Plaintiff Barbara Good appeals from the district court’s grant of summary judgment in favor of defendant University of Chicago Medical Center in Good’s reverse race discrimination case. Good was employed in UCMC’s Radiology Department as a lead technologist in the Computerized Tomography Department. She admits that there were issues with her job performance, but she contends that UCMC discriminated against her on account of her race (white) when it terminated her employment rather than demoting her as it had some employees of other races. Good sued UCMC under Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., and 42 U.S.C. § 1981.
An employee alleging racial discrimination under these statutes may elect to proceed via either the indirect or the di*673rect methods of proof, or a combination of the two. See Egonmwan v. Cook County Sheriffs Dep’t, 602 F.3d 845, 849-50 & n. 7 (7th Cir.2010). Good invoked both methods. The district court found that she failed to present sufficient evidence to withstand UCMC’s motion for summary judgment using either method. Good has appealed, and we affirm.
I. Standard of Review and Factual Background
We review the district court’s grant of summary judgment de novo, construing all facts and drawing all reasonable inferences in favor of Good, the non-moving party. See Winsley v. Cook County, 563 F.3d 598, 602 (7th Cir.2009). Summary judgment is appropriate if the evidence demonstrates that there is no genuine issue of material fact and the moving party is entitled to judgment as a matter of law. Fed.R.Civ.P. 56(a). In keeping with these standards, the following facts are set forth in a light most favorable to Good, as the non-moving party. We do not vouch for their truth in any other sense.
The Radiology Department of UCMC is composed of the Computerized Tomography, or “CT” Department, where Good worked, as well as the Magnetic Resonance Imaging Department, the Ultrasound Department, and the Nuclear Medicine/Positron Emission Technology Department. Each department employed staff technologists to perform imaging scans and who worked on either the first, second, or third shift. Each shift of each department had a designated lead technologist. Staff technologists and lead technologists were biweekly employees. Each department also had a manager, who was a monthly employee.
Each year, the department managers evaluated their biweekly lead and staff technologists and generated annual performance reviews. Managers evaluated their employees’ performance in several categories, each on a scale of 1 to 5, with 5 being the highest. If an employee’s overall annual performance review was less than 3, UCMC would place the employee on a Performance Improvement Plan or PIP. The PIP would specify areas of improvement, measurable expectations, and consequences of an employee’s failure to improve in the targeted areas within a designated timeframe of either 30, 60, or 90 days.
UCMC maintained a four-step corrective Progressive Action' Policy that provided that an employee’s failure to complete a PIP was grounds for termination. But according to UCMC’s Policy and Procedure Manual, “it is the policy of the University of Chicago Hospitals to demote [an] individual ]” who “cannot perform ... her assigned job responsibilities” because “her skills are not matched to the requirements of the job” or she “lack[s] ... motivation to perform up to standards.”
UCMC Radiology Department managers were to be held to “a higher standard of performance, due to their added responsibilities.” Managerial employees could be terminated at any time, and the record adequately demonstrates that, like biweekly employees, UCMC subjected its managers to annual performance reviews and dealt with some managers’ performance deficiencies by demoting them in lieu of harsher corrective treatment such as a PIP or a probationary period, or even termination.
Good was hired as a lead technologist in the CT Department in May 1994. She resigned in 1999 to take another position, but UCMC rehired her three months later as a staff technologist. In 2004, she was promoted back to lead technologist, and she was assigned to the second shift. In 2005, Cliff Sissel became CT Manager and *674Good’s immediate supervisor. In April 2006, Monica Geyer became the Assistant Director of Specialty Imaging Services. Ed Smith was the Executive Director of Radiology. Like plaintiff Good, both Sissel and Geyer are white. Smith’s race is hot disclosed by the record.
In July 2007, Sissel reviewed Good’s performance for the year ending June 30, 2007 and gave her an overall rating of 2.65. Good did not dispute Sissel’s evaluation. Because her overall score was below 3, Sissel and Geyer developed a 90-day PIP designed to improve Good’s performance. Pursuant to the PIP, Good needed to improve in three. areas: (1) timely patient service; (2) improvement in staff efficiency; and (3) minimizing staff overtime in her department. When she received the PIP, Good told Sissel and Geyer that she “would be ... happy to step down to a staff tech position.” Geyer responded, “That’s a possibility. We might think about that.”
