Faced with a defendant's motion for summary judgment, a plaintiff asserting an intentional-discrimination claim under Title VII of the Civil Rights Act of 1964, the Equal Protection Clause, or 42 U.S.C. § 1981 must make a sufficient factual showing to permit a reasonable jury to rule in her favor. She can do so in a variety of ways, one of which is by navigating the now-familiar three-part burden-shifting framework established by the Supreme Court in McDonnell Douglas Corp. v. Green , 411 U.S. 792 , 93 S.Ct. 1817 , 36 L.Ed.2d 668 (1973). Under that framework, the plaintiff bears the initial burden of establishing a prima facie case of discrimination by proving, among other things, that she was treated differently from another "similarly situated" individual-in court-speak, a "comparator." Texas Dep't of Cmty. Affairs v. Burdine , 450 U.S. 248 , 258-59, 101 S.Ct. 1089 , 67 L.Ed.2d 207 (1981) (citing McDonnell Douglas , 411 U.S. at 804 , 93 S.Ct. 1817 ). The obvious question: Just how "similarly situated" must a plaintiff and her comparator(s) be?

To date, our attempts to answer that question have only sown confusion. In some cases, we have required a proper comparator to be "nearly identical" to the plaintiff. See , e.g. , Nix v. WLCY Radio/Rahall Commc'ns , 738 F.2d 1181 , 1185 (11th Cir. 1984) (citations omitted). In others, we have expressly rejected a nearly-identical standard. See, e.g. , Alexander v. Fulton Cty ., 207 F.3d 1303 , 1333-34 (11th Cir. 2000). In still others, without even mentioning the nearly-identical benchmark, we have deemed it sufficient that the plaintiff and the comparator engaged in the "same or similar" conduct. See, e.g. , Holifield v. Reno , 115 F.3d 1555 , 1562 (11th Cir. 1997) (per curiam). And to make matters worse, in still others we have applied both the *1218 nearly-identical and same-or-similar standards simultaneously. See, e.g. , Maniccia v. Brown , 171 F.3d 1364 , 1368 (11th Cir. 1999). It's a mess.

In an effort to clean up, and to clarify once and for all the proper standard for comparator evidence in intentional-discrimination cases, we took this case en banc and instructed the parties to address the following issue:

The Supreme Court has held that in order to make out a prima facie case of discrimination under Title VII of the Civil Rights Act of 1964, the Equal Protection Clause of the Fourteenth Amendment, or 42 U.S.C. § 1981 , a plaintiff must prove, among other things, that she was treated differently from another "similarly situated" individual. What standard does the phrase "similarly situated" impose on the plaintiff: (1) "same or similar," (2) "nearly identical," or (3) some other standard?

Our plaintiff-appellant's position is twofold. First, as a procedural matter, she urges us to "move" any qualitative analysis of comparator evidence out of the initial prima facie stage of the McDonnell Douglas analysis, where it historically has resided, and into the third-tier pretext stage. Second, as a substantive matter, she contends that we should jettison both the same-or-similar and nearly-identical tests in favor of what she calls a "flexible, common-sense" standard, which the Seventh Circuit seems to have embraced: "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." Appellant's En Banc Br. at 34 (quoting Coleman v. Donahoe , 667 F.3d 835 , 846 (7th Cir. 2012) ).

Not surprisingly, the defendants-appellees see things differently. First, they insist that the comparator evaluation should remain part of the prima facie stage of the McDonnell Douglas analysis. Second, they urge us to keep the nearly-identical standard, which, they say, reflects the dominant rule in our case law and most accurately captures the Supreme Court's understanding of the phrase "similarly situated."

For the reasons that follow, we hold, as an initial matter, that a meaningful comparator analysis must be conducted at the prima facie stage of McDonnell Douglas 's burden-shifting framework, and should not be "move[d]" to the pretext stage. With respect to the standard itself, we hold that the proper test for evaluating comparator evidence is neither plain-old "same or similar" nor "nearly identical," as our past cases have discordantly suggested. Nor is it the Seventh Circuit's so-long-as-the-comparison-isn't-useless test. Rather, we conclude that a plaintiff asserting an intentional-discrimination claim under McDonnell Douglas must demonstrate that she and her proffered comparators were "similarly situated in all material respects."

I

A

Jacqueline Lewis, an African-American woman, started working for the Union City Police Department as a patrol officer in 2001 and was promoted to detective in 2008. She suffered a heart attack the following year but was cleared to return to work without any restrictions.

In 2010, then-Police Chief Charles Odom announced a new policy requiring all officers to carry Tasers. As part of the training associated with the new policy, officers had to receive a five-second Taser shock. 1

*1219 After hearing about the Taser policy and being scheduled for separate pepper-spray training, Lewis became concerned that she might be at an increased risk of injury because of her earlier heart attack. Lewis's doctor agreed and, due to what she described as "several chronic conditions including a heart condition," she informed Chief Odom that she "would not recommend" that either a Taser or pepper spray be used either "on or near" Lewis.

Because as a detective Lewis would inevitably be (at the very least) "near" pepper spray-and under the new policy, Tasers, as well-Chief Odom concluded that the restrictions described by Lewis's doctor prevented her from performing the essential duties of her job. Accordingly, Lewis was placed on unpaid administrative leave "until such time [as her doctor] release[d] [her] to return to full and active duty." Lewis was instructed "to complete the necessary FMLA paperwork concerning [her] absence" and told that she could use her accrued paid leave until it was expended. After a few weeks, Lewis had exhausted all of her accrued leave but still hadn't completed the necessary FMLA paperwork. As a result, her absence was deemed "unapproved" and she was terminated pursuant to the Union City Personnel Policy, which stated that "[a]ny unapproved leave of absence [is] cause for dismissal."

B

Lewis brought this action against Union City and Chief Odom (together, "the City"), alleging, as relevant here, race and gender discrimination in violation of Title VII, the Equal Protection Clause, and 42 U.S.C. § 1981 . 2 The City moved for summary judgment, and Lewis filed a response in which she identified as comparators two other Union City police officers whom she claimed had been treated more favorably. The first was Sergeant Cliff McClure, a white man who failed the "balance" portion of a physical-fitness test in 2014 (nearly four years after the events culminating in Lewis's termination) and was given 90 days of unpaid administrative leave to remedy the conditions that caused him to fail. McClure retook (and passed) the test within the 90-day period and returned to work. The second was Officer Walker Heard, a white man who failed an "agility" test in 2013 (almost three years after Lewis's firing) and was also placed on unpaid administrative leave for 90 days. Heard was offered (and ultimately declined) a position as a dispatcher. The offer, however, remained open for approximately 11 months while Heard's attorney negotiated with the City regarding allegations that he suffered from a disability. In the end, Heard was terminated after 449 days on unpaid administrative leave because he was unable to demonstrate his fitness to be a patrol officer. 3

The district court granted summary judgment to the City, concluding as to the race- and gender-discrimination claims that Lewis's "proffered comparators d[id] not qualify under either [the 'nearly identical' or the 'same or similar'] standard."

*1220 Over Judge Tjoflat's dissent, a panel of this Court reversed, holding in relevant part that Lewis had presented enough evidence to create a genuine issue of material fact as to her race- and gender-discrimination claims. Lewis v. City of Union City , 877 F.3d 1000 , 1004 (11th Cir. 2017), reh'g en banc granted, opinion vacated , Lewis v. City of Union City , 893 F.3d 1352 (11th Cir. 2018). In particular, the panel determined that McClure and Heard were valid comparators for purposes of assessing Lewis's prima facie case. 877 F.3d at 1015-18 . In the course of so holding, the panel acknowledged that "[t]his Circuit often has applied or referred to a 'nearly identical' standard to determine whether proposed comparators are similarly situated," but it rejected that test, which, it reasoned, applies only in workplace-misconduct cases. Id. at 1017 n.11.

We vacated the panel's opinion and took this case en banc to clarify the proper standard for comparator evidence in intentional-discrimination cases. 4

II

This case arises principally under Title VII of the Civil Rights Act of 1964, which makes it unlawful for an employer "to fail or refuse to hire or to discharge any individual, or otherwise to discriminate against any individual with respect to his compensation, terms, conditions, or privileges of employment, because of such individual's race, color, religion, sex, or national origin." 42 U.S.C. § 2000e-2(a)(1). "The language of Title VII makes plain the purpose of Congress to assure equality of employment opportunities and to eliminate those discriminatory practices and devices" that have been used to disadvantage racial, gender, and religious minorities in the workplace. McDonnell Douglas Corp. v. Green , 411 U.S. 792 , 800, 93 S.Ct. 1817 , 36 L.Ed.2d 668 (1973). 5

In order to survive summary judgment, a plaintiff alleging intentional discrimination must present sufficient facts to permit a jury to rule in her favor. One way that she can do so is by satisfying the burden-shifting framework set out in McDonnell Douglas . 6 When proceeding under McDonnell Douglas , the plaintiff bears the initial burden of establishing a prima facie case of discrimination by showing (1) that she belongs to a protected class, (2) that she was subjected to an adverse employment action, (3) that she *1221 was qualified to perform the job in question, and (4) that her employer treated "similarly situated" employees outside her class more favorably. See, e.g. , Holifield v. Reno , 115 F.3d 1555 , 1561-62 (11th Cir. 1997) (citing McDonnell Douglas, 411 U.S. at 802 , 93 S.Ct. 1817 ). If the plaintiff succeeds in making out a prima facie case, the burden shifts to the defendant to articulate a legitimate, nondiscriminatory reason for its actions. Tex. Dep't of Cmty. Affairs v. Burdine, 450 U.S. 248 , 253, 101 S.Ct. 1089 , 67 L.Ed.2d 207 (1981). Finally, should the defendant carry its burden, the plaintiff must then demonstrate that the defendant's proffered reason was merely a pretext for unlawful discrimination, an obligation that "merges with the [plaintiff's] ultimate burden of persuading the [factfinder] that she has been the victim of intentional discrimination." Id. at 256 , 101 S.Ct. 1089 .

The question before the en banc court is whether Lewis adequately showed that the City treated "similarly situated" employees outside her class more favorably than her. That question, in turn-and in the light of the parties' contentions-comprises two subsidiary questions. First, should the "similarly situated"- i.e. , comparator-analysis be conducted at the prima facie stage of the McDonnell Douglas framework, as we (and the Supreme Court) have traditionally held, or should it instead be reserved for the pretext stage? Second, and in either event, what is the proper standard for determining whether a plaintiff and her comparators are "similarly situated"? We turn to a consideration of those questions. 7

A

All here seem to agree that the qualitative assessment of comparator evidence has historically occurred in-and been an integral part of-the plaintiff's prima facie case. In her brief, for instance, Lewis recites the black-letter principle that in order to make out a prima facie case in a wrongful-termination action like this, "a plaintiff must show that she: (1) is a member of a protected class; (2) was qualified for the position from which she was terminated; (3) was terminated; and (4) was treated less favorably than similarly situated employees outside her protected class ." Appellant's En Banc Br. at 21 (citing, e.g. , St. Mary's Honor Ctr. v. Hicks , 509 U.S. 502 , 506, 113 S.Ct. 2742 , 125 L.Ed.2d 407 (1993) ) (emphasis added). Even so, Lewis now asks us to "move" the similarly-situated comparison out of the initial prima facie stage of the McDonnell Douglas analysis and into the tertiary pretext stage. Id . at 15, 30. For the reasons that follow, we decline to do so.

For starters, the Supreme Court has repeatedly (and consistently) included a comparator-evidence assessment-using one formulation or another-as an element of a plaintiff's prima facie case. Beginning in McDonnell Douglas itself, the Court emphasized that a Title VII plaintiff-there bringing a failure-to-hire claim-carries his prima facie burden "by showing (i) that he belongs to a racial minority; (ii) that he applied and was qualified for a job for which the employer was seeking applicants; (iii) that, despite his qualifications, he was rejected; and"-importantly here-"(iv) that, after his rejection, the position remained open and the employer continued to seek applicants from persons of complainant's qualifications." 411 U.S. at 802 , 93 S.Ct. 1817 . The Court reiterated a similar four-part prima facie test in *1222 Furnco Construction Corp. v. Waters , 438 U.S. 567 , 575, 98 S.Ct. 2943 , 57 L.Ed.2d 957 (1978), and again in St. Mary's Honor Center v. Hicks , 509 U.S. at 506 , 113 S.Ct. 2742 . Most recently, in Young v. United Parcel Service, Inc. , the Court applied the McDonnell Douglas framework to a failure-to-accommodate claim arising under the Pregnancy Discrimination Act, and in so doing restated yet again the required four-step test: "[A] plaintiff ... may make out a prima facie case by showing, as in McDonnell Douglas , [1] that she belongs to the protected class, [2] that she sought accommodation, [3] that the employer did not accommodate her, and [4] that the employer did accommodate others 'similar in their ability or inability to work.' " --- U.S. ----, 135 S.Ct. 1338 , 1354, 191 L.Ed.2d 279 (2015) (enumeration added).

