); Patterson v. McLean Credit Union, 491 U.S. 164, 187-88, 109 S.Ct. 2363, 105 L.Ed.2d 132 (1989) (plaintiff “might seek to demonstrate that [the employee’s claim to have promoted a better qualified applicant was pretextual by showing that she was in fact better qualified than the person chosen for the position”), superseded on other grounds by 42 U.S.C. § 1981(b). Our precedents also teach that the similarly-situated inquiry and the pretext *858 inquiry are not hermetically sealed off from one another. We have often noted that “the prima facie case and pretext analyses often overlap.” Scruggs v. Garst Seed Co., 587 F.3d 832, 838 (7th Cir.2009); accord, Adelman-Reyes v. St. Xavier University, 500 F.3d 662, 665, (7th Cir.2007); Olsen, 267 F.3d at 600. Where the plaintiff argues that an employer’s discipline is meted out in an uneven manner, the similarly-situated inquiry dovetails with the pretext question. Evidence that the employer selectively enforced a company policy against one gender but not the other would go to both the fourth prong of the prima facie case and the pretext analysis. Thus, the “same inquiry into similarly situated employees has been made at the pretext stage.” Morrow v. Wal-Mart Stores, Inc., 152 F.3d 559, 561 (7th Cir.1998); accord, e.g., Buie v. Quad/Graphics, Inc., 366 F.3d 496, 508 (7th Cir.2004) (“The disparate treatment of similarly-situated employees who were involved in misconduct of comparable seriousness, but did not have a similar disability, could establish pretext.”); O’Regan v. Arbitration Forums, Inc., 246 F.3d 975, 985 (7th Cir.2001) (“to show pretext (as well as the fourth element of a prima facie case) the inquiry remains the same: the plaintiff must show that similarly situated employees were treated more favorably than the plaintiff’), citing Morrow, 152 F.3d at 561; Hiatt v. Rockwell Inti Corp., 26 F.3d 761, 770 (7th Cir.1994) (“In order to demonstrate pretext under the McDonnell Douglas analysis, a plaintiff may put forth evidence that (1) employees outside of the protected class ..., (2) who were involved in acts of comparable seriousness, (3) were nevertheless retained or rehired (while the plaintiff was not).”). A good example is Gordon v. United Airlines, where the airline fired an African American male flight attendant after he deviated from his flight schedule without authorization. 246 F.3d at 880. The district court granted summary judgment for United. We reversed: “Our review of the record reveals inconsistencies in definition and disparities in application [of the unauthorized deviation rule] that calls into question United’s proffered justification ....” Id. at 889. As evidence of pretext, the court pointed to Gordon’s showing that a similarly situated employee had been disciplined less harshly: “[T]he weakness of the proffered justification for the termination is further emphasized by the fact that the only other time that United has categorized an action as an unauthorized deviation, the involved employee, a white female, was not terminated.” Id. at 892. We explained: “A showing that similarly situated employees belonging to a different racial group received more favorable treatment can also serve as evidence that the employer’s proffered legitimate, nondiscriminatory reason for the adverse job action was a pretext for racial discrimination.” Id., quoting Graham v. Long Island R.R., 230 F.3d 34, 43 (2d Cir.2000). The reasoning and result in Gordon confirm what we have stated in many other cases: that comparator evidence can do “double-duty” at both the prima facie and pretext stages. Several other circuits agree. See, e.g., Hawn v. Executive Jet Management, Inc., 615 F.3d 1151, 1158 (9th Cir.2010) (“The concept of ‘similarly situated’ employees may be relevant to both the first and third steps of the McDonnell Douglas framework.”); Graham v. Long Island R.R., 230 F.3d 34, 43 (2d Cir.2000) (same); EEOC v. Horizon/CMS Healthcare Corp., 220 F.3d 1184, 1195 n. 6 (10th Cir.2000) (“while evidence that a defendant treated a plaintiff differently than similarly-situated employees is certainly sufficient to establish a prima facie case, it is ‘[especially relevant’ to show pretext if the defendant proffers a *859 legitimate, nondiscriminatory reason for the adverse employment action”); see also Rodgers v. U.S. Bank, N.A., 417 F.3d 845, 852-53 (8th Cir.2005) (finding comparator evidence relevant to both the prima facie and pretext phases, but imposing a more “rigorous” standard at the pretext stage), abrogated on other grounds, Torgerson, 643 F.3d at 1058; Simpson v. Kay Jewelers, Div. of Sterling, Inc., 142 F.3d 639, 646 (3d Cir.1998) (same). 