Over the course of the next 90 days, Good failed to improve sufficiently. Sissel discussed these issues with Good in August and again in September 2007, but on October 12, 2007, Geyer gave Good a Final Written Warning, put her on a 30-day PIP, and transferred her to the third shift, which was less busy than the second shift. The warning stated that Good had not met “the majority of the goals [of the 90-day PIP] impacting patient care.” The warning also stated that “this is UCMC’s final effort to work with [Good] to bring her performance to an acceptable level. If [Good] fails to meet these new goals on a less busy shift, further corrective action may be taken up to and including termination of employment.” Good again asked to be demoted to a staff technologist position, and Geyer told her, “we’re thinking about it.”
Sissel and Geyer expected that Good would properly handle inpatient scan orders, properly handle the timing of emergency scans, and properly maintain the CT Department work area. Good did not contest the 30-day PIP or her need to improve in those areas. In spite of these corrective efforts, however, Good’s performance did not improve.
In late October or early November 2007, Good again asked Sissel and Geyer to give her a demotion. Geyer told her that UCMC had “changed [its] policies” and had “decided not to do that anymore.” Contrary to Geyer’s statement, UCMC had not amended the demotion policy and it was still in force. On November 2, 2007, Ed Smith (Director of Radiology and Geyer’s immediate supervisor) sent Geyer an e-mail, instructing: “No more e-mails about [Good],” and “have her removed.” Geyer understood Smith to be instructing her and Sissel to terminate Good’s employment. Accordingly, on November 16, 2007, Geyer sent an e-mail to Employee/Labor Relations at UCMC stating that Good had failed to successfully complete her 30-day PIP and recommending that UCMC terminate Good’s employment. Sissel agreed with this decision. On November 27, 2007, UCMC terminated Good’s employment. UCMC replaced Good with Kristin Runion, who is also white.
II. Direct Method of Proof
To withstand UCMC’s motion for summary judgment under the direct method of proof, Good must present “direct or circumstantial evidence that creates a convincing mosaic of discrimination on the basis of race.” Winsley v. Cook County, 563 F.3d at 604 (quotation marks omitted). Good has no direct -evidence that race played any role at all in UCMC’s treatment of her. She relies on circumstantial *675evidence, which we have said typically falls into one of three categories:
(1) suspicious timing, ambiguous oral or written statements, or behavior toward or comments directed at other employees in the protected group; (2) evidence, whether or not rigorously statistical, that similarly situated employees outside the protected class received systematically better treatment; [or] (3) evidence that the employee was qualified for the job in question but was passed over in favor of a person outside the protected class and the employer’s reason is a pretext for discrimination.
Darchak v. City of Chicago Board of Educ., 580 F.3d 622, 631 (7th Cir.2009). A plaintiff may survive a motion for summary judgment based only on circumstantial evidence under the direct method, but only if the circumstantial evidence presented points “directly to a discriminatory reason for the employer’s action.” Cerutti v. BASF Corp., 349 F.3d 1055, 1061 (7th Cir.2003), quoting Adams v. Wal-Mart Stores, Inc., 324 F.3d 935, 939 (7th Cir.2003); see also Lim v. Trustees of Indiana Univ., 297 F.3d 575, 580 (7th Cir.2002) (requiring “direct evidence” to “prove the particular fact in question without reliance upon inference or presumption”) (internal quotation and emphasis omitted).
To satisfy this burden, Good points to three UCMC employees of different races or ethnicities from hers who were allowed to take demotions from their positions rather than receive corrective discipline for their deficient performance. She also argues that UCMC provided inconsistent reasons for departing from its demotion policy in her case. This, she argues, is sufficient circumstantial evidence to satisfy the direct method of proof. We disagree. Simply stated, the circumstantial evidence on which Good relies is insufficient because it does not point to a discriminatory reason for UCMC’s decision to end her employment rather than demoting her as she would have wished. From this evidence, one might guess or speculate that perhaps Good’s race might have made a difference in the decision, but guesswork and speculation are not enough to avoid summary judgment.