The Supreme Court has located the comparator analysis in McDonnell Douglas 's preliminary stage for good reason. Lest we forget, the plaintiff's burden at step one is to show a prima facie case of something in particular-namely, unlawful intentional "discrimination." See, e.g. , McDonnell Douglas , 411 U.S. at 802 , 93 S.Ct. 1817 ("The complainant in a Title VII trial must carry the initial burden under the statute of establishing a prima facie case of racial discrimination ." (emphasis added)); Burdine , 450 U.S. at 252-53 , 101 S.Ct. 1089 ("discrimination"); St. Mary's , 509 U.S. at 506 , 113 S.Ct. 2742 ("discrimination"). "The prima facie case serves an important function," the Supreme Court has said, in that "it eliminates the most common nondiscriminatory reasons" for the employer's treatment of the plaintiff-and in so doing "give[s] rise to an inference of unlawful discrimination." Burdine , 450 U.S. at 253-54 , 101 S.Ct. 1089 ; see also Furnco , 438 U.S. at 577 , 98 S.Ct. 2943 (same); Int'l Bhd. of Teamsters v. United States , 431 U.S. 324 , 358, 97 S.Ct. 1843 , 52 L.Ed.2d 396 (1977) (same). A successful prima facie showing thus establishes a "legally mandatory, rebuttable presumption" of intentional discrimination, Burdine , 450 U.S. at 254 n.7, 101 S.Ct. 1089 , that "produces a required conclusion in the absence of explanation" by the employer, St. Mary's , 509 U.S. at 506 , 113 S.Ct. 2742 (quotation marks omitted). In other words, establishing a prima facie case of discrimination entitles the plaintiff to judgment-to victory-if the employer either can't, won't, or doesn't provide a nondiscriminatory explanation for its actions. See Burdine , 450 U.S. at 254 , 101 S.Ct. 1089 ("[I]f the employer is silent in the face of the presumption, the court must enter judgment for the plaintiff because no issue of fact remains in the case."). It follows, therefore, that at the prima facie stage the plaintiff must show a potential "winner"- i.e. , enough to give rise to a valid inference that her employer engaged in unlawful intentional "discrimination." 8

So what exactly is this "discrimination," an inference of which is the object of the plaintiff's prima facie case? As we have said many times-and as all of us know intuitively-"[d]iscrimination consists of treating like cases differently ." N.L.R.B. v. Collier , 553 F.2d 425 , 428 (5th Cir. 1977) (emphasis added); see also, e.g. , Frosty Morn Meats, Inc. v. N.L.R.B. , 296 F.2d 617 , 621 (5th Cir. 1961). The converse, of course, is also true: Treating different *1223 cases differently is not discriminatory, let alone intentionally so. See Nix v. WLCY Radio/Rahall Commc'ns , 738 F.2d 1181 , 1186 (11th Cir. 1984) ("[I]f an employer applies a rule differently to people it believes are differently situated, no discriminatory intent has been shown.") (quoting Chescheir v. Liberty Mut. Ins. Co. , 713 F.2d 1142 , 1148 (5th Cir. 1983) ) (emphasis added).

By its very nature, therefore, discrimination is a comparative concept-it requires an assessment of whether "like" (or instead different) people or things are being treated "differently." In light of that reality-and because a sufficient prima facie showing gives rise to an inference of unlawful discrimination-"mov[ing]" the comparator analysis out of the initial prima facie stage and into the tertiary pretext stage, as Lewis requests, would make no sense. Were we to do so, a plaintiff could demonstrate a potential winner at step one by showing only (1) that she belongs to a protected class, (2) that she suffered some adverse employment action, and (3) that she was qualified to perform the job in question. But even conclusive proof of those three elements can't entitle a plaintiff to judgment on a claim for unlawful discrimination . Absent a qualitative comparison at the prima facie stage- i.e. , without determining whether the employer treated like cases differently-there's no way of knowing (or even inferring) that discrimination is afoot. Think about it: Every qualified minority employee who gets fired, for instance, necessarily satisfies the first three prongs of the traditional prima facie case. But that employee could have been terminated because she was chronically late, because she had a foul mouth, or for any of a number of other nondiscriminatory reasons. It is only by demonstrating that her employer has treated "like" employees "differently"- i.e. , through an assessment of comparators-that a plaintiff can supply the missing link and provide a valid basis for inferring unlawful discrimination .

You see the problem: Lewis's proposal that the qualitative assessment of comparator evidence be "move[d]" out of the prima facie stage and into the pretext stage would allow the plaintiff to proceed-and potentially to win-without any good ground for presuming that discrimination has occurred. Doing so would effectively shift to the defendant the burden of disproving discrimination-which is precisely what the Supreme Court has forbidden. See, e.g. , Burdine , 450 U.S. at 253, 254-58 , 101 S.Ct. 1089 (rejecting rules that placed too stringent a burden on a gender-discrimination defendant and emphasizing that the "ultimate burden of persuading the trier of fact that the defendant intentionally discriminated against the plaintiff remains at all times with the plaintiff"). 9

*1224 * * *

We have no trouble concluding, therefore, that a meaningful comparator analysis must remain part of the prima facie case. In order to defeat summary judgment, a Title VII plaintiff proceeding under McDonnell Douglas must prove, as a preliminary matter, not only that she is a member of a protected class, that she suffered an adverse employment action, and that she was qualified for the job in question, but also that she was treated less favorably than "similarly situated" individuals outside her class.

B

Having resolved that a comparator analysis must be conducted at the prima facie stage of the McDonnell Douglas framework-and not relegated to the pretext phase-we must now decide exactly what sort of showing the phrase "similarly situated" requires a plaintiff to make and, in so doing, begin to flesh out the standard's parameters.

As we've already confessed, we've made something of a hash of the "similarly situated" issue, bouncing back and forth (and back and forth) between two standards-"nearly identical" and "same or similar." Compare, e.g. , Nix , 738 F.2d at 1185 ("nearly identical"), with, e.g. , Holifield , 115 F.3d at 1562 ("same or similar"), and, e.g. , Maniccia v. Brown , 171 F.3d 1364 , 1368 (11th Cir. 1999) (both). We took this case en banc principally to determine whether we should continue to apply one of those standards, or instead adopt some other test.

1

As an initial matter, no one seems to be advocating the same-or-similar standard-presumably because it is simultaneously too strict and too lenient, and thus incoherent. The terms "same" and "similar" denote different things. In ordinary usage, "same" means precise, jot-for-jot correspondence-clearly too exacting. See Webster's Second New International Dictionary 2209 (1944). "Similar," a much flabbier concept that includes among its definitions "somewhat like," id . at 2340, has its own problems-chief among them that it adds absolutely nothing to the general " similarly situated" standard.

Lewis's principal concern is that we reject the nearly-identical test, which she finds too "rigid." She urges us instead to adopt the Seventh Circuit's "flexible, common-sense" standard pursuant to which the "similarly situated" requirement is satisfied "[s]o long as the distinctions between the plaintiff and the proposed comparators are not 'so significant that they render the comparison effectively useless.' " Appellant's En Banc Br. at 34 (quoting Coleman v. Donahoe , 667 F.3d 835 , 846 (7th Cir. 2012) ). The City, by contrast, wants us to keep the nearly-identical standard, which, it says, reflects the dominant theme in our case law and most accurately captures the Supreme Court's understanding of the phrase "similarly situated."

We conclude that neither party's proposal quite fits the bill. For reasons we will explain, Lewis's not-useless standard is too lax, and the City's (and occasionally our own) nearly-identical test is too strict. We hold, instead-without trying to force an artificial gloss-that a plaintiff must show that she and her comparators are "similarly situated in all material respects." That standard, we think, best and most fairly implements federal statutory prohibitions on "discrimination," properly balances the need to protect employees from invidious discrimination with the deference owed to *1225 employers' rational business judgments, and sensibly serves considerations of sound judicial administration by making summary judgment available in appropriate (but by no means all) cases.

2

In assessing the parties' positions-and adopting our own-we take as our lodestars (1) the ordinary meaning of the term "discrimination" and (2) the twin policies that the Supreme Court has said animate the McDonnell Douglas framework's prima facie case. First, "discrimination": Discrimination, as we have explained, is the act of "treating like cases differently." E.g. , Collier , 553 F.2d at 428 (emphasis added). Quintessential discrimination-the Platonic form, if you will-therefore requires true "like[ness]," perfect apples-to-apples identity. As Lewis and her amici correctly explain, however, in the real world (and real workplaces) "doppelgangers are like unicorns"-they don't exist. So practically speaking, at least, perfect identity is a non-starter. But in adopting a comparator standard, we must not stray too far from paradigmatic notions of discrimination, lest we sanction a regime in which treating different things differently violates Title VII, which clearly it does not. Second, and relatedly, we must be mindful of the two functions that the Supreme Court has said the prima facie case is designed to serve-(1) to eliminate "the most common nondiscriminatory reasons" for an employer's conduct, and (2) to provide a sound basis for an "inference of unlawful discrimination." Burdine , 450 U.S. at 253-54 , 101 S.Ct. 1089 ; see also Furnco , 438 U.S. at 577 , 98 S.Ct. 2943 ; Young , 135 S.Ct. at 1354 .

It's clear, we think, that Lewis's proposed standard-which, again, deems the similarly-situated standard satisfied "[s]o long as the distinctions between the plaintiff and the proposed comparators are not 'so significant that they render the comparison effectively useless' "-fails these tests. In its looseness, Lewis's standard departs too dramatically from the essential sameness that is necessary to a preliminary determination that the plaintiff's employer has engaged in unlawful "discrimination." A plaintiff and a comparator might be alike in some (even random) sense, such that a comparison wouldn't necessarily be irrelevant or crazy. But without closer correspondence, the comparison wouldn't provide any sound basis for eliminating legitimate reasons for an employer's conduct or validly inferring discriminatory animus.

Separately, Lewis's not-useless standard risks giving courts too much leeway to upset employers' valid business judgments. In applying McDonnell Douglas , the Supreme Court has stressed the importance of striking an appropriate balance between employee protection and employer discretion. In McKennon v. Nashville Banner Publishing Co. , for instance, the Court observed that federal antidiscrimination statutes do "not constrain employers from exercising significant other prerogatives and discretions in the course of the hiring, promoting, and discharging of their employees," and emphasized that courts deciding cases arising under those laws "must recognize" not only "the important claims of the employee[s]" but also "the legitimate interests of the employer." 513 U.S. 352 , 361, 115 S.Ct. 879 , 130 L.Ed.2d 852 (1995). The reason for deference, the Court has explained, is that "[c]ourts are generally less competent than employers to restructure business practices, and unless mandated to do so by Congress they should not attempt it." Furnco , 438 U.S. at 578 , 98 S.Ct. 2943 . By permitting cases to proceed on the most meager showing of similarity between a *1226 plaintiff and her comparators, Lewis's not-useless standard would thrust courts into staffing decisions that bear no meaningful indicia of unlawful discrimination.

Finally, it seems to us inevitable that Lewis's proposal would effectively eliminate summary judgment as a tool for winnowing out meritless claims. Indeed, forestalling summary judgment appears to be a feature of the not-useless standard, not a bug. Lewis repeatedly touts her standard as not only "flexible" but also inherently "factual." E.g. , Appellant's En Banc Br. at 16, 34; Appellant's En Banc Reply Br. at 9. So long, she says, as there are "enough common features to allow meaningful comparison, the question of equivalence should be left to a jury to decide." Appellant's En Banc Br. at 16. Lewis's amici likewise stress that "[a]ny comparison will necessarily entail fact-finding and weighing of evidence, which, of course, is the job of the jury"-and therefore contend that it is "only in cases of the most dissimilar proposed comparators, without any other indicia of discriminatory intent, that summary judgment may be supported." Br. of Amici for Appellant at 16. That's just too low a bar. Without prejudging, it seems to us inconceivable, for instance, that nearly every one of the hundreds of Title VII cases filed in this Circuit last year warranted a full trial.

While Lewis's standard is clearly too lenient, the City's is too strict-or at least has the appearance of being too strict. Although we have employed it for some time now-albeit inconsistently-the nearly-identical standard gives off the wrong "vibe." Despite the adverb "nearly"-and our repeated reassurances that "comparators need not be the plaintiff's doppelgangers," Flowers v. Troup County, Georgia, School District , 803 F.3d 1327 , 1340 (11th Cir. 2015), and, even more explicitly, that " '[n]early identical' ... does not mean 'exactly identical,' " McCann v. Tillman , 526 F.3d 1370 , 1374 n.4 (11th Cir. 2008) -there is a risk that litigants, commentators, and (worst of all) courts have come to believe that it requires something akin to doppelganger-like sameness. Although we must take care not to venture too far from the form-"apples should be compared to apples"-we must also remember that "[e]xact correlation is neither likely nor necessary." Dartmouth Review v. Dartmouth Coll. , 889 F.2d 13 , 19 (1st Cir. 1989), overruled on other grounds by Educadores Puertorriqueños en Acción v. Hernandez , 367 F.3d 61 (1st Cir. 2004). And we are not willing to take the risk that the nearly-identical test is causing courts reflexively to dismiss potentially valid antidiscrimination cases. 10

3

So, we are left to try to find the sweet spot between Lewis's squishy not-useless standard and the City's preferred nearly-identical standard. For reasons explained below, we hold that a plaintiff proceeding under McDonnell Douglas must show that she and her comparators are "similarly situated in all material respects." 11

*1227 a

As an initial matter, what does it mean for a plaintiff and her comparators to be "similarly situated in all material respects"? Fair question. Of course, precisely what sort of similarity the in "all material respects" standard entails will have to be worked out on a case-by-case basis, in the context of individual circumstances. But we are not without guideposts. 12

We know, for instance, that the plaintiff and her comparators need not be "similar in all but the protected ways." Young , 135 S.Ct. at 1354 . A plaintiff needn't prove, therefore, that she and her comparators are identical save for their race or gender, as the case may be. Not even the nearly-identical standard requires that level of exactitude. Nor is it necessary for a plaintiff to prove purely formal similarities- e.g. , that she and her comparators had precisely the same title. See Lathem v. Dep't of Children & Youth Servs., 172 F.3d 786 , 793 (11th Cir. 1999) ("The relevant inquiry is not whether the employees hold the same job titles, but whether the employer subjected them to different employment policies.") (citation omitted). Nor will minor differences in job function disqualify a would-be comparator. See, e.g. , Ercegovich v. Goodyear Tire & Rubber Co. , 154 F.3d 344 , 353 (6th Cir. 1998) (concluding that minor "differences in ... job activities" did not "automatically" disqualify plaintiff's proffered comparators, at least where they held "related human resources positions").