7 In this case, Coleman has offered evidence sufficient to support a finding that Arient and Pelletier were situated similarly to her, are outside her protected classes, and received more lenient punishment for a comparably serious violation of the same rule. Together with the evidence identified by the arbitrator concerning the seriousness of the supposed threat and the Postal Service’s response to it, this evidence of selective enforcement was enough to create a genuine issue of fact as to whether the Postal Service’s asserted reason for terminating Coleman was pretextual. We must reverse summary judgment for the Postal Service on Coleman’s claims of sex and race discrimination. III. Retaliation Claims Coleman also appeals the district court’s grant of summary judgment to the Postal Service on her Title VII retaliation claims. Like discrimination, retaliation may be established by either the direct or indirect methods of proof. See Weber v. Universities Research Ass’n, 621 F.3d 589, 592 (7th Cir.2010). In the district court and in her appellate briefs, Coleman relied on both methods. In oral argument, however, Coleman’s counsel conceded that she lacked sufficient evidence to show a prima facie case of retaliation under the indirect method. We therefore consider Coleman’s retaliation claims under only the direct method of proof. To establish retaliation under the direct method, Coleman must show that; (1) she engaged in activity protected by Title VII; (2) the Postal Service took an adverse employment action against her; and (3) there was a causal connection between her protected activity and the adverse employment action. See Leitgen v. Franciscan Skemp Healthcare, Inc., 630 F.3d 668, 673 (7th Cir.2011). The first two elements are not disputed. Her formal EEOC charges were “the most obvious form of statutorily protected activity.” Silverman v. Board of Educ. of City of Chicago, 637 F.3d 729, 740 (7th Cir.2011); see 42 U.S.C. § 2000e-3(a). She also offered evidence that she had complained of race and sex discrimination to her supervisors as early as May 2005, and her requests for pre-complaint counseling before filing EEO charges also qualify as protect *860 ed activity. Coleman’s placement on unpaid off-duty status and termination were both adverse employment actions. The parties dispute only whether Coleman has evidence supporting an inference that her protected activity caused the Postal Service’s adverse actions. Coleman can show causation by showing that her complaints and EEO filings were a “substantial or motivating factor” in the Postal Service’s decisions to place her in off-duty status and/or to fire her. Gates, 513 F.3d at 686, quoting Culver v. Gorman & Co., 416 F.3d 540, 545 (7th Cir.2005). This may be done via direct evidence, which would “entail something akin to an admission by the employer (T’m firing you because you had the nerve to accuse me of sex discrimination!’).” O’Leary, 657 F.3d at 630. It may also be done by presenting a “ ‘convincing mosaic’ of circumstantial evidence” that would permit the same inference without the employer’s admission. Rhodes v. Illinois Dep’t of Transportation, 359 F.3d 498, 504 (7th Cir.2004), quoting Troupe v. May Dep’t Stores Co., 20 F.3d 734, 737 (7th Cir.1994). Coleman has presented no direct evidence of retaliation, so she relies on a mosaic of circumstantial evidence. In both retaliation and discrimination cases, we have recognized three categories of circumstantial evidence available to a plaintiff using the “convincing mosaic” approach. See, e.g., Volovsek v. Wisconsin Dep’t of Agriculture, Trade & Consumer Protection, 344 F.3d 680, 689 (7th Cir. 2003). One includes “suspicious timing, ambiguous statements oral or written, ... and other bits and pieces from which an inference of [retaliatory] intent might be drawn.” Silverman, 637 F.3d at 734, quoting Troupe, 20 F.3d at 736. Another is “evidence, but not necessarily rigorous statistical evidence, that similarly situated employees were treated differently.” Volovsek, 344 F.3d at 689. Another type is “evidence that the employer offered a pretextual reason for an adverse employment action.” Dickerson v. Board of Trustees of Community College Disk No. 522, 657 F.3d 595, 601 (7th Cir.2011); Diaz v. Kraft Foods Global, Inc., 653 F.3d 582, 586-87 (7th Cir.2011). 