To determine whether a plaintiffs co-worker was similarly situated for purposes of this analysis, a court must make a “flexible, common-sense” evaluation of the relevant factors. Henry v. Jones, 507 F.3d 558, 564 (7th Cir.2007). “All things being equal, if an employer takes an action against one employee in a protected class but not another outside that class, one can infer discrimination. The ‘similarly situated’ prong establishes whether all things are in fact equal.” Filar v. Board of Educ. of City of Chicago, 526 F.3d 1054, 1061 (7th Cir.2008) (internal citation omitted). The purpose is to eliminate other possible explanatory variables, “such as differing roles, performance histories, or decision-making personnel, which helps isolate the critical independent variable” — discriminatory animus. Humphries v. CBOCS West, Inc., 474 F.3d 387, 405 (7th Cir.2007), aff'd, 553 U.S. 442, 128 S.Ct. 1951, 170 L.Ed.2d 864 (2008). Similarly situated employees “must be ‘directly comparable’ to the plaintiff ‘in all material respects,’ ” but they need not have identical employment files. Patterson v. Indiana Newspapers, Inc., 589 F.3d 357, 365-66 (7th Cir.2009), quoting Raymond v. Ameritech Corp., 442 F.3d 600, 610 (7th Cir.2006). So long as the distinctions between the plaintiff and the proposed comparators are not “so significant that they render the comparison effectively useless,” the similarly situated requirement is satisfied. Humphries, 474 F.3d at 405. Which factors are material is a case-specific inquiry that depends on the specifics of the *676defendant’s decision and the stated reason for it. See Coleman v. Donahoe, 667 F.3d 835, 847-52 (7th Cir.2012) (finding that coworkers were sufficiently similarly situated for meaningful comparison in spite of having different immediate supervisors, different job titles, and different duties); Crawford v. Indiana Harbor Belt Railroad Co., 461 F.3d 844, 846 (7th Cir.2006) (the question is whether “members of the comparison group are sufficiently comparable to [the plaintiff] to suggest that [the plaintiff] was singled out for worse treatment”).
Of the three employees that Good put forward, the district court found that only one, an African American named Balderos-Mason, could be considered similarly situated to Good. Like Good, BalderosMason was a lead technologist. The performance histories of Good and BalderosMason suggest that they had similar deficiencies in performance, and Geyer was involved in their disciplinary actions. But, unlike Good, Balderos-Mason was not terminated. Instead, approximately one year before Good was terminated, BalderosMason was demoted to a staff technologist position. Balderos-Mason was not put on a PIP before she was demoted, but the evidence suggests that Balderos-Mason chose to be demoted rather than be put on a PIP. We agree with the district court’s analysis and consider Balderos-Mason to be similarly situated to Good.
However, the district court determined that two other non-white employees whom Good proposed as comparators were not sufficiently similar to Good to offer meaningful comparison. Here, our analysis departs from the district court. UCMC does not contest that these employees’ performance deficiencies were sufficiently similar to Good’s to be comparable, but instead attempts to distinguish them on the basis of their status as managers and the decision-makers involved in the respective employment decisions. True, these two employees were monthly, managerial employees and not biweekly supervisory employees, as Good had been. UCMC’s policies provided that these managers could be terminated at any time and were to be held “to a higher standard of performance.” Neither manager was on a PIP at the time of his demotion, and neither manager reported to Good’s supervisors, Geyer or Sissel.
But there was a common decision-maker involved — Ed Smith, Director of Radiology. Smith was involved in the decisions to demote the two managers in light of their undisputed performance issues, and he prompted Good’s termination with his instruction to Geyer to “have her removed.” In short, the two managers’ status as managerial, monthly employees does not eliminate them as comparators. Given their similar performance deficiencies, the fact that they were held to a “higher standard” than Good, and the fact that a common decision-maker decided to demote them but decided to terminate Good, the two managers offered meaningful comparison, at least for purposes of summary judgment.
But we agree with the district court’s ultimate conclusion that UCMC’s demotions of these non-white employees were insufficient circumstantial evidence under the direct method of proof, which requires evidence leading directly to the conclusion that an employer was illegally motivated, without reliance on speculation. See Cerutti, 349 F.3d at 1061; Adams, 324 F.3d at 939; Lim, 297 F.3d at 580; see also Coleman, 667 F.3d at 863 (Wood, J., concurring) (“Like a group of Mesopotamian scholars, [judges] work hard to see if a ‘convincing mosaic’ can be assembled that would point to the equivalent of the blatantly discriminatory statement.”). It can be a high threshold, particularly in a *677reverse discrimination case. Good’s evidence of similarly situated employees of different races, even expanded to accommodate three comparable non-white employees instead of only one, does not rise to this level. This record simply does not contain a hint of race-based animus. A jury could not reasonably conclude that Good was treated differently because of her race without relying on speculation.