Having said that, we can also envision the sorts of similarities that will, in the main, underlie a valid comparison. Ordinarily, for instance, a similarly situated comparator-

• will have engaged in the same basic conduct (or misconduct) as the plaintiff, see, e.g. , Mitchell v. Toledo Hosp. , 964 F.2d 577 , 580, 583 (6th Cir. 1992) (holding that a plaintiff terminated for "misuse of [an employer's] property" could not rely on comparators allegedly guilty of "absenteeism" and "insubordination") 13 ;

• will have been subject to the same employment policy, guideline, or rule as the plaintiff, see, e.g. , Lathem, 172 F.3d at 793 (holding that a plaintiff's proffered comparators were valid where all were subject to the same "workplace rules or policies");

• will ordinarily (although not invariably) have been under the jurisdiction of the same supervisor as the *1228 plaintiff, see, e.g. , Jones v. Gerwens , 874 F.2d 1534 , 1541 (11th Cir. 1989) (observing that "disciplinary measures undertaken by different supervisors may not be comparable for purposes of Title VII analysis"); and

• will share the plaintiff's employment or disciplinary history, see, e.g. , Tennial v. United Parcel Serv., Inc. , 840 F.3d 292 , 304 (6th Cir. 2016) (explaining that "[d]ifferences in experience and disciplinary history" can disqualify a plaintiff's proffered comparators).

In short, as its label indicates-"all material respects"-a valid comparison will turn not on formal labels, but rather on substantive likenesses. To borrow phrasing from a recent Supreme Court decision, a plaintiff and her comparators must be sufficiently similar, in an objective sense, that they "cannot reasonably be distinguished." Young , 135 S.Ct. at 1355 . 14

b

Now, why "similarly situated in all material respects"? Most importantly, we think that it (unlike the parties' dueling proposals) hews closely, but also realistically, to the ordinary meaning of "discrimination" and, when met, satisfies the twin aims of the prima facie case-"eliminat[ing] the most common nondiscriminatory reasons" for an employer's action and thereby "giv[ing] rise to an inference of unlawful discrimination," Burdine , 450 U.S. at 253-54 , 101 S.Ct. 1089 . An all-material-respects standard also leaves employers the necessary breathing space to make appropriate business judgments. See McKennon , 513 U.S. at 361 , 115 S.Ct. 879 ; Furnco , 438 U.S. at 578 , 98 S.Ct. 2943 . An employer is well within its rights to accord different treatment to employees who are differently situated in "material respects"- e.g. , who engaged in different conduct, who were subject to different policies, or who have different work histories. Finally, the all-material-respects standard *1229 serves the interest of sound judicial administration by allowing for summary judgment in appropriate cases-namely, where the comparators are simply too dissimilar to permit a valid inference that invidious discrimination is afoot. 15

* * *

We conclude, therefore, that it is not enough for a plaintiff proceeding under McDonnell Douglas to show only that a comparison with identifiable employees outside her class wouldn't be "useless." We also hold, though, that a plaintiff needn't show that she and her comparators were "nearly identical." Instead, she must demonstrate-as part of her prima facie case-that she and her comparators are "similarly situated in all material respects." 16

III

Having settled on-and begun to flesh out-a standard to govern the "similarly situated" issue, we must now apply it to Lewis's case. We conclude that Lewis has not made out a prima facie case because she and her proffered comparators-Sergeant McClure and Officer Heard-were not "similarly situated in all material respects." To be clear, in so concluding, we consider only Lewis's own presentation and facts not reasonably subject to dispute. 17

Lewis contends that she and her comparators were "similarly situated"-even, she says, under a strict nearly-identical test-because they "were all placed on administrative leave by the City when they could not meet a physical qualification of the job of police officer." Appellant's En Banc Br.

*1230 at 13. But Lewis's broad-brush summary glosses over critical differences. Lewis and her comparators were placed on leave years apart and pursuant to altogether different personnel policies and, perhaps even more importantly, for altogether different conditions.

McClure and Heard were placed on leave pursuant to the Police Department's Physical Fitness/Medical Examinations policy-which, as the dissent acknowledges, "had not yet been issued when Lewis was fired" and, indeed, wouldn't be promulgated for another two years. Dissenting Op. at 1257-58. That policy states that an officer who is "not deemed fit for duty will be placed on unpaid administrative leave for a period of ninety (90) days"-the purpose being to allow the officer time to improve his or her health and fitness through training, diet, and exercise. Lewis, on the other hand, was placed on leave pursuant to the Union City Personnel Policy, which provides in relevant part that "[a]ny unapproved leave of absence [is] cause for dismissal." As already explained, after Lewis's doctor determined that a Taser-which Lewis would be required to carry-couldn't be used "on" or even "near" her, Lewis was instructed to apply for FMLA leave and told that she could use any other accumulated leave "until ... expired." When Lewis failed to apply for FMLA leave and exhausted her other leave, she was terminated on the ground that, under the Personnel Policy, she was absent without permission. 18

The fact that the City applied different personnel policies to Lewis and her comparators is unsurprising, because they were contending with very different underlying conditions. McClure and Heard flunked physical-fitness requirements related, respectively, to "balance" and "agility." Both were placed on leave for 90 days to remedy the problems that caused their failures. McClure was cleared by his doctor to re-take the balance test, took and passed it, and returned to work; Heard was unable to demonstrate his fitness and (following an extended negotiation with the City) was eventually terminated. For present purposes, the important point is that the deficient physical-fitness benchmarks that sidelined McClure and Heard were at least theoretically (and in McClure's case, actually) remediable. Lewis, by contrast, failed a training requirement-pertaining specifically to the carriage and use of a deadly weapon-on the ground that she suffered from what her doctor described as a "chronic" heart condition, see Doc. 58-3 at 6, and what she herself has called a "permanent" heart injury, see Doc. 72 at 8. And unlike McClure, for instance, Lewis was never cleared by her doctor to participate in the required training. 19

*1231 For these reasons-because they were subject to different personnel policies and placed on leave for different underlying conditions-we conclude that McClure and Heard were not similar to Lewis "in all material respects," and thus were not valid comparators for purposes of Lewis's prima facie case.

IV

In sum, we hold that when a plaintiff relies on the McDonnell Douglas burden-shifting framework to prove an intentional-discrimination claim using circumstantial evidence, she must demonstrate-as part of her prima facie case-that she was treated differently from other individuals with whom she was similarly situated in all material respects. We further hold that Lewis has failed to do so.

The case is REMANDED to the panel for proceedings consistent with this opinion. 20

ROSENBAUM, Circuit Judge, joined by MARTIN and JILL PRYOR, Circuit Judges, concurring in part and dissenting in part:

The Supreme Court created the McDonnell Douglas 1 framework as a delicate balance between an employee's right to work free from discrimination and an employer's right to take action against an employee for any nondiscriminatory reason. Today, the Majority Opinion drops an anvil on the employer's side of the balance. Though the Majority Opinion correctly defines "similarly situated" within the McDonnell Douglas framework as "similarly situated in all material respects," Maj. Op. at 1218, it one-sidedly implements and interprets this standard to the employer's redounding benefit. As a result, plaintiffs proceeding by circumstantial evidence in this Circuit will have a difficult time budging the now-off-kilter balance and surviving summary judgment.

First, faced with how and where to implement the "similarly situated" standard, the Majority Opinion rigorously 2 applies *1232 the standard entirely 3 at the prima facie stage of the McDonnell Douglas analysis. Yet that construction of the prima facie case rebukes its parent: McDonnell Douglas and its progeny explicitly and implicitly require a generalized application of the "similarly situated" standard at the initial, prima facie juncture and a more particularized one at the pretext phase of the framework-after the employer has satisfied its burden of coming forward with its nondiscriminatory reason for adverse action. In ratcheting up the prima facie stage's "similarly situated" standard, the Majority Opinion defies the purpose of Title VII and the McDonnell Douglas framework.

How the "similarly situated" inquiry is implemented matters: if it is turned up too high at the prima facie stage, it sweeps in the employer's nondiscriminatory reasons. And considering the employer's nondiscriminatory reasons at the prima facie stage flouts Supreme Court precedent. It also affects whether the employee ever has a chance to demonstrate that the employer's reasons were pretextual or whether the court must instead blindly accept the employer's untested assertions as a non-discriminatory basis for the employer's decision. So by locating a rigorous "similarly situated" requirement at the prima facie stage of the McDonnell Douglas framework, the Majority Opinion shrinks the number of potentially discriminated-against plaintiffs who will have an opportunity to see trial-or even to challenge their employers' proffered reasons for taking action against them.

And the errors do not end there. In applying the "similarly situated" standard to Lewis's facts, the Majority Opinion overly broadly construes the term "material" in that standard. As a result, it requires comparators to be similarly situated in immaterial ways. The Majority Opinion also omits key facts showing that Lewis and her two chosen comparators were "similarly situated" in all material ways, while violating an elementary principle of summary-judgment review by assuming facts in favor of the Department that are not supported by the record.

We will almost always be able to find a difference between a plaintiff and her proposed comparators if we strain our eyes looking hard enough. But as the Majority Opinion stresses, the "similarly situated" test is meant to ensure only that "apples *1233 [are] compared to apples." Maj. Op. at 1226. By focusing on trivial differences, though, the Majority Opinion uses the word "material" to bar meaningful comparisons between Honeycrisp and SnapDragon apples. 4 So the Majority Opinion finds that Lewis's claim fails because her doctor's note recommending against Tasing and pepper-spraying her rendered her too dissimilar from her comparators who had balance and agility issues-even though the upshot of all three officers' conditions was that the City deemed them all physically unfit to patrol the streets of Union City and, in turn, relied on that fact to put them all on administrative leave. Meanwhile, Lewis, an African-American woman, received only 21 days of administrative leave before being fired, but one white male comparator enjoyed 449 days before his dismissal and the other got up to 90 days to demonstrate he had become physically fit enough to resume his duties.

In applying the "similarly situated" standard to Lewis's facts to eliminate comparators who actually were similarly situated to her in all material respects, the Majority Opinion contradicts 46 years of Supreme Court precedent, elevates the "similarly situated" requirement to one of form over substance, and incorrectly concludes that the district court did not err in entering summary judgment against Lewis.

I respectfully dissent from these errors.

I.

I begin with the Majority Opinion's error in how and where it has implemented the "similarly situated" standard within the McDonnell Douglas framework.

In Section A, I explain that the Supreme Court established the McDonnell Douglas burden-shifting framework to allow plaintiffs who lack direct evidence of discrimination to nonetheless obtain fair scrutiny of their claims that they have been discriminated against in the workplace-that is, as a delicate balance between employers' rights to make legitimate personnel decisions and employees' rights to be free from "those discriminatory practices ... which have fostered racially stratified job environments ...." McDonnell Douglas Corp. v. Green , 411 U.S. 792 , 800, 93 S.Ct. 1817 , 36 L.Ed.2d 668 (1973). Section B demonstrates that the Supreme Court's remarks about how the McDonnell Douglas analysis works demand the conclusion that the comparator requirement is minimal and generalized at the plaintiff's initial burden-the prima facie stage-and progresses to more specificity only at the last stage-the pretext inquiry-of the framework. Next, Section C shows that the Supreme Court's actual analyses in McDonnell Douglas itself and in its progeny followed this funnel-like application of the "similarly situated" requirement in the specificity they required of comparator evidence. In Section D, I explain how the Majority Opinion's decision to move the entire "similarly situated" inquiry into the prima facie case inflicts a significant blow on a plaintiff's ability to establish discrimination in circumstantial cases and upsets the delicate balance struck by the McDonnell Douglas framework. And finally, Section E explores how the Majority Opinion errs and concludes that conducting the complete and rigorous comparator analysis at *1234 the prima-facie-case stage is fatally flawed.

A.

"[I]t is abundantly clear that Title VII tolerates no racial discrimination, subtle or otherwise." McDonnell Douglas , 411 U.S. at 801 , 93 S.Ct. 1817 . Title VII's prohibitions against discrimination undoubtedly "reflect an important national policy," U.S. Postal Serv. Bd. of Governors v. Aikens , 460 U.S. 711 , 716, 103 S.Ct. 1478 , 75 L.Ed.2d 403 (1983), because of "society's consensus that discrimination ... is a profound wrong of tragic dimension," Patterson v. McLean Credit Union , 491 U.S. 164 , 188, 109 S.Ct. 2363 , 105 L.Ed.2d 132 (1989). To effectuate Title VII's broad, nationally important remedial purposes, the Supreme Court has cautioned against issuing opinions that "signal[ ] one inch of retreat from Congress' policy to forbid discrimination in the private, as well as the public, sphere." Id.

Yet despite this unyielding mandate, plaintiffs often face a paradox: "Unless the employer is a latter-day George Washington, employment discrimination is as difficult to prove as who chopped down the cherry tree," Thornbrough v. Columbus & Greenville R. Co. , 760 F.2d 633 , 638 (5th Cir. 1985), because employers of "even minimal sophistication will neither admit discriminatory animus nor leave a paper trail demonstrating it ...." Riordan v. Kempiners , 831 F.2d 690 , 697 (7th Cir. 1987) ; see also Grigsby v. Reynolds Metals Co. , 821 F.2d 590 , 595 (11th Cir. 1987) ("[P]laintiffs ... rarely are fortunate enough to have access to direct evidence of intentional discrimination."). To chip away at that paradox, "[c]ourts today must be increasingly vigilant" because "[t]he sophisticated would-be violator has made our job a little more difficult." Aman v. Cort Furniture Rental Corp. , 85 F.3d 1074 , 1082 (3d Cir. 1996).

That's where McDonnell Douglas comes in. The "entire purpose" of McDonnell Douglas "is to compensate for the fact that direct evidence of intentional discrimination is hard to come by." Price Waterhouse v. Hopkins , 490 U.S. 228 , 271, 109 S.Ct. 1775 , 104 L.Ed.2d 268 (1989) (O'Connor, J., concurring), superseded in part by The Civil Rights Act of 1991, Tit. I, § 107(a), 105 Stat. 1075 (codified at 42 U.S.C. § 2000e-2(m) ); see also Grigsby , 821 F.2d at 595 ("The McDonnell Douglas - Burdine patterns of proof were designed to ease the evidentiary burdens on employment discrimination plaintiffs ...."). So the McDonnell Douglas framework is "designed to assure that the 'plaintiff [has] his day in court despite the unavailability of direct evidence.' " Trans World Airlines, Inc. v. Thurston , 469 U.S. 111 , 121, 105 S.Ct. 613 , 83 L.Ed.2d 523 (1985) (alteration in original) (quoting Loeb v. Textron, Inc. , 600 F.2d 1003 , 1014 (1st Cir. 1979) ).