8 “Each type of evidence is sufficient by itself (depending of course on its strength in relation to whatever other evidence is in the case) to support a judgment for the plaintiff; or they can be used together.” Troupe, 20 F.3d at 736. Coleman has offered evidence of suspicious timing and pretext, and that evidence is sufficient to present a genuine issue of fact as to the Postal Service’s motives in suspending and then firing her. Timing: We have often invoked the general rule that “temporal proximity between an employee’s protected activity and an adverse employment action is rarely sufficient to show that the former caused the latter.” O’Leary, 657 F.3d at 635, citing Leitgen, 630 F.3d at 675. When temporal proximity is one among several tiles in an evidentiary mosaic depicting retaliatory motive, however, “[s]uspicious timing ... can sometimes raise an inference of a causal connection.” Magyar v. St. Joseph Regional Medical Center, 544 F.3d 766, 772 (7th Cir.2008); see Scaife v. Cook County, 446 F.3d 735, 742 (7th Cir. 2006) (“Close temporal proximity provides *861 evidence of causation and may permit a plaintiff to survive summary judgment provided that there is other evidence that supports the inference of a causal link.”), quoting Lang v. Illinois Dep’t of Children & Family Services, 361 F.3d 416, 419 (7th Cir.2004). Our cases reject any bright-line numeric rule, but when there is corroborating evidence of retaliatory motive, as there is here, an interval of a few weeks or even months may provide probative evidence of the required causal nexus. See Magyar, 544 F.3d at 772 (“This court has found a month short enough to reinforce an inference of retaliationciting Lang, 361 F.3d at 419. “Deciding when the inference is appropriate cannot be resolved by a legal rule; the answer depends on context.... A jury, not a judge, should decide whether the inference is appropriate.” Loudermilk, 636 F.3d at 315. Coleman’s protected activity began with informal complaints of race and sex discrimination that reached Sove, one of the relevant decision-makers, in May 2005. In June, Coleman received a new and unpleasant work assignment, which she refused, resulting in discipline. Then, after her request for advance sick leave was denied, she filed an EEO request for counseling, she was asked to.work in a storeroom, she checked herself into the hospital, and she was suspended — all within a span of about six weeks. The suspension came on August 3, 2005. That was the day she was released from the hospital and the day her psychiatrist told Berry of Coleman’s homicidal thoughts. But the suspension also occurred a few weeks after the friction between Coleman and Berry, which followed her complaints of discrimination, had built up to the point that she checked herself into the hospital. Later in August 2005, she filed her first formal EEOC charge. She filed her second formal EEOC charge in December 2005. Five weeks after that, she was fired. Even if the sequence of events alone would not be enough by itself, this sequence of protected activity and punitive action could lend some support to a reasonable juror’s inference of retaliation. See, e.g., Hunt-Golliday v. Metropolitan Water Reclamation Dist., 104 F.3d 1004, 1014 (7th Cir.1997) (“Interpreting the facts in [the plaintiffs] favor, she can show a pattern of criticism and animosity by her supervisors following her protected activities ... [that] supports the existence of a causal link.”). Pretext: Coleman’s timing evidence does not stand alone. She has also presented evidence that the employer’s stated reason for acting was pretextual, which also tends to support an inference of retaliation. The Postal Service’s explanation for both the suspension on August 3, 2005 and the termination on January 13, 2006 is Coleman’s supposed violation of the rule against threats and violence. If that explanation were beyond reasonable dispute, we would agree with the district court and affirm summary judgment on the retaliation claims. As we explained above in detail, however, Coleman has offered substantial evidence that the supposed rule violation was only a pretext for unlawful motives. A jury could reasonably conclude (though of course it would not be required to conclude) that the Postal Service acted for reasons other than its stated reason. Without repeating that discussion in detail, we conclude that when combined with the fairly close sequence of Coleman’s protected activity and the actions taken against her, that evidence of pretext could support a reasonable inference of retaliatory intent, thus precluding summary judgment. 