Good attempts to bridge this gap by arguing that UCMC deviated from its demotion policy and that it gave “shifting” reasons for her termination. But she offers nothing to dispute UCMC’s evidence that the demotion policy did not give every employee an absolute right to be demoted. Without some evidence from which we could reasonably infer that UCMC exercised its discretion to terminate Good rather than demote her based on her race, the fact that it deviated from a highly discretionary demotion policy, standing alone, is not probative of improper motivation.
Good further argues that UCMC’s reasons for her termination “shifted,” but here again we disagree that the record supports that conclusion. Geyer told Good that she could not be demoted because UCMC’s policy had changed and it was no longer demoting employees. Good has contradicted that point by offering evidence that UCMC’s demotion policy had not changed. Perhaps Geyer was misinformed, or perhaps she attempted to mislead Good. In either case, though, Good fails to explain how Geyer’s statement sheds any light on the motives underlying Good’s termination — particularly on this record, which shows that Smith, and not Geyer, triggered Good’s termination and that Good was replaced by another white supervisor.
Although UCMC has used different words at different times to describe Good’s performance issues, we also disagree with Good’s argument that UCMC’s semantics could hide an improper motive. Good’s June 2007 performance review stated:
Things that are holding [Good] back are the communication she has with her shift, fair work distribution and controlling the schedule____[She] also needs to work on pre-scheduling transports and coordinating exams with the nursing units to minimize delays. This also means communicating in advance to the staff about what is pending and when it is expected to be in route to the department. Staying positive and keeping staff informed will help her improve in her role.
Good’s 30-day PIP reiterated these concerns. Finally, in Geyer’s e-mail to Employee/Labor Relations in which she carried out Smith’s instruction to have Good terminated, Geyer wrote:
1. Staff want to be transferred off of [Good’s] shift.
2. [She] does not communicate well with the staff.
3. [She] gets short tempered with the staff when they remind her of things that need to be done.
4. [She] does not provide direction and does not assist the staff in ensuring smooth workflow.
5. [She] does not provide the staff with complete sets of paperwork prior to patient arrival, which delays patient care.
6.....[She] demonstrates no sense of urgency in getting the patients completed in a timely manner.
7. There is no sense of team between [her] and the technologists.
8. [She] is not providing training to the new technologist on the shift.
So while [Good] has met some of the measurable goals, the larger picture indicates that [Good] does not have the *678leadership skills needed to run an efficient, productive, and happy shift.
It is our recommendation that we terminate [Good] at this time, since she has not improved her leadership skills despite concerted mentoring and coaching.
By comparison, in its brief before the district court, UCMC justified its decision to terminate Good based on its record of Good’s “poor attitude.” Good argues that these are different reasons, but the phrase “poor attitude” is not inappropriate or inconsistent shorthand for the performance issues UCMC had thoroughly documented before and at the time of Good’s termination. Although UCMC did not use identical language to describe Good’s deficiencies, a reasonable jury viewing this record could not find that UCMC’s rationale for Good’s termination shifted or changed over time in such a way that could suggest that its decision was actually motivated by her race.
In sum, Good has presented evidence that three similarly situated, non-white coworkers received better treatment when they were permitted to take demotions from their managerial roles, yet Good, who is white, was terminated from her position as a supervisor. Under the direct method, however, we cannot conclude that Good’s disparate treatment was racially motivated without evidence pointing more directly to a discriminatory motive without reliance on speculation. See Cerutti, 349 F.3d at 1061. Nothing in this record indicates that UCMC was motivated to terminate Good because of her race. Good has not presented evidence that anyone at UCMC had an anti-white bias, nor has she presented evidence that UCMC has a history of discrimination against white people. She cannot, for example, point to a formal or informal affirmative action policy in her workplace that might have required or encouraged UCMC to deviate from its demotion policy in her case because of her race. She never heard any anti-white slurs or jokes in the workplace, be they “stray remarks” or not. In addition, the undisputed facts show that Good’s position was filled by another white person. In this reverse race discrimination case, the requisite “mosaic” from which a reasonable jury might conclude that UCMC was motivated to terminate Good based on her race is simply not in the evidence.