B.

To accomplish this aim, McDonnell Douglas established a delicately balanced three-step framework, allocating the burden of production and the order for the presentation of proof in a funnel-like way that narrows from the first step of the framework to the third. See St. Mary's Honor Ctr. v. Hicks , 509 U.S. 502 , 506, 113 S.Ct. 2742 , 125 L.Ed.2d 407 (1993). The Supreme Court has described the McDonnell Douglas framework as "progressively ... sharpen[ing] the inquiry into the elusive factual question of intentional discrimination." Burdine , 450 U.S. at 255 n.8, 101 S.Ct. 1089 .

At the first step of the McDonnell Douglas framework, the plaintiff must set forth a prima facie case of discrimination.

*1235 Tex. Dep't of Cmty. Affairs v. Burdine , 450 U.S. 248 , 252-53, 101 S.Ct. 1089 , 67 L.Ed.2d 207 (1981). We have construed that requirement to mean that a plaintiff must show that (1) she belongs to a protected group, (2) she was subjected to an adverse employment action, (3) she was qualified to perform the job in question, and (4) her employer more favorably treated "similarly situated" employees outside her protected group. See, e.g. , Holifield v. Reno , 115 F.3d 1555 , 1561-62 (11th Cir. 1997) (citation omitted). The Supreme Court has noted that at the first phase of the framework, "[t]he burden ... is not onerous," 5 Burdine , 450 U.S. at 253 , 101 S.Ct. 1089 , and, in fact, is "minimal," St. Mary's , 509 U.S. at 506 , 113 S.Ct. 2742 .

Following the Court's lead, we in the past have ourselves described the plaintiff's prima facie burden as "light," Turlington v. Atlanta Gas Light Co. , 135 F.3d 1428 , 1432 (11th Cir. 1998), and as "a low bar to hurdle," Flowers v. Troup Cty., Sch. Dist. , 803 F.3d 1327 , 1336 (11th Cir. 2015). Other Circuits have also observed that the prima facie case requires only a "small showing" that can be "easily made," Caraballo-Caraballo v. Corr. Admin. , 892 F.3d 53 , 57 (1st Cir. 2018), because it was "not meant to stymie plaintiffs," Cline v. Catholic Diocese of Toledo , 206 F.3d 651 , 660 (6th Cir. 2000). As the Supreme Court has explained, the prima facie phase anticipates that the plaintiff need make only a "generalized" showing of the elements of its case. St. Mary's , 509 U.S. at 516 , 113 S.Ct. 2742 .

This "minimal" burden to make a "generalized" showing at the prima facie stage ensures the prima facie case acts as a hill and not a mountain in these notoriously difficult-to-prove cases. An employee often finds herself at a significant disadvantage to an employer when it comes to knowing the reasons for the employer's employment decision and to having access to information concerning both that decision and potential comparators. So we have recognized "that the prima facie case is designed to include only evidence that is objectively verifiable and either easily obtainable or within the plaintiff's possession." Vessels v. Atlanta Indep. Sch. Sys. , 408 F.3d 763 , 769 (11th Cir. 2005) (citing Walker v. Mortham , 158 F.3d 1177 , 1192-93 (11th Cir. 1998) ); see also Price Waterhouse , 490 U.S. at 271 , 109 S.Ct. 1775 (O'Connor, J., concurring). 6

This, of course, makes even more sense in light of the prima facie case's design to *1236 eliminate only "the most common nondiscriminatory reasons" for the employment action. Burdine , 450 U.S. at 253-54 , 101 S.Ct. 1089 ; see also Maj. Op. at 1222 (citations omitted).

Once a plaintiff clears that "low bar" by establishing a prima facie case, the employer-who is in the best position to know why it took action-must come forward with a nondiscriminatory reason for its employment action. See Burdine , 450 U.S. at 255 , 101 S.Ct. 1089 . Only after the defendant offers its nondiscriminatory reason at the second stage of the McDonnell Douglas framework does "the factual inquiry ['sharpen' and] proceed[ ] to a new level of specificity" at the third McDonnell Douglas juncture-the pretext phase. Id. at 255 & n.8, 101 S.Ct. 1089 .

Significantly, the Supreme Court has explained that at this third stage, "the inquiry now turns from the few generalized factors that establish a prima facie case to the specific proofs and rebuttals of discriminatory motivation the parties have introduced." St. Mary's , 509 U.S. at 516 , 113 S.Ct. 2742 . Only here does the "plaintiff's burden of showing pretext 'merge[ ]' with the ultimate burden of demonstrating unlawful discrimination." Clark v. Huntsville City Bd. of Educ. , 717 F.2d 525 , 529 (11th Cir. 1983) (quoting Burdine , 450 U.S. at 256 , 101 S.Ct. 1089 ).

Seizing on Supreme Court jurisprudence, several of our sister Circuits have expressly observed that the funnel-like nature of the McDonnell Douglas framework demands a funnel-like inquiry into comparators, narrowing from the "minimal," "generalized," and "not onerous" burden at the prima facie stage to a burden of more heightened specificity at the pretext stage. For example, the Sixth Circuit has explained that "a general weighing of the qualifications of [the plaintiff] and [her comparator] is necessary at the prima facie stage; however, this light review must be distinguished from the more rigorous comparison conducted at the later stages of the McDonnell Douglas analysis." Provenzano v. LCI Holdings, Inc. , 663 F.3d 806 , 813-14 (6th Cir. 2011).

Other Circuits have also arrived at this same conclusion. For instance, the Eighth Circuit has opined that "the low-threshold standard [for determining at the prima facie stage whether employees are similarly situated] more accurately reflects Supreme Court precedent." Wimbley v. Cashion , 588 F.3d 959 , 962 (8th Cir. 2009) (citation and internal quotation marks omitted). Similarly, the Tenth Circuit has remarked that "while evidence that a defendant treated a plaintiff differently than similarly-situated employees is certainly sufficient to establish a prima facie case, it is '[e]specially relevant' to show pretext if the defendant proffers a legitimate, nondiscriminatory reason for the adverse employment action." EEOC v. Horizon/CMS Healthcare Corp. , 220 F.3d 1184 , 1195 n.6 (10th Cir. 2000). 7

In fact, before today, we ourselves had caselaw that implicitly recognized that McDonnell Douglas required the "similarly situated" standard to be applied more generally at the prima facie stage before *1237 applying that standard more rigorously at the pretext juncture. For example, in Walker v. Mortham , 158 F.3d 1177 , 1192 (11th Cir. 1998), a failure-to-promote case, we warned that courts are "prohibit[ed] from requiring a plaintiff to prove" at the prima facie stage that she had the same qualifications as the applicant that the employer promoted. As Judge Tjoflat cogently explained, forcing the plaintiff to make such a rigorous showing at the prima facie stage "runs contrary to the policies underlying the McDonnell Douglas prima facie case." Id. Based on the language and purpose of McDonnell Douglas and its progeny, as I have explained, we were right to reach this conclusion.

This dichotomous approach to the rigor applied to proposed comparators accords with the difference in what the prima facie and pretext stages are meant to do. In contrast to the pretext stage, plaintiffs do not have the "ultimate burden of demonstrating unlawful discrimination" at the prima facie phase. Burdine , 450 U.S. at 256 , 101 S.Ct. 1089 . Instead, the plaintiff's prima facie burden is "quite different," Furnco Const. Corp. v. Waters , 438 U.S. 567 , 575, 98 S.Ct. 2943 , 57 L.Ed.2d 957 (1978), and "not as burdensome as succeeding on 'an ultimate finding of fact as to' a discriminatory employment action," Young v. United Parcel Serv., Inc. , --- U.S. ----, 135 S.Ct. 1338 , 1354, 191 L.Ed.2d 279 (2015) (quoting Furnco , 438 U.S. at 576 , 98 S.Ct. 2943 ). So it is critical for a court to require a plaintiff to satisfy only her "not onerous," "generalized" burden at the prima facie juncture, instead of applying the "more rigorous standard at the prima facie stage," which " 'conflate[s] the prima facie case with the ultimate issue of discrimination ....' " Rodgers v. U.S. Bank, N.A. , 417 F.3d 845 , 852 (8th Cir. 2005), abrogated on other grounds by Torgerson v. City of Rochester , 643 F.3d 1031 (8th Cir. 2011) (quoting Williams v. Ford Motor Co. , 14 F.3d 1305 , 1308 (8th Cir. 1994) ). Yet the Majority Opinion, after introducing its hyper-specific prima facie comparator analysis, provides no indication as to how it expects the inquiry to "proceed[ ] to a new level of specificity" at the pretext stage. Burdine , 450 U.S. at 255 , 101 S.Ct. 1089 . By maxing out the specificity at the prima facie stage, it simply ignores Supreme Court precedent.

C.

Courts' practices of reserving the specificity of the comparator inquiry for the pretext stage comport not only with the Supreme Court's language in McDonnell Douglas and its progeny but also with those cases' actual analyses. In McDonnell Douglas , the plaintiff, a mechanic who had previously been laid off by McDonnell Douglas, applied for reemployment and was turned down. 411 U.S. at 794-96 , 93 S.Ct. 1817 . The plaintiff filed suit, alleging, as relevant here, race discrimination. Id. at 796-97 , 93 S.Ct. 1817 . The Supreme Court concluded that the plaintiff had successfully met his prima facie burden by showing that McDonnell Douglas "sought mechanics, [the plaintiff's] trade, and continued to do so after [the plaintiff's] rejection"; that McDonnell Douglas did not dispute the plaintiff's qualifications; and that McDonnell Douglas acknowledged that the plaintiff's prior work performance was satisfactory. Id. at 802 , 93 S.Ct. 1817 .

At the second stage, McDonnell Douglas stated that it had declined to rehire the plaintiff because he had previously participated in civil disobedience against McDonnell Douglas. Id. at 803 , 93 S.Ct. 1817 .

Turning to the pretext phase of the inquiry, the Supreme Court remanded the case to allow the plaintiff "a fair opportunity to show that [McDonnell Douglas's] stated reason for [the plaintiff's] rejection *1238 was in fact pretext." Id. at 804 , 93 S.Ct. 1817 . And for the first time in the three-step analysis, the Supreme Court became more specific about comparator evidence. Whereas at the prima facie stage the Supreme Court accepted as comparators any white mechanic who had been rehired, at the pretext juncture, the Supreme Court explained, "Especially relevant to [a pretext showing] would be evidence that white employees involved in acts against petitioner of comparable seriousness to the [plaintiff's civil disobedience] were nevertheless retained or rehired." Id. (emphasis added). So only at the pretext stage did the Supreme Court determine that considering McDonnell Douglas's treatment of white mechanics who engaged in acts of "comparable seriousness," as opposed to simply white mechanics, was appropriate.

Indeed, the Supreme Court has never rigorously scrutinized comparators at the prima facie stage. 8 For instance, in Furnco , the Court determined that the plaintiffs satisfied their prima facie burden by showing only that they were qualified, "and the employer continued to seek persons of similar qualifications." 438 U.S. at 576 , 98 S.Ct. 2943 . And in St. Mary's , the Court found the prima facie burden satisfied when the plaintiff demonstrated only that he "was qualified for the position ... , that he was demoted from that position ... , and ... that the position remained open and was ultimately filled by a white man." 509 U.S. at 506 , 113 S.Ct. 2742 .

In short, the purpose of the McDonnell Douglas framework, the language of McDonnell Douglas and its progeny, and the analysis set forth in McDonnell Douglas itself all unmistakably demonstrate that the "similarly situated" requirement must be applied in a "generalized," "minimal" way at the prima facie stage and in a more "particulari[zed]" way at the pretext phase.

D.

The Majority Opinion's insistence on creating a rigorous one-stop shop for analyzing comparators at the prima facie stage throws the McDonnell Douglas framework's delicate balance out of whack. It effectively considers the employer's nondiscriminatory reasons (the second-stage inquiry) at the first stage-the prima facie case-without providing the protection of allowing the employee to present evidence and argument challenging those reasons as pretext for discrimination. As a result, the Majority Opinion's heightened application of the "similarly situated" standard at the *1239 prima facie juncture significantly reduces the employee's chances of surviving summary judgment.

Supreme Court precedent precludes what the Majority Opinion does. In particular, the Supreme Court prohibits consideration, at the prima facie stage, of the employer's reason for its actions, since allowing the employer's reason for its actions to seep into the prima facie inquiry unfairly weights McDonnell Douglas 's delicate balance in favor of the employer.

But before we can fully appreciate Supreme Court precedent on this point, we must pause briefly to consider the interconnected relationship between the rigor applied to the "similarly situated" comparator inquiry and the employer's stated nondiscriminatory reason for its actions. When we do, we find that the "similarly situated" comparator requirement and the employer's nondiscriminatory reason for its actions can be two sides of a single coin.

Employers often argue as their nondiscriminatory reason that the plaintiff and her comparator's situations differed in some way that justified the different treatment. In other words, employers tend to assert that the plaintiff and her proposed comparator were not "similarly situated." So if we allow the employer's view of who is "similarly situated" to the plaintiff to govern the "similarly situated" inquiry at the prima facie stage, we necessarily import into the prima facie phase the employer's reason for its actions-even though McDonnell Douglas prohibits such consideration until the second and third junctures of the framework.