9 *862 Under the convincing mosaic approach, a retaliation case can “be made by assembling a number of pieces of evidence none meaningful in itself, consistent with the proposition of statistical theory that a number of observations each of which supports a proposition only weakly can, when taken as a whole, provide strong support if all point in the same direction.” Cole v. Illinois, 562 F.3d 812, 815 n. 2 (7th Cir. 2009), quoting Sylvester v. SOS Children’s Villages Illinois, 453 F.3d 900, 903 (7th Cir.2006). On their own, Coleman’s evidence of suspicious timing and pretext might not be enough to show a causal connection between her protected activities and her suspension or termination. Together, however, they are sufficient to withstand summary judgment and create a question for the jury. IV. Conclusion In adjudicating claims under federal employment discrimination statutes, a court does not sit as a “super-personnel department,” second-guessing an employer’s “business decision as to whether someone should be fired or disciplined because of a work-rule violation.” Ptasznik v. St. Joseph Hosp., 464 F.3d 691, 697 (7th Cir.2006), quoting Ballance v. City of Springfield, 424 F.3d 614, 621 (7th Cir.2005). But we must also resist the temptation to act as jurors when considering summary judgment motions. Plaintiff Coleman has offered enough evidence of race and sex discrimination and retaliation to withstand summary judgment. The judgment of the district court is therefore Reversed and the case is Remanded for further proceedings consistent with this opinion. WOOD, Circuit Judge, with whom TINDER and HAMILTON, Circuit Judges, join, concurring. The lead opinion carefully analyzes Denise Coleman’s claims that the Post Office’s decision to fire her violated Title VII’s prohibitions against discriminatory employment decisions (here, on grounds of race and sex) and retaliatory actions. See 42 U.S.C. §§ 2000e-2(a)(l) (discrimination), 2000e-3(a) (retaliation). For the discrimination claim, the opinion meticulously applies the so-called indirect method of proof, which originated with the Supreme Court’s 1973 decision in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973); for the retaliation claim the opinion turns to the so-called direct method of proof, and more particularly to the indirect (or “mosaic”) way of directly proving retaliation. It concludes succinctly that Coleman managed to put enough in the record to defeat the defendant’s motion for summary judgment. A jury might find in Coleman’s favor, given the inconsistencies in the Post Office’s treatment of other workers who also violated the violence rule, even though the odds may be against Coleman here. Summary judgment, however, is not about odds, once a threshold has been crossed. I agree with my colleagues that Coleman has presented enough on both theories to move forward with her case. *863 I write separately to call attention to the snarls and knots that the current methodologies used in discrimination cases of all kinds have inflicted on courts and litigants alike. The original McDonnell Douglas decision was designed to clarify and to simplify the plaintiffs task in presenting such a case. Over the years, unfortunately, both of those goals have gone by the wayside. We now have, for both discrimination and retaliation cases, two broad approaches — the “direct” and the “indirect.” But the direct approach is not limited to cases in which the employer announces “I have decided to fire you because you are a woman [or a member of any other protected class].” Instead, the direct method permits proof using circumstantial evidence, as we acknowledged in Troupe v. May Dep’t Stores Co., 20 F.3d 734 (7th Cir.1994). Like a group of Mesopotamian scholars, we work hard to see if a “convincing mosaic” can be assembled that would point to the equivalent of the blatantly discriminatory statement. If we move on to the indirect method, we engage in an allemande worthy of the 16th century, carefully executing the first four steps of the dance for the prima facie case, shifting over to the partner for the “articulation” interlude, and then concluding with the examination of evidence of pretext. But, as my colleagues correctly point out, evidence relevant to one of the initial four steps is