III. Indirect Method of Proof
We turn now to Good’s argument that she presented a prima facie case of discrimination under the indirect, burden-shifting method initially set forth in McDonnell Douglas Corp. v. Green, 411 U.S. 792, 93 S.Ct. 1817, 36 L.Ed.2d 668 (1973). Under the indirect method, to establish a prima facie case of racial discrimination in a reverse discrimination suit involving disparate discipline such as this one, the plaintiff bears the burden of establishing: (1) “ ‘background circumstances’ that demonstrate that a particular employer has ‘reason or inclination to discriminate invidiously against whites’ or evidence that ‘there is something “fishy” about the facts at hand’ (2) that she suffered an adverse employment action; and (3) that she was treated less favorably than similarly situated individuals who are not members of the protected class. Phelan v. City of Chicago, 347 F.3d 679, 684-85 (7th Cir.2003) (altering first prong of the indirect case to account for reverse nature of race discrimination claim), quoting Mills v. Health Care Service Corp., 171 F.3d 450, 455 (7th Cir.1999); Peele v. Country Mut. Ins. Co., 288 F.3d 319, 329 (7th Cir.2002) (“When a plaintiff produces evidence sufficient to raise an inference that an employer applied its legitimate expectations in a disparate manner ... the second and fourth prongs of McDonnell Douglas merge — allowing the plaintiff to *679establish a prima facie case, stave off summary judgment for the time being, and proceed to the pretext inquiry.”). Summary judgment in the defendant’s favor is proper if a plaintiff fails to set forth a prima facie case. See Burks v. Wisconsin Dep’t of Transportation, 464 F.3d 744, 751 (7th Cir.2006). Otherwise, if the plaintiff satisfies her initial burden, the burden shifts to the defendant to present a legitimate, nondiscriminatory reason for its decision. Id. If the defendant does so, the burden returns to the plaintiff to show that the defendant’s explanation was pretextual. Id.
We find, like the district court before us, that Good has failed to present sufficient evidence of the first prong for this reverse race discrimination case: evidence of “background circumstances” demonstrating that UCMC has “reason or inclination to discriminate invidiously against whites,” or evidence that there is something “fishy” about her termination. See Phelan, 347 F.3d at 684-85. In Phelan, the plaintiff was a white man who argued that he had been unfairly treated because of his race. His reverse discrimination case failed because he was unable to present any facts from which a jury could infer that his white superiors were inclined to discriminate against their fellow whites; indeed, the plaintiff had been replaced by a white person. Id. Good, too, is white and was replaced by another white person. Also like the plaintiff in Phelan, Good has offered no facts in the summary judgment record that could suggest to a reasonable jury that UCMC had any reason or inclination to discriminate against white persons. See, e.g., Hague v. Thompson Distribution Co., 436 F.3d 816, 822-23 (7th Cir.2006) (five white plaintiffs satisfied the “background circumstances” prong by presenting evidence that after their African American boss fired them, they were replaced by three African American employees, an African American employee was assigned duties of the fourth, and the fifth was not replaced); Ballance v. City of Springfield, 424 F.3d 614, 618 (7th Cir.2005) (background circumstances shown by law firm’s report confirming that police chief took minority race and female gender into account when hiring, assigning, promoting, and disciplining officers).
Good argues that this prong is satisfied by her evidence that UCMC departed from its demotion policy in an “unprecedented fashion.” Good Br. 28. She relies on Mills v. Health Care Service Corp., 171 F.3d 450, 457 (7th Cir.1999). In that case, we found that Mills, a male, made a sufficient showing of suspicious “background circumstances” by showing that over a seven-year span, nearly all promotions in his office went to women instead of men and that women dominated the supervisory positions in the relevant office. See id. at 457. As explained above, Good’s evidence does not rise to this level. Good has evidence that in three instances, UCMC gave a minority employee with deficient performance the opportunity to take a demotion rather than suffer more onerous discipline. But again, nothing in the record demonstrates that UCMC has an anti-white bias or a history of discrimination against white people. Good has no evidence that a formal or informal “affirmative action” policy was in force in her workplace. After Good’s termination, her position was filled by another white person. We agree with the district court that Good has not offered evidence of any fishy “background circumstances” from which a reasonable finder of fact could conclude that UCMC was motivated by improper, racially-based motives when it terminated her employment rather than demoting her. Accordingly, we find that Good has failed to establish a prima facie case of discrimination under the indirect method.
*680Finally, we recognize that the direct and indirect methods for proving and analyzing employment discrimination cases are subject to criticism. They have become too complex, too rigid, and too far removed from the statutory question of discriminatory causation. See Coleman, 667 F.3d at 862-63 (Wood, J., concurring). If we look away from the intricacies of the direct and indirect methods here and focus on the summary judgment evidence as a whole, we still do not see evidence that would allow a reasonable finding of reverse race discrimination in favor of Good.
The judgment of the district court is Affirmed.