In fact, we need not look any further than the Majority Opinion to see this principle at work. Though Lewis met her "generalized," "not onerous," and "minimal" prima facie burden to establish she was "similarly situated" to McClure and Heard when she showed that all three were patrol officers put on administrative leave by the Department because they were not physically fit for duty, the Majority Opinion nonetheless concludes that she failed to satisfy her burden. And it does so because it reasons that Lewis "was placed on leave pursuant to the Union City Personnel Policy" on administrative leave, while "McClure and Heard were placed on leave pursuant to the Police Department's Physical Fitness/Medical Examinations policy," and Lewis "suffered ... a 'chronic' heart condition," while McClure and Heard had "balance" and "agility" problems, respectively. Maj. Op. at 1229-31.

But, of course, these are not only ways of purporting to show that Lewis was not similarly situated to her proposed comparators; they also reflect the Department's stated reasons for why it took the action of putting Lewis on administrative leave. 9

*1240 Now, to be clear, and as explained further at Section II.B. (pages 1243-47), infra , that Lewis met her prima facie burden does not mean we do not consider these offered nondiscriminatory reasons for the Department's actions, which effectively propose a more particularized application of the "similarly situated" standard. We must. But we do so at the pretext stage, after the employer presents them at McDonnell Douglas 's second stage.

The Majority Opinion's failure to follow this sequence is wrong for at least two reasons. First, the Supreme Court in U.S. Postal Service Board of Governors v. Aikens , 460 U.S. 711 , 103 S.Ct. 1478 , 75 L.Ed.2d 403 (1983), prohibits us from using the employer's nondiscriminatory reason to find that the plaintiff failed to establish her prima facie case. In Aikens , the district court reversed its initial findings that the plaintiff had satisfied his prima facie burden, upon considering the defendant's explanation for why the employee was not promoted. See id. at 716 , 103 S.Ct. 1478 . The Supreme Court found this analysis fundamentally flawed. Once the employer "responds to the plaintiff's proof by offering evidence of the reason for the plaintiff's rejection," the Court explained, whether or not the plaintiff made out a prima facie case "is no longer relevant." Id. at 714 , 103 S.Ct. 1478 .

Rather, by providing its nondiscriminatory reason, the employer moves the inquiry to the pretext stage, where we must decide the "ultimate question" of whether its action against the plaintiff was discriminatory. Id. At this point, courts cannot resolve the case by returning to the prima facie stage. Id. at 717 , 103 S.Ct. 1478 . So under Aikens , courts may not consider the employer's nondiscriminatory reasons offered at the second McDonnell Douglas stage to determine that a plaintiff has failed to establish a prima facie case. 10 Yet *1241 disappointingly, the Majority Opinion never even attempts to explain how it can be squared with Aikens .

Second, by collapsing the first two stages of the McDonnell Douglas inquiry into the prima facie phase, the Majority Opinion not only strays from Aikens , it also inflicts significant consequences on McDonnell Douglas 's delicate balance-and therefore on plaintiffs. If the rigorous application of the "similarly situated" requirement occurs at the prima facie stage, it forecloses the plaintiff from challenging the employer's stated reasons for why it deems the plaintiff not similarly situated to more favorably treated potential comparators. Instead, the court must simply accept the employer's reason because it masquerades as the plaintiff's failure to show that she was "similarly situated" to her proposed comparator and prevents the plaintiff from establishing her prima facie case. And if the court concludes the plaintiff has not made out a prima facie case, the plaintiff cannot reach the third stage of the McDonnell Douglas framework, which is the first place a plaintiff can challenge the employer's stated reasons for its actions.

If, on the other hand, the heightened inquiry into "similarly situated" comparators occurs at the pretext stage, the plaintiff receives the chance to demonstrate pretext. And to the extent that a plaintiff can raise a material issue of fact concerning the employer's stated reason for its employment action-that is, cast doubt on the truthfulness of the employer's classification of the plaintiff as not "similarly situated" to more favorably treated employees-the plaintiff has the chance to do so and can survive summary judgment. See Burdine , 450 U.S. at 256 , 101 S.Ct. 1089 (reasoning that at the pretext stage, the employee "now must have the opportunity to demonstrate that the proffered reason was not the true reason for the employment decision."); cf. id . at 256 n.10, 101 S.Ct. 1089 ("Indeed, there may be some cases where the plaintiff's initial evidence [its prima facie case], combined with effective cross-examination of the defendant, will suffice to discredit the defendant's explanation").

So locating the more onerous "similarly situated" standard at the pretext juncture instead of frontloading it in the prima facie case effectuates McDonnell Douglas 's purpose to allow a plaintiff who lacks direct proof of discrimination-by definition, the only type of plaintiff who relies on the McDonnell Douglas framework-a chance to show that the same employer who might not have been forthright about its reasons for taking action against her (thus causing her to have to use the McDonnell Douglas analysis to prove her case) may not be entirely candid in litigation where it defends its actions.

Before today, we recognized the problem and impermissibility of collapsing the McDonnell Douglas framework. For instance, because an employee's "job performance is bound up" with the pretext inquiry, we deferred heavily scrutinizing that issue until the pretext stage to avoid eliding the stages of the McDonnell Douglas framework.

*1242 Alvarez v. Royal Atl. Developers, Inc., 610 F.3d 1253 , 1265 (11th Cir. 2010). In the same vein, we warned against holding that nondiscriminatory reasons are "sufficient to defeat the prima facie case" because such a rule "cannot be squared with the structure and purpose of the McDonnell Douglas framework." Vessels , 408 F.3d at 769 .

Other Circuits have identified this problem, too. For example, in Ortiz v. Norton , 254 F.3d 889 , 895 (10th Cir. 2001), the Tenth Circuit explained, "Short-circuiting the analysis at the prima facie stage frustrates a plaintiff's ability to establish that the defendant's proffered reasons were pretextual and/or that [the protected status] was the determining factor ...." (citation and internal quotation marks omitted); see also Horizon/CMS Healthcare Corp ., 220 F.3d at 1193-95 & 1195 n.7 (explaining that applying a rigorous "similarly situated" standard in the prima facie case "compress[es] the three stages of the McDonnell Douglas analysis" and "denie[s] [the plaintiff] the opportunity to show that the reasons advanced by the defendant were pretextual").

Simply put, in the context of summary judgment, requiring the plaintiff to make a heightened showing of the "similarly situated" requirement at the prima facie stage heavily weights down the employer's side of McDonnell Douglas 's delicate employee-employer balance. So by applying a searching "similarly situated" inquiry that considers the employer's nondiscriminatory reasons at the prima facie stage, the Majority Opinion severely errs. And its error ensures that at least some plaintiffs with potentially valid claims of discrimination will never get the opportunity to show that the employer's offered reason is a cover for discrimination. 11

E.

Three errors the Majority Opinion makes lead it to mistakenly conclude that the "similarly situated" requirement must be employed rigorously at the prima facie stage: (1) it emphasizes the employer's interest over the employee's interest, rather than balancing them like McDonnell Douglas 's framework was intended to do; (2) it misses the point of McDonnell Douglas and its progeny; and (3) it imagines a non-existent deluge of potentially undeserving victorious employment plaintiffs if the "similarly situated" standard is not rigorously applied at the prima facie stage. I discuss each error, in turn, below.

First , the Majority Opinion contends that imposing a rigorous "similarly situated" standard at the prima facie stage is warranted because it furthers "the twin aims of the prima facie case-eliminating the most common nondiscriminatory reasons for an employer's action and thereby giving rise to an inference of unlawful discrimination." Maj. Op. at 1228 (citation and internal quotation marks omitted). The Majority Opinion also identifies a third aim of the prima facie standard: "winnowing out meritless claims" while "leav[ing] employers the necessary breathing space to *1243 make appropriate business judgments." Maj. Op. at 1225-26, 1228-29.

No doubt all of these things are important. But as the Supreme Court has recognized, so are the employee's interests in being able to earn a livelihood without being discriminated against and in having a chance to prove her case, despite the difference in her access (versus that of her employer) to information about the employment action. Indeed, "Title VII[ ] balance[s] between employee rights and employer prerogatives." Price Waterhouse , 490 U.S. at 243 , 109 S.Ct. 1775 (plurality opinion). And in turn, "[the McDonnell Douglas ] balance of burdens is the direct result of Title VII's balance of rights." Id. at 245 , 109 S.Ct. 1775 .

Yet critically, the Majority Opinion and its decision to locate the heightened "similarly situated" standard in the prima facie case fail to properly account for employee interests and circumstances. So contrary to what McDonnell Douglas and its progeny anticipated, by focusing on the employer's interests, the Majority Opinion naturally winds up with a system decidedly weighted towards the employer instead of delicately balanced between the interests of both parties. See supra at Section I.D. That system is irreconcilable with the robust protection against workplace discrimination Congress intended Title VII to secure. See, e.g. , Teamsters v. United States , 431 U.S. 324 , 348, 97 S.Ct. 1843 , 52 L.Ed.2d 396 (1977) ("The primary purpose of Title VII was to assure equality of employment opportunities and to eliminate ... discriminatory practices ...." (internal quotation marks omitted)).

Second , the Majority Opinion clashes with McDonnell Douglas and its progeny. Perhaps the most obvious illustration of this comes from the Majority Opinion's exile of McDonnell Douglas 's "not onerous" prima facie standard-a centerpiece of its delicate balance of employers' and employees' rights-to a single footnote in a misguided effort to render it meaningless. See Maj. Op. at 1229 n.15.

The Majority Opinion tries to hide the "not onerous" elephant in the McDonnell Douglas framework by wishing it away as a mere "descriptor that doesn't pertain to the substantive standard that governs the prima facie analysis." Id . In the Majority Opinion's view, then, the plaintiff's burden to establish a prima facie case is "not onerous" with respect to something other than her burden to establish a prima facie case under the substance of the standards governing the prima facie case.

Wait ... what?

Of course, whether standards are onerous or not governs whether the plaintiff's burden on the prima facie case is onerous or not. It makes no sense to speak of the plaintiff's prima facie burden as "not onerous" if the plaintiff must, in fact, satisfy an onerous substantive standard to meet her prima facie burden.

Nor does the Majority Opinion's reliance on the quotation from Young , 135 S.Ct. at 1354 (quoting Furnco , 438 U.S. at 576 , 98 S.Ct. 2943 ), respond to this problem. The Majority Opinion opines that McDonnell Douglas 's description of the plaintiff's prima facie burden as "not onerous" is "simply an acknowledgement that, by its nature, a prima facie case is preliminary, and thus is 'not as burdensome as succeeding on "an ultimate finding of fact as to" a discriminatory employment action.' " Maj. Op. at 1229 n.15 (quoting Young , 135 S.Ct. at 1354 (quoting Furnco , 438 U.S. at 576 , 98 S.Ct. 2943 )).

But the Young quotation comes from Young 's recognition that a plaintiff presents a prima facie case by "showing actions taken by the employer from which one can infer, if such actions remain unexplained, *1244 that it is more likely than not that such actions were based on a discriminatory criterion illegal under Title VII," and that required showing "is not as burdensome as succeeding on an ultimate finding of fact as to a discriminatory employment action." Young , 135 S.Ct. at 1354 (cleaned up). Neither that statement nor anything in Young 's (or any other Supreme Court precedent's) discussion somehow supports the notion that the "not onerous" language does not apply to the substantive standards governing the prima facie stage or that the "not onerous" language applies to something other than the substantive standards at the prima facie stage. They just don't. 12

And even overlooking these hiccups, dismissing the Supreme Court's "not onerous" statement as some supposedly superfluous "descriptor" writes off and literally relegates to footnote status a central tenet of the McDonnell Douglas framework. As I have noted, McDonnell Douglas and its progeny demonstrate in case after case that, in fact, the plaintiff's "minimal" burden at the prima facie stage refers to the plaintiff's "minimal" burden under the substantive standards that govern the prima facie case. See supra at Section I.C. The Supreme Court has never once, either implicitly or explicitly, moved one iota away from this standard in the over forty years since it decided McDonnell Douglas . 13 Yet the Majority Opinion treats McDonnell Douglas as an out-of-Circuit intermediate appellate decision, leaving its main point unobserved and proceeding as if that case were, at most, potentially persuasive authority, yet of no actual persuasive value. So much for Justice Scalia's reminder that "what is required to establish the McDonnell Douglas prima facie case is infinitely less than what a directed verdict demands." St. Mary's , 509 U.S. at 515 , 113 S.Ct. 2742 . Forget "infinitely less"; the holding the Majority Opinion announces today is nearly indistinguishable from what a directed verdict demands.

I cannot agree with this. So long as we remain bound by the Supreme Court and the "descriptors" that Court gives us in instructing us how to conduct the McDonnell Douglas analysis, we have no authority to implement standards in a way that expressly contradicts those "descriptors."

Third , the Majority Opinion misconceives the consequences of the plaintiff's establishment of a prima facie case. The Majority Opinion states that "[a] successful prima facie showing ... entitles the plaintiff to judgment-to victory-if the employer ... doesn't provide a nondiscriminatory explanation for its actions," Maj. Op. at 1222, intimating that we will be overcome by a flood of undeserving victorious employment plaintiffs if we do not rigorously apply the "similarly situated" standard at the prima facie stage. But establishing a prima facie case at summary judgment is not the golden ticket to automatic *1245 recovery that the Majority Opinion indicates.

First, in the real world, if a plaintiff establishes a prima facie case, virtually no employer completely declines at the second stage of McDonnell Douglas to offer a nondiscriminatory reason for its actions. Indeed, that is the point of why the McDonnell Douglas framework is necessary-no employer is likely to admit it took employment action for discriminatory reasons or sit back and do nothing if so accused. So it is not surprising that we appear to have no opinion at all in our Circuit precedent where an employer altogether failed to try to offer a nondiscriminatory reason for its action. 14

Second, even if that unicorn of an employer who offers no nondiscriminatory reason for its actions exists somewhere out there, establishing a prima facie case at summary judgment still does not necessarily automatically end the case. Rather, as the Supreme Court noted in St. Mary's , even if the defendant does not meet its burden of production, "[i]f reasonable minds could differ as to whether a preponderance of the evidence establishes the facts of a prima facie case, then a question of fact does remain, which the trier of fact will be called upon to answer." Id. at 509-10 , 113 S.Ct. 2742 .