often (and is here) equally helpful for showing pretext. Perhaps McDonnell Douglas was necessary nearly 40 years ago, when Title VII litigation was still relatively new in the federal courts. By now, however, as this case well illustrates, the various tests that we insist lawyers use have lost their utility. Courts manage tort litigation every day without the ins and outs of these methods of proof, and I see no reason why employment discrimination litigation (including cases alleging retaliation) could not be handled in the same straightforward way. In order to defeat summary judgment, the plaintiff one way or the other must present evidence showing that she is in a class protected by the statute, that she suffered the requisite adverse action (depending on her theory), and that a rational jury could conclude that the employer took that adverse action on account of her protected class, not for any non-invidious reason. Put differently, it seems to me that the time has come to collapse all these tests into one. We have already done so, when it comes to the trial stage of a case. See, e.g., EEOC v. Bd. of Regents of Univ. of Wise. Sys., 288 F.3d 296, 301 (7th Cir.2002). It is time to finish the job and restore needed flexibility to the pre-trial stage. With those observations, I concur in my colleagues’ opinion. Footnotes [1] The Postal Service calls Coleman’s version an "embellished” “misstatement of the facts” because they only “pulled” a knife and did not hold it to the victim’s throat. We doubt that the difference between holding a knife to a man’s throat and merely displaying it while holding him down is material for purposes of summary judgment. [2] For cases from other circuits, see, for example, Russell v. City of Kansas City, 414 F.3d 863, 868 (8th Cir.2005) (employing "comparable seriousness” standard); Graham v. Long Island R.R., 230 F.3d 34, 40 (2d Cir.2000) (same); Kendrick v. Penske Transportation Services, Inc., 220 F.3d 1220, 1230 (10th Cir. 2000) (same); Holbrook v. Reno, 196 F.3d 255, 261 (D.C.Cir.1999) (same); Taylor v. Virginia Union University, 193 F.3d 219, 234 (4th Cir.1999) (en banc) (same), abrogated in part on other grounds by Desert Palace, Inc. v. Costa, 539 U.S. 90, 92, 123 S.Ct. 2148, 156 L.Ed.2d 84 (2003). [3] Perhaps if the situation were reversed, if Coleman had threatened another employee with a weapon while Arient and Pelletier had only made alarming statements to a third-party, their conduct would be less serious. For example, the Tenth Circuit once found that a proposed comparator who had threatened a co-worker with assault and then arguably threatened his supervisor with physical violence “did not violate work rules of comparable seriousness” as the plaintiff, who had physically assaulted his supervisor by pushing him to the ground. See Kendrick, 220 F.3d at 1232. [4] The Postal Service may mean simply that the Arient and Pelletier suspensions are not comparable to Coleman’s termination be *852 cause the plant leadership did not honestly regard them, but did regard Coleman, as presenting a serious ongoing threat. An employer’s honest belief about its motives for disciplining a Title VII disparate treatment plaintiff is relevant, but at the pretext stage, not for the plaintiff's prima facie case. The similarly-situated inquiry is about whether employees are objectively comparable, while the pretext inquiry hinges on the employer’s subjective motivations. As discussed below, however, there are reasons to doubt even that the Postal Service subjectively believed Coleman was dangerous. The Postal Service also cites Bodenstab v. County of Cook, 569 F.3d 651, 657 n. 2 (7th Cir.2009), for the proposition that fighting with other employees and bringing a gun to the workplace are not comparable to threatening to kill a supervisor. But Bodenstab’s passing discussion of the similarly-situated prong in footnote 2 is dicta; the court chose to "skip over” the prima facie analysis, and its central holding was that the plaintiff had failed to establish pretext. Id. at 657. [5] For scholarly criticism of this phenomenon, see Suzanne B. Goldberg, Disciimination by Comparison, 120 Yale. LJ. 728, 734 (2011) ("The judicial demand for comparators continues largely unabated ..., sharply narrowing both the possibility of success for individual litigants and, more generally, the very meaning of discrimination.”); Charles A. Sullivan, The Phoenix from the Ash: Proving Discrimination by Comparators, 