So the plaintiff's establishment of an unrebutted prima facie case does not hand her a ticket to skip the trial and go directly to the Clerk's Office to collect a judgment; that happens only when the employer offers no nondiscriminatory reason for its actions and reasonable minds could not disagree about whether the employer intentionally discriminated against the plaintiff. See Reeves v. Sanderson Plumbing Prod., Inc. , 530 U.S. 133 , 146, 120 S.Ct. 2097 , 147 L.Ed.2d 105 (2000) ("[I]n St. Mary's ... we held that the factfinder's rejection of the employer's legitimate, nondiscriminatory reason for its action does not compel judgment for the plaintiff.") (emphasis in original); see also Walker , 158 F.3d at 1184 ("The prima facie case, standing alone, puts the evidence in equipoise-although one could reasonably conclude that the plaintiff was not hired because of her sex, one could just as reasonably conclude that the plaintiff was not hired because the employer did not like the suit she was wearing ....").

Because the Majority Opinion makes these three critical errors in its analysis, it incorrectly concludes that the "similarly situated" standard must be rigorously applied during the prima facie stage of the McDonnell Douglas analysis. In doing so, it disrupts McDonnell Douglas 's delicate balance by playing favorites, elevating the interests of employers while ignoring the interests of employees. As a result, the Majority Opinion's location of the rigorous "similarly situated" application will leave victims of discrimination remediless at the first stage of the McDonnell Douglas framework, denying them the chance to show that the employer's stated reason for *1246 its action may be nothing more than a smokescreen to cover discriminatory intent. Title VII requires more than that.

II.

The Majority Opinion compounds employee-adverse error in how it applies the term "material" in the "similarly situated" standard. Section A below explains the Majority Opinion's mistake, and Section B shows how a correct application of the term to all relevant facts in Lewis's case requires the conclusion that, for purposes of surviving summary judgment, Lewis sufficiently demonstrated she was similarly situated to her two chosen comparators, McClure and Heard.

A.

"Material," of course, means "[b]eing both relevant and consequential; crucial ...." Material , The American Heritage Dictionary (5th ed. 2011). So in the context of the McDonnell Douglas framework, a comparator who is "similarly situated in all material respects" is a comparator who was similarly situated to the employee in all ways necessarily crucial to-but only in those ways necessarily crucial to-the employer's decision to take action against the employee. And if any material issue of fact 15 exists concerning whether a nondiscriminatory factor (or reason) was necessarily crucial to the employer's employment action, it raises a material issue of fact concerning whether the employer's proposed narrower comparator class is actually "similarly situated" to the plaintiff. As a result, the case must survive summary judgment.

To see what this means in practice, we must look to what the Supreme Court has said about comparators and how it has applied the concept. Once again, we begin with McDonnell Douglas . As I have noted, McDonnell Douglas , a failure-to-hire case, discussed the comparator inquiry in terms of "comparable seriousness." It instructed that, at the pretext stage, the district court should consider "evidence that white employees involved in acts against [the company] of comparable seriousness to the 'stall-in' [the plaintiff participated in before being laid off by the company] were nevertheless retained or rehired." Id. at 804, 93 S.Ct. 1817 . In other words, a person need not have engaged in a "stall-in" to be a proper comparator in McDonnell Douglas . Rather, he must have been involved in only acts of "comparable seriousness"-that is, behavior that, when considered by the employer, would necessarily trigger the same employment decision as an employee's participation in a "stall-in" would.

And more recently, in Young v. United Parcel Service, Inc. , --- U.S. ----, 135 S.Ct. 1338 , 191 L.Ed.2d 279 (2015), which the Majority Opinion points out holds that "the plaintiff and her comparators need not be similar in all but the protected ways," Maj. Op. at 1227 (quoting Young , 135 S.Ct. at 1354 ) (internal quotation marks omitted), the Supreme Court further illustrated how similar a proper comparator must be. 16 There, the plaintiffs *1247 sued under the Pregnancy Discrimination Act, alleging that UPS had discriminated against her on the basis of her pregnancy. Id. at 1344 .

UPS required drivers like Young to be able to lift up to 70 pounds. Id. But when Young became pregnant, her doctor told her not to lift more than 20 pounds. Id. So UPS told Young she could not work while she was under a lifting restriction. Id. Young brought suit, arguing that UPS more favorably treated non-pregnant workers "who were similar in their inability to work." Id . (internal quotation marks and alteration omitted). The Supreme Court concluded that, at the prima facie stage, a genuine dispute existed "as to whether UPS provided more favorable treatment to at least some employees whose situation cannot reasonably be distinguished from Young's," so it returned Young's case to the district court for determination of the pretext inquiry. Id. at 1355 (emphasis added). Among those employees Young held up as comparators were the following:

1. union members whose collective-bargaining agreement with UPS promised to provide temporary alternative work assignments if the employees were "unable to perform their normal work assignments due to an on-the-job injury";

2. union members whose collective-bargaining agreement with UPS provided that UPS would "make a good faith effort to comply ... with requests for a reasonable accommodation because of a permanent disability" under the Americans With Disabilities Act;

3. union members whose collective-bargaining agreement stated that UPS would give "inside" jobs to drivers who had lost their Department of Transportation ("DOT") certifications because of a failed medical exam, a lost driver's license, or involvement in a motor vehicle accident;

4. employees who received accommodations, such as a 10-pound lifting limitation) after suffering varied disabilities, such as a foot injury or an arm injury, on the job;

5. employees who received accommodations after injury, where the record was unclear as whether the injury occurred on or off the job, such as a recurring knee injury, an ankle injury, a stroke, or a leg injury;

6. employees who received "inside" jobs after losing their DOT certifications; and

7. employees who received accommodations even though they incurred their disabilities, such as an ankle injury or cancer, off the job.

Id. at 1346-47 .

Young highlights the problems with the Majority Opinion's analysis. First, Young applied a generalized comparator standard at the prima facie stage of the McDonnell Douglas analysis. To be sure, the Court's opinion did not reveal precisely which, if not all, of the widely varied proposed comparators' situations it deemed "[ ]not reasonably ... distinguish[able] from Young's." Id. at 1346-47 . But it nonetheless allowed all of them to proceed to the pretext inquiry. And in doing so, the Court found genuine issues of material fact about whether all of these other employees were proper comparators at the generalized prima facie stage. That means it could not *1248 conclude that the differences between the proposed comparators and Young were necessarily "material" for purposes of the prima facie "similarly situated" inquiry.

So to really understand what the Supreme Court has considered "material" for purposes of the prima facie comparator inquiry, we must look more closely at who the Supreme Court concluded could not necessarily be ruled out as comparators. And when we do that, we find that the Court cast a pretty wide net in performing the comparator inquiry at the prima facie stage.

For example, at this prima facie phase of the inquiry, among others, the Court found the following groups of people appropriate comparators for a pregnant woman who temporarily could not lift more than 20 pounds: (1) non-pregnant employees who could not drive because they literally lacked the qualifications (a license or a DOT certification); (2) non-pregnant employees who could not drive because they had permanent disabilities that made them physically unable to drive; and (3) nonpregnant employees who permanently could not lift 70 pounds. Young , 135 S.Ct. at 1346-47 . That's a broad variety of fish caught in the comparator net. And it contrasts significantly with the Majority Opinion's view of the scope of the "similarly situated" inquiry at the prima facie stage.

Second, although at the prima facie stage the Supreme Court found a material issue of fact concerning whether all of the proposed comparators were sufficiently similarly situated, it nonetheless sent the case to the district court to determine at the pretext phase whether a material question of fact existed regarding UPS's reason for treating pregnant women differently-that is, UPS's reason why it did not view Young's proposed comparators as "similarly situated" to her. So based on this ruling, Young received the opportunity to present evidence and argument showing that UPS's more narrowly proposed comparators class was pretextual. That means that the Supreme Court recognized that employees as differently situated from Young as non-pregnant drivers who lost their DOT certifications or licenses and permanently injured non-pregnant drivers could conceivably be appropriate comparators for the temporarily pregnant Young-even at the pretext phase of the McDonnell Douglas framework.

The Majority Opinion, though it mentions McDonnell Douglas and Young more than once, dismisses these cases' lessons about proper comparators. And it doesn't even try to explain how its interpretation of what the "similarly situated" requirement demands is consistent with the comparators Young and McDonnell Douglas actually allowed. Perhaps that is because it is not. Instead, the Majority Opinion attempts to distinguish Young in two ways: (1) on the basis of the language of the "similarly situated" standard that appears in the Pregnancy Discrimination Act ("PDA"), which governed in Young , and (2) on the basis that the "sheer numbers" of non-pregnant employees UPS accommodated were "overwhelming," Maj. Op. at 1228 n.14, while the Department treated only Heard and McClure differently than Lewis. Id. Neither succeeds.

First, contrary to the Majority Opinion's wishful thinking, the language of the PDA's "similarly situated" requirement does not render Young 's lessons concerning the "similarly situated" requirement inapplicable in the gender- and race-discrimination context. True, the PDA phrases its "similarly situated" requirement differently than we state ours today, since under the PDA, employers must treat "women affected by pregnancy ... the same for all employment-related purposes ... as other persons not so affected but *1249 similar in their ability or inability to work." 42 U.S.C. § 2000e(k). But the PDA's focus on the plaintiff and her comparators' inability to work is precisely what makes Young applicable to Lewis's case. Indeed, Lewis's claim is that she was "similar [to Heard and McClure] in [her] ability or inability to work" when the Department involuntarily put her, McClure, and Heard on administrative leave for being physically unfit for duty. So Young is instructive on this key point in Lewis's case. And even the Majority Opinion concedes Young 's relevance in the context of evaluating comparators for purposes of determining whether they are "similar in their ability or inability to work." See Maj. Op. at 1228 n.14, 1227.

As to the Majority Opinion's second basis for attempting to distinguish Young , it fares no better than the first. The Majority Opinion contends that Young is not instructive because it involved "overwhelming" "numbers" of comparators, whereas Lewis identifies only two comparators. 17 So by the Majority Opinion's reasoning, an employer can discriminate with impunity, as long as the employer is careful to treat fewer than "overwhelming" "numbers" of similarly situated comparators better than the plaintiff.

Clearly, that is unacceptable under the law. And the Majority Opinion's efforts to distinguish Young in a footnote, see Maj. Op. at 1228 n.14, fail miserably.

Because the Majority Opinion does not engage with McDonnell Douglas and Young 's actual applications of the comparator requirement, we are left to discern for ourselves the guiding principles concerning the meaning of "material" that we can take from the analyses of McDonnell Douglas and Young .

We know from McDonnell Douglas 's "comparable seriousness" requirement that appropriate comparators include individuals with circumstances or who engaged in conduct that would necessarily trigger the same employment decision by the employer-at least until and unless, at the pretext stage, the employer can demonstrate a nondiscriminatory reason why such individuals are not appropriate comparators, for which the plaintiff does not raise a material issue of fact.

And Young goes even further. It suggests at the prima facie stage that in a case where a plaintiff seeks a job accommodation, employees whose circumstances cause them to be unable to perform essential aspects of the same job that the plaintiff holds may be appropriate comparators for a plaintiff who cannot perform an essential part of her job-even if the essential function that the employee cannot do or the reason the plaintiff cannot perform an essential function is different than that of the comparators. Of course, at the pretext stage, some or all of these proposed comparators may prove not to be appropriate comparators. But that depends on whether the plaintiff can point to a material issue of fact concerning whether the employer's reason for treating the proposed comparators differently than the plaintiff-raised at McDonnell Douglas 's second stage-is pretextual.

Despite the guidance McDonnell Douglas and Young offer, they cannot provide complete guidance to resolve every case. Indeed, "[t]he facts necessarily will vary in Title VII cases,"

*1250 McDonnell Douglas , 411 U.S. at 802 n.13, 93 S.Ct. 1817 , so "case-by-case" analysis will be required, "in the context of individual circumstances," Maj. Op. at 1227. And unfortunately, neither I nor anyone else I'm aware of can articulate a checklist that will necessarily apply in every single case. Even the Supreme Court itself has recognized this fact. See Swierkiewicz v. Sorema N. A. , 534 U.S. 506 , 506-07, 122 S.Ct. 992 , 152 L.Ed.2d 1 (2002) ("The precise requirements of the prima facie case can vary with the context and were 'never intended to be rigid, mechanized, or ritualistic.' ") (quoting Furnco , 438 U.S. at 577 , 98 S.Ct. 2943 ); see also Sprint/United Mgmt. Co. v. Mendelsohn , 552 U.S. 379 , 387-88, 128 S.Ct. 1140 , 170 L.Ed.2d 1 (2008) ("The question whether evidence of discrimination by other supervisors is relevant in an individual ADEA case is fact based and depends on many factors, including how closely related the evidence is to the plaintiff's circumstances and theory of the case.").

Nevertheless, besides the considerations we can draw from the McDonnell Douglas and Young analyses, factors such as whether the plaintiff and the proposed comparators had the same supervisor or a similar employment or disciplinary history, and whether the proposed comparator and the plaintiff had fairly comparable positions may also be relevant, even at the prima facie stage. For example, an employee whose job it is to drive and an employee whose job does not require any driving may be subject to materially different consequences when it comes to the significance to their employment status of an off-duty car accident, and if that is the case, the driver would obviously know that. Such employees are not "similarly situated" at the prima facie stage.

But the biggest principle we take from Supreme Court precedent on the level of specificity required in applying the "similarly situated" standard is that the standard applies in a "generalized," "not onerous," and "minimal" way at the prima facie stage and in a more rigorous way at the pretext stage. So at the second phase of the McDonnell Douglas framework, an employer's stated reasons for its action can fairly and severely narrow-and even completely eliminate-the field of "similarly situated" comparators. And if they do, that ends the action, unless the plaintiff can raise a material issue of fact concerning whether the employer's stated reason was pretextual.

Instead of following this principle, the Majority Opinion requires courts to perform the heightened "similarly situated" inquiry at the prima facie stage, and it engages in a hyper-detailed application of the standard at that juncture. 18 So legitimate questions about how detailed similarities between a plaintiff and her comparators must be before losing meaning get disposed of without the employer's reasons ever being subjected to pretext testing.

*1251 If the Majority Opinion's analysis were correct, in Young , at the prima facie stage, the Supreme Court would have upheld summary judgment to UPS for all the reasons the Fourth Circuit noted: some of Young's proposed comparators were " 'disabled under the ADA' " whereas Young was not; other comparators were " 'injured on the job,' " whereas she was not. Young , 135 S.Ct. at 1348 (quoting Young v. United Parcel Serv., Inc. , 707 F.3d 437 , 450 (4th Cir. 2013) ). But, of course, that did not happen. Rather, the Supreme Court held that the Fourth Circuit erred and that Young should be allowed to proceed to the pretext stage to present the whole menu of comparators enumerated above. Id. at 1355-56. It is simply impossible to square the Majority Opinion with how the Supreme Court has instructed lower courts to effectuate McDonnell Douglas.

The Majority Opinion makes this error because it allows the matching of characteristics of comparators to become an exercise for its own sake. We must always remember our purpose in identifying "similarly situated" comparators. At the prima facie stage, we match characteristics to determine only whether the plaintiff and her comparators are similar enough that we can confidently rule out discrimination as a basis for the employer's action-without hearing the employer's reason for its action and without allowing the plaintiff an opportunity to cast doubt upon that articulated reason. And at the pretext stage, we evaluate characteristics to ascertain only whether the plaintiff and her comparators are similar enough that we cannot confidently rule out discrimination as a basis for the employer's action-even considering the plaintiff's efforts to show pretext. If we are any more demanding than that, we fail to restrict our comparison to "material" similarities and violate the purpose of Title VII and the McDonnell Douglas framework.

B.

To demonstrate how the "similarly situated" standard should be applied under McDonnell Douglas and its progeny, I apply it in this section to Lewis's facts. Before I can do that, though, I must take a moment to identify some important facts the Majority Opinion leaves out and to correct some inferences the Majority Opinion mistakenly draws from the facts it does mention. For these reasons, Section 1 sets forth the relevant facts, and Section 2 applies the correct "similarly situated" standard to them in conducting the McDonnell Douglas analysis.

1.

a. The Facts Leading to Lewis's Termination

Lewis began working at the Union City Police Department as a police officer in 2001. In 2008, the Department promoted her to the position of detective.

The following year, on January 29, 2009, Lewis suffered a heart attack. Lewis's cardiologist described her heart attack as "relatively small because it hadn't affected her heart function," and she had no blockage. After taking the month of February 2009 off, Lewis returned to full, active duty on March 2, 2009.

When she started back, her lieutenant, Jerry Hester, told her that detectives did not respond directly to calls but that they waited to be called out specifically. Hester assigned "children and women crimes" to the "lady" detectives and gave "the more aggressive stuff" to himself or Sergeant Cliff McClure.

In June 2010, the Department implemented a policy requiring all officers in a position to make an arrest or supervise an arrestee, which included detectives, to carry *1252 a Taser. To become certified to carry a Taser, the Department required, among other things, for officers to receive a five-second Taser shock.

Similarly, pepper-spray training, which was necessary for pepper-spray certification, involved exposure to pepper spray. Though Lewis had been exposed to pepper spray when she was at the police academy, she was not certified in its use.

To remedy this circumstance, on June 14, 2010, the Department scheduled Lewis for pepper-spray training, including pepper-spray exposure, three days later, on June 17, 2010.

The very next day, June 15, 2010, Lewis's doctor gave her a letter that "recommend[ed] that a Taser gun or [pepper] spray [not] be used on or near [Lewis] secondary to her chronic conditions [including a heart condition]." The letter further requested that the Department "[p]lease take [the doctor's recommendation] into consideration when making any decisions about occupational training." Lewis provided the Department with that letter the following day, June 16, 2010.

And the day after that, Thursday, June 17, 2010, Assistant Chief of Police Lee Brown issued a letter to Lewis "plac[ing] [her] on administrative leave without compensation until such time your physician releases you to return to full and active duty." The letter further advised, "You need to contact Tracie McCord in Human Resources to complete the necessary FMLA [Family Medical Leave Act] paperwork concerning your absence. You presently have accrued leave that you may utilize until the time such leave is expended."

Lewis contacted McCord's office the next Monday, June 21, 2010. On June 22, McCord's assistant called Lewis back and scheduled an appointment for Lewis and McCord to meet on Wednesday, June 23, 2010.

At the June 23 meeting, McCord gave Lewis FMLA paperwork and told Lewis to ask her doctor to call Brown to see whether there was any way that Lewis could be safely Tased.

That same day, Lewis called her doctor and left a message seeking to have her doctor schedule a time to call Brown and to fill out the FMLA forms. Lewis's doctor's office returned the call and made a July 7 appointment for Lewis to bring the FMLA paperwork.

In the meantime, on July 1, 2010, Lewis sent Police Chief Charles Odom a letter asking that he permit her to "resume [her] duties as [she] had been doing without limitation until [she] was recently placed on administrative leave." She clarified that she was "only asking for an accommodation on the [T]aser and [pepper-spray] training until everything [wa]s cleared up with [her] doctor." Later that day, Brown responded on behalf of himself and Odom. He stated, "[U]ntil your doctor releases you for duty, your request is denied."

The following day, Friday, July 2, 2010, Lewis emailed Brown to advise him that her doctor was out on vacation until Wednesday, July 7, the date upon which she had already scheduled an appointment with the doctor and would provide her with the FMLA paperwork. Lewis also requested that Brown give her his cellphone number so she could put her doctor directly in touch with Brown.

The next communication in the record occurred after the July 4 holiday, on Tuesday, July 6, 2010. In that email from Lewis to Brown, Lewis sent a "reminder/follow up" to her Friday email to Brown, reminding Brown that her doctor was on vacation until Wednesday and that she had an appointment at noon on Wednesday, so she *1253 could give the doctor the FMLA paperwork. Lewis again requested Brown's cellphone number to provide to her physician so the doctor could call him directly. Later that day, Brown instructed Lewis to have her doctor call the office number to schedule a time to "have a conversation about the situation."

The following day, July 7, 2010, Lewis went to see her doctor and provided the doctor with the FMLA paperwork to fill out. Although the doctor did not prepare the paperwork before Lewis left, the doctor advised Lewis that she was going to contact the Police Department. So Lewis called McCord and left her a message that the doctor would be calling that day. But the doctor did not call, and no one told Lewis.

Instead, the next morning, at 11:15 a.m., Captain Mike Jones and Lieutenant Ken Hester visited Lewis and served her with a letter from Brown immediately terminating her employment. The letter explained,

As of Thursday, July 8, 2010, at 1000 hours, you have not provided the City of Union City with any paperwork requesting Family and Medical Leave. You have not contacted me or any other City official concerning your status in relation to returning to active duty. You are absent without leave.

Union City Personnel Policy Chapter 6, Section 1, c, (Leaves of Absence) states, "Any unapproved leave of absence will be cause for dismissal." Because you have exhausted all of your accrued leave and have failed to complete and turn in the necessary paperwork to be placed on Family and Medical Leave, your absence is unapproved and you are terminated effective immediately.

Nobody told Lewis before her termination that if she did not turn in her FMLA paperwork by July 7, any adverse action would befall her.

After receiving the termination letter, Lewis called her doctor to ask whether the doctor had called the Police Department, as the two had agreed the doctor would. The doctor's assistant said that the doctor intended to call the Department on her lunch hour that day. So Lewis emphasized the importance, explaining that she had been fired from her position.

Lewis's doctor then called Odom that same day. Odom told the doctor "he was concerned about [Lewis's] being released back to full duty and still being under [the doctor's] care." The doctor responded that she "release[s] people back to work duty and still care[s] [for] them in the office." Odom "also wondered if ... Lewis influenced [the doctor's] decision to recommend that [Lewis] not use the Taser or pepper spray." The doctor replied that "pepper spray was not as much [of] a concern" as "be[ing]" Tased. As the doctor explained the problem, the doctor "did not know how the 'Taser' would affect Lewis."

In the course of the litigation in this case, Odom later conceded in his deposition testimony that if Lewis had brought in a document from her doctor stating that she could be exposed to pepper spray, "she probably would have been put back to work ...." He further admitted that during the July 8 conversation between Brown and Lewis's doctor, in light of the doctor's expressed withdrawal of the pepper-spray concerns, nothing precluded Brown from asking the doctor for a note stating that Lewis could proceed with pepper-spray training. Yet Brown did not do so. Odom explained this fact by noting that the conversation did not occur until after Lewis's termination earlier that same day.

Also on July 8, 2010, Lewis notified Odom of her intent to appeal her termination.

*1254 The City Manager affirmed Lewis's termination on July 23, 2010.

b. The Proposed Comparators

As comparators, Lewis proposed Walker Heard, a white male who served as a master patrol officer, and Cliff McClure, a white male who worked as a sergeant in the Criminal Investigations Division.

Heard was placed on administrative leave after failing a physical-agility test, since he was not fit for duty. He was given 90 days of administrative leave within which to work with medical professionals and pass the fitness-for-duty test. But instead of Heard passing the test, near the end of the 90-day period, Heard's attorney sent a letter to the Department stating that Heard had a disability and requesting that the fitness-for-duty test be waived.

Although Odom did not waive the test, he also did not terminate Heard's employment at the end of the 90-day administrative-leave period. Rather, Odom offered Heard the chance to transfer to a dispatcher position, which did not require him to pass the fitness-for-duty test that patrol officers had to take. Odom held the dispatcher position open for Heard for eleven months before demanding a decision and finally terminating him when he declined the transfer. In all, Heard was on administrative leave for 449 days before he was finally fired.

The Department placed McClure on administrative leave because he failed a balance test, so he was rendered unfit for duty. McClure received up to a 90-day period of administrative leave within which to work with medical professionals to remedy his balance test and requalify as fit for duty. Before the 90-day period expired, McClure successfully completed his physical-fitness test and was deemed fit to return to duty.

c. The Policies the Department Invoked

The Department relies on portions of two policies in this case. I summarize each in turn.

i. Union City Employee Handbook: Personnel Policy Chapter 6 (Leaves of Absence)

Odom testified that the Department placed Lewis on leave under the Union City Employee Handbook's Personnel Policy on leaves of absence, which is found at Chapter 6.

Section 1.A. of that policy provides, in relevant part, "A regular full-time employee may be granted leave of absence without pay at the discretion of the Department Director and the City Manager. An employee may also be placed on leave of absence status without application." As relevant here, § 1.A. imposes a 180-day limit on any such leave. Finally, if an employee is placed on administrative leave without pay, she may choose to take any portion of it as paid leave to the extent she has paid leave available. But whether or not she chooses to take any of her unpaid administrative leave as available paid leave is "up to the individual employee," since the option to use paid leave during unpaid administrative leave is for the benefit of the employee. As a result, according to the City Manager, who was responsible for reviewing and revising the Union City Employee Handbook, whether the employee opts to use paid leave for any portion of her unpaid administrative leave is "irrelevant" to and does not affect the length of the administrative-leave period.

Turning to § 7 of the City's leave-of-absence policy, that covers FMLA absences. That section has a subsection called "Notification and Reporting Requirements," § 7.F. But absent from that section is any deadline for the submission *1255 of FMLA paperwork where, as in this case, the employee is put on administrative leave by the City and told to subsequently "complete the necessary FMLA paperwork concerning [the] absence" (as opposed to scheduling FMLA leave herself). Section 7.B. also authorizes "[a]n eligible employee [to] apply available accrued leave to all approved absences where all or a portion of their normal base salary would otherwise be unpaid."

The last part of Chapter 6 that is pertinent is § 1.C., which the Department relied on in the letter terminating Lewis's employment. That provision states, "Any unapproved leave of absence will be cause for dismissal."

ii. The Union City Police Department's Physical-Fitness Policy

In January 2013, after Lewis left but before Heard and McClure were placed on administrative leave, the Department created a policy requiring all sworn officers to pass an annual physical-fitness test. 19 Under the policy, any officer who did not pass the annual fitness test was sent to physicians for a less strenuous fitness-for-duty test. If the officer failed that test, the officer was placed on administrative leave without pay for 90 days. The officer then had that 90-day period to improve his or her health to the point where the officer could pass the fitness-for-duty test. At the end of the 90-day period, physicians administered a second fitness-for-duty test. If the officer passed, the officer was reinstated. But if the officer failed, the officer was "deemed unfit for duty and [was] subject to termination."

The "governing force" for the Department's authority to involuntarily place officers on administrative leave came from the Union City Personnel Policy Handbook, which governed all City employees. The Handbook "set the baseline" for the Department. As a result, the Department generally could not exceed its proscriptions, since the Handbook and the Department's policies are "[v]ery similar" to the relationship between "State laws" and "County laws."

The Department's issuance of the physical-fitness policy's 90-day administrative-leave provision and the Department's placing of Lewis on administrative leave both involved the Department's exercise of its authority to involuntarily put an officer on administrative leave for being physically unfit for duty. Viewed in the light most favorable to Lewis, the record yields the reasonable inference that the Department's authority to involuntarily place Lewis on administrative leave for being physically unfit for duty and the Department's decision to put physically unfit officers involuntarily on 90 days of administrative leave were exercises of the same power: the authority referenced in Chapter 6, § 1.A., in the Union City Personnel Policy Handbook for the Department to put an employee involuntarily on administrative leave for up to 180 days. In fact, both exercises were "governed" by the Union City Personnel Policy Handbook since the Handbook put a 180-day restrictor cap on the Department's discretion.