60 Ala. L.Rev. 191, 216 (2009) (criticizing the tendency of courts "to require the comparator to be the almost-twin of the plaintiff before the comparison is sufficiently probative”); Ernest F. Lidge III, The Courts’ Misuse of the Similarly Situated Concept in Employment Discrimination Law, 67 Mo. L.Rev. 831, 832 (2002) (noting that courts find "that potential comparators are not similarly situated because of relatively minor, or irrelevant, distinctions between the comparators and the plaintiff”). [6] In her report, Dr. Ionescu indicated that Coleman gave "verbal agreement” to the conversation she had with Berry. Such consent would negate what might otherwise raise a serious issue of physician-patient confidentiality. See 735 ILCS § 5/8-802 ("No physician or surgeon shall be permitted to disclose any information he or she may have acquired in attending any patient in a professional character, necessary to enable him or her professionally to serve the patient.”). The Illinois Mental Health and Developmental Disabilities Confidentiality Act requires that "(a]ll records and communications” made in the course of therapy "shall be confidential and shall not be disclosed,” with certain exceptions. 740 ILCS § 110/3(a). One such exception applies "when ... a therapist, in his or her sole discretion, determines that disclosure is necessary to ... protect the recipient or other person against a clear, imminent risk of serious physical or mental injury or disease or death.” 740 ILCS § 110/11. In light of the fact that Dr. Ionescu discharged Coleman as "stable” the very day she spoke with Berry, it seems highly unlikely that she considered Coleman a "clear, imminent risk” to his safety at the time of their conversation. [7] Other circuit courts, hewing more closely to McDonnell Douglas, channel comparator evidence into the pretext phase of the sequence. See, e.g. Rioux v. City of Atlanta, 520 F.3d 1269, 1277 (11th Cir.2008) ("We, too, address the sufficiency of any comparator evidence in our examination of pretext, rather than as an element of Rioux’s prima facie case....”); Conward v. Cambridge Sch. Comm., 171 F.3d 12, 19 (1st Cir.1999) ("the time to consider comparative evidence in a disparate treatment case is at the third step of the burden-shifting ritual, when the need arises to test the pretextuality vel non of the employer’s articulated reason for having acted adversely to the plaintiff's interests"). This approach makes sense because the probative value of a proposed comparator depends largely on the specific non-discriminatory reason the employer has put forward. As one commentator argues: "It makes no sense ... to require the plaintiff to choose comparison cases based on their relevance to the employer’s not-yet-'articulated’ justification. It would make far more sense for courts to consider the presence or absence of good comparative data as part of a review of the evidence as a whole____” Deborah C. Malamud, The Last Minuet: Disparate Treatment after Hicks, 93 Mich. L.Rev. 2229, 2293 (1995). [8] The latter two categories are similar to required elements under the indirect method, so that "our analyses overlap.” Egonmwan v. Cook County Sheriff s Dep’t, 602 F.3d 845, 851 (7th Cir.2010). The mosaic approach provides parties and courts with a little more flexibility and room for common sense than the indirect method sometimes allows. See Hasan v. Foley & Lardner LLP, 552 F.3d 520, 529 (7th Cir.2008) ("under the indirect method of proof, a plaintiff must produce evidence of how the employer treats similarly situated employees,” while "the direct method of proof imposes no such constraints”). [9] In making her argument for retaliation based on circumstantial evidence, Coleman *862 also offers Arient and Pelletier, the white men involved in the knife incident, as comparators who were outside her protected class. Such comparator evidence can be relevant in showing retaliation under the "mosaic” approach. See Volovsek, 344 F.3d at 689. But this record is simply silent as to whether either of these two white men ever complained of unlawful discrimination. Even without the use of those comparators, Coleman has enough evidence to avoid summary judgment. We will not speculate further on the matter, but note only that it should be fairly easy for a plaintiff in such a case to serve an interrogatory asking whether the relevant decision-makers had any knowledge of protected activity on the part of the proposed comparators.