This is unsurprising since both cases involve the same action-placing an officer involuntarily on administrative leave-for the same reason-being physically unfit for duty. Plus, both exercises fall within the policy's boundaries requiring fewer than 180 days' leave. Nor did the Department put the physical-fitness policy or any *1256 other evidence suggesting that anything other than the discretion referenced in the Personnel Policy Handbook was the source of its power to put officers involuntarily on administrative leave for 90 days for being physically unfit for duty into the summary judgment record. Rather, the record contains only one source of authority for the Department to put someone involuntarily on administrative leave: Chapter 6 of the Personnel Policy Handbook. 20

Odom attested that both Heard and McClure were placed on administrative leave pursuant to the 90-day administrative-leave portion of the physical-fitness policy. 21 While McClure was able to pass the second fitness-for-duty test and return to work, Heard was not. But as explained in Section II.B.1.b above, Heard's employment was not terminated at the end of the 90-day period. Instead, he remained on administrative leave for a total of 449 days before Odom ended his employment.

2.

Lewis satisfied her comparator burden at both the prima facie stage and the pretext stage, allowing her to survive summary judgment. As I explain below, material questions of fact exist about whether the Department's stated nondiscriminatory reasons were necessarily crucial to the Department's dissimilar treatment of Lewis and her comparators.

Significantly, Heard and McClure's situations cannot "reasonably be distinguished" from Lewis's. Young , 135 S.Ct. at 1355 ; see also Caraballo-Caraballo , 892 F.3d at 60 (stating that in the "in all relevant respects" standard, "reasonableness is the touchstone" and "[t]he court must decide whether a prudent person, looking objectively at the plaintiff and her comparator would think them roughly equivalent") (internal citations and quotation marks omitted). Lewis and her comparators had similar job responsibilities, had similar job titles, had similar seniority, and were supervised by the same person, Odom. And most important, the Department put all three on involuntary administrative leave for being physically unfit for duty. Yet Lewis presented evidence that the Department gave Lewis just 21 days of administrative leave while affording McClure and Heard 90 and 449 days of administrative leave, respectively.

This is plenty to meet Lewis's "generalized" and "minimal" prima facie burden. To put it in McDonnell Douglas 's "comparable *1257 seriousness" terms, see 411 U.S. at 804 , 93 S.Ct. 1817 , Lewis's physical condition triggered the same employment decision by the Department as did McClure's and Heard's-the Department placed all three involuntarily on administrative leave because they were not fit for duty and indicated that they could not return to duty until they were fit to do so.

And if we compare Lewis's situation to that of Young in Young , Lewis bears a closer resemblance to her proposed comparators than Young did to her accepted comparators. As I have noted, in Lewis's case, she and her comparators all could not do what the Department deemed an essential aspect of their work for the same reason-they were all physically unfit for duty. But in Young , where the plaintiff could not deliver packages because she temporarily had a physical restriction concerning how much weight she could lift, some of her approved comparators could not work as drivers for reasons unrelated to their health-they had been stripped of their licenses or DOT certifications.

In light of Lewis's similarities to her comparators, at the prima facie stage, we cannot confidently rule out discrimination as a basis for the employer's action, without hearing the employer's reason for its action and without allowing the plaintiff an opportunity to cast doubt upon that articulated reason.

Since these facts establish a prima facie case for Lewis, we turn to the Department's nondiscriminatory reasons for firing Lewis after giving her only 21 days of administrative leave. The Department has offered four reasons for its action, and Lewis asserts each is pretextual. Before I address each reason in turn, I note a few ways in which a plaintiff may create a material issue of fact concerning pretext.

Among others, a plaintiff can survive the pretext stage by casting sufficient doubt on the defendant's proffered nondiscriminatory reasons to permit a reasonable factfinder to conclude that the employer's proffered "reasons were not what actually motivated its conduct." Cooper-Houston v. Southern Ry. Co. , 37 F.3d 603 , 605 (11th Cir. 1994) (citation omitted). The plaintiff can also demonstrate pretext if she shows the employer's articulated reason is false and that the false reason hid discrimination. See St. Mary's , 509 U.S. at 515 , 113 S.Ct. 2742 ; Hurlbert v. St. Mary's Health Care Sys., Inc. , 439 F.3d 1286 , 1298 (11th Cir. 2006) (citing as evidence of pretext "an employer's failure to articulate clearly and consistently the reason for an employee's discharge"). And the plaintiff can show pretext if she establishes that the employer has failed to clearly articulate and follow its formal policies. See Hinson v. Clinch Cty., Ga. Bd. of Educ. , 231 F.3d 821 , 831 (11th Cir. 2000) (finding issue of fact on pretext reversing because employer failed to raise a concern directly with the employee). All of these ways of demonstrating pretext come into play in this case.

First, the Department asserts that it provided the comparators with more than 21 days' administrative leave under the Department's physical-fitness policy, which did not exist when Lewis worked there. And to be sure, the physical-fitness policy had not yet been issued when Lewis was fired.

But viewing the facts in the light most favorable to Lewis, as we must, the physical-fitness policy is nothing more than a statement of how the Department exercises its authority as identified in Chapter 6, section 1.A. of the City Employee Handbook, to place an officer on administrative leave when that officer is not physically fit to perform her duties. So the record shows that Odom had just two opportunities to use the discretion afforded to him by *1258 Chapter 6, section 1.A. of the City Employee Handbook to promulgate administrative-leave policies for officers who were physically incapable of policing. He couldn't have behaved more differently.

First, when Odom had Lewis, an African-American woman before him, he, at best, gave her 21 days of administrative leave before firing her. But when he exercised his discretion to grant leave without an African-American woman in front of him, he provided officers with 90 days of administrative leave-more than four times as much as he gave Lewis.

Second, when Odom had Lewis, an African-American woman to consider, he gave no warning before firing her after 21 days of administrative leave. But under the policy Odom promulgated after Lewis's departure to apply generally to all officers who were physically unfit for duty, he provided an express warning that their leave would expire in 90 days if they did not remedy their fitness-for-duty deficiency.

And third, in Heard's case, Odom skated right by Chapter 6, § 1.A's 180-day cap and gave Heard 449 days of unpaid administrative leave-21 times the number of days he gave Lewis and 269 days more than the City's administrative-leave policy allowed (and 359 days more than the Department's physical-fitness policy permitted). A reasonable jury could find that the Department did not consistently exercise its authority in placing physically unfit officers on administrative leave and that the Department did not comply with its own policies. A reasonable jury could further conclude that these inconsistencies reveal discriminatory motivation.

The Department offers as a second nondiscriminatory reason for its actions that Lewis's condition was permanent, while Heard's and McClure's were not necessarily so. But this reason also has three holes in it.

First, as I have noted, in Young , the Supreme Court concluded that temporarily disabled employees could be proper comparators for permanently disabled employees. Young , 135 S.Ct. at 1346-47 . We have no authority to demand more similarity than the Supreme Court. And the Majority Opinion does not say why the permanently disabled were acceptable comparators to the temporary disabled in Young , yet they are not here.

Second, even if we could set aside Young , in the light most favorable to Lewis, the record shows the Department itself did not appear to believe that Lewis's lack of physical fitness for duty was permanent. For example, in the June 17, 2010, letter putting Lewis on administrative leave, Brown stated that Lewis was on such leave "until such time your physician releases you to return to full and active duty." And Lewis also advised the Department that she did not expect her lack of physical fitness to be permanent, asking the Department on July 1, 2010, for "an accommodation ... until everything is cleared up with my doctor." Even in the letter firing Lewis, the Department said it placed Lewis on administrative leave "until [her] physician released [her] to return to full and active duty."

Plus, the Department had a history of working with others with heart conditions to allow them to receive a milder version of a Tasing than officers without heart conditions received. 22 So the Department had firsthand knowledge and experience that a heart condition did not necessarily mean that Lewis would never be fit for *1259 duty. And Lewis was in daily contact with the Department about the possibility of having her doctor see if something could be worked out with Brown to resolve the Tasing and pepper spray issues. 23 In the light most favorable to Lewis, these facts demonstrate that both parties viewed Lewis's return to fitness for duty to be a possibility. 24 So a reasonable jury could conclude that the alleged permanence of Lewis's condition was not the true reason for the Department's actions and rather, that excuse hid discriminatory motivation.

And third, even assuming, arguendo , the Department considered Lewis's physical unfitness for duty to be permanent-a conclusion that is not supported by the record, when Heard's unfitness for duty was determined not to be temporary, the Department offered him the opportunity to take a job with the Department that did not require him to go out on patrol. Odom offered no such alternative to Lewis before or after firing her, even though he admitted that the Department had enough information on the same day Lewis was fired to offer her a different position that would not have required her to patrol the streets. So a reasonable jury could conclude that the Department did not consistently apply its policies and that it failed to do so for discriminatory reasons.

The Department offered yet another (a third) reason for firing Lewis at the time it terminated her employment-that she was "absent without leave" because her paid leave supposedly expired. But a reasonable jury could find that that reason has no support in the record. Rather, on June 17, 2010, the Department placed Lewis on "administrative leave without compensation until such time your physician releases you." Lewis remained on that leave until the moment the Department fired her. And according to the City Manager, *1260 whether she had or chose to take paid leave for any portion of her otherwise-unpaid administrative-leave period was "irrelevant" to the administrative-leave status on which the Department placed her. So contrary to the Department's stated reason, a reasonable jury could ascertain that Lewis was not "absent without leave" when the Department fired her; she was on the very administrative leave the Department forced her to take. And based on that, a reasonable jury could conclude that the Department's third stated reason was pretext for discrimination.

Closely related to the Department's contention that Lewis was absent without leave is its fourth stated reason for firing Lewis: that Lewis missed some unspoken deadline by which to file her FMLA paperwork. But that's just it: no deadline was ever stated-not in the daily communications between the parties, not in the City's Employee Handbook, and not anywhere else in the record. Plus, under the terms of the June 17, 2010, letter, everyone knew Lewis would take FMLA leave after she exhausted her paid leave, 25 and it's not like Lewis disappeared from the Department's radar screen without a trace. On the contrary, Lewis was in regular communication with the Department, informing it on a daily basis of her progress in obtaining the FMLA paperwork from her doctor. A reasonable jury could infer that the Department's sudden imposition of an apparently previously non-existent deadline for submitting FMLA paperwork suggests a cover for discrimination. 26

Indeed, the Department's failure to warn Lewis that once her accrued leave expired she would be terminated rather than placed on unpaid-leave status is just the sort of arbitrary action that Justice Rehnquist noted could be probative of discriminatory intent. See Furnco , 438 U.S. at 577 , 98 S.Ct. 2943 ("[W]e know from our experience that more often than not people do not act in a totally arbitrary manner, without any underlying reasons, especially in a business setting. Thus, when all legitimate reasons for [the adverse employment action] have been eliminated as possible reasons for the employer's actions, it is more likely than not [that] the employer, who we generally assume acts only with some reason, based his decision on an impermissible consideration such as race.").

So in sum, Lewis established both at the prima facie and pretext stages that, at the very least, material questions of fact exist concerning whether she was "similarly situated" to Heard and McClure in all "material" respects. But the Majority Opinion engages in a rigorous application of the "similarly situated" standard at the prima facie case, and it construes the term "material" to include immaterial comparisons. So Lewis's case, which should survive summary judgment under the McDonnell Douglas framework, dies at the prima facie stage. And Lewis does not even get the chance to contest the Department's reasons for terminating her employment. That's unfortunate since the City's reasons for its action are, when viewed in the light most favorable to Lewis, suspect.

*1261 III.

In sum, I respectfully dissent from the Majority's decisions to apply a rigorous "similarly situated" standard at the prima facie stage and to overbroadly construe the term "material" within the "similarly situated" standard. These errors, which contravene Title VII's purpose, are particularly troubling in light of recent studies on employment discrimination. Even though Title VII has been at work for close to 45 years, we cannot, unfortunately, say that discrimination has been eradicated. To the contrary, a recent study in the Harvard Business Review found that hiring discrimination against African-Americans and Latino-Americans has not declined at all in the last 25 years. See Lincoln Quillian, et. al, Hiring Discrimination Against Black American Hasn't Decline in 25 Years , Harvard Business Review, Oct. 11, 2017, available at https://hbr.org/2017/10/hiring-discrimination-against-black-americans-hasnt-declined-in-25-years . Another study found that Asian-Americans and African-Americans were twice as likely to receive interviews if they scrubbed all racial cues from their résumés. Dina Gerdeman, Minorities Who 'Whiten' Job Resumes Get More Interviews , Harvard Business School, May 17, 2017, available at https://hbswk.hbs.edu/item/minorities-who-whiten-job-resumes-get-more-interviews .

And a recent study from Pew Research concluded that four in ten women have reported that their employers discriminated against them because of their gender. See Kim Parker & Cary Funk, Gender Discrimination Comes In Many Forms for Today's Working Women , Pew Research, Dec. 14, 2017, available at http://www.pewresearch.org/fact-tank/2017/12/14/gender-discrimination-comes-in-many-forms-for-todays-working-women/. 27 Plus, women are "still systematically sideline[d]" once they become pregnant, seeing their hourly wages chopped by four percent. Natalie Kitroeff and Jessica Silver-Greenberg, "Pregnancy Discrimination Is Rampant Inside America's Biggest Companies," The New York Times , Feb. 8, 2019, available at https://www.nytimes.com/interactive/2018/06/15/business/pregnancy-discrimination.html .

Of course, employers usually don't post help-wanted signs reading "blacks need not apply," and they are generally astute enough not to ask women about plans to start a family. Instead, discrimination today often surreptitiously sits behind a veil of subtlety, with the boss handing out the plum assignments to male officers while relegating the "lady" detectives to "less aggressive" "children crimes." See supra at 1237. Using codewords or subtlety does not make discrimination any less of a problem under Title VII. And because Title VII tolerates none of it, we should be particularly cautious before ratcheting up plaintiffs' "not onerous" prima facie burden, in violation of McDonnell Douglas and its progeny.

I respectfully dissent from the Majority Opinion's errors.