This is the paradigmatic “hard” case, and we have labored for many months to reach the right result. On appeal, Defendant, Joe’s Stone Crab, Inc. (“Joe’s”), challenges the district court’s entry of judgment in favor of Plaintiff, the Equal Employment Opportunity Commission (the “EEOC”), on its gender-based disparate impact claims under Title VII. Joe’s is a landmark Miami Beach seafood restaurant which from 1986 to 1990 hired 108 male food servers and zero female food servers. After the EEOC filed its discrimination charge in June 1991, Joe’s hired 88 food servers from 1991 to 1995, nineteen, or roughly 21.7%, of whom were female. The district court concluded that while Joe’s was not liable for intentional discrimination, it was liable for disparate impact discrimination based on these statistical disparities. After thorough review, we vacate the district court judgment, and remand for reconsideration of the EEOC’s *1268intentional discrimination claim consistent with this opinion.
In our view, the facts of this case render a disparate impact finding inappropriate. A disparate impact claim requires the identification of a specific, facially-neutral, employment practice causally responsible for an identified statistical disparity. On this record, the district court has identified no facially-neutral practice responsible for the gender disparity in Joe’s food server population and we can find none. However, some of the district court’s subsidiary findings suggest that there may have been facially-discriminatory practices of Joe’s that were responsible for the identified hiring disparity, although the district court expressly rejected the EEOC’s intentional discrimination claim in summary fashion. Several powerful prudential considerations, including the fact that the record is replete with conflicting witness testimony permitting more than one resolution of this claim, and the fact that some of the district court’s subsidiary factual findings are in apparent conflict with its conclusion that Joe’s was not liable for intentional discrimination, persuade us that the wisest course is a remand to the district court so that it may consider further its factual findings and conclusions of law in light of this opinion.
I.
The facts of this case are reasonably straightforward and are fully outlined by the district court in EEOC v. Joe’s Stone Crab, Inc., 969 F.Supp. 727 (S.D.Fla.1997). Joe’s Stone Crab, Inc. is a fourth-generation, family-owned seafood restaurant and Miami Beach landmark. During the stone crab season, which lasts from October to May, the restaurant is extremely busy— serving up to 1450 patrons each weeknight and up to 1800 patrons each weekend night. Today, the restaurant employs between 230 and 260 employees; of those, approximately 70 are food servers. Throughout its history, Joe’s has experienced extremely low food server turnover — a result of Joe’s family ethos, generous salary and benefits package, and its seven-month employment season. From 1950 onward, however, the food servers have been almost exclusively male.
On June 25, 1991, the Equal Employment Opportunity Commission (“EEOC”) filed a discrimination charge, under sections 706 and 707 of Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e, et. seq., alleging that Joe’s discriminated on the basis of sex in the hiring and recruiting of food servers. On April 17, 1992, the EEOC issued its Decision, finding a pattern and practice of intentional sex discrimination in Joe’s hiring and recruiting practices. Specifically, the EEOC determined that a word-of-mouth recruiting system and Joe’s reputation for hiring only male food servers resulted in almost no women actually applying for food server positions at Joe’s. The EEOC also found that Joe’s subjective hiring practices were responsible for the gross statistical disparity between the percentage of female food servers in the Miami Beach community and the percentage of female food servers working at Joe’s. As required by Title VII, see 42 U.S.C. § 2000e-5(b), the EEOC and Joe’s attempted to conciliate the Decision’s findings but were unsuccessful.
On June 8, 1993, the EEOC filed a complaint in the Southern District of Florida alleging that Joe’s violated Title VII through both intentional disparate treatment discrimination as well as unintentional disparate impact discrimination. The gravamen of the complaint centered around the EEOC’s findings with respect to Joe’s hiring and recruiting practices for food servers. The EEOC sought permanent injunctive relief, back pay, and prejudgment interest for qualified claimants.1 Over fifteen days interspersed from August 1996 to December 1996, the district *1269court held a liability bench trial. The analysis at trial focused on two discrete time periods: first, the pre-EEOC charge period from 1986 to 1990; and finally, the post-EEOC charge period from 1991 to 1995. On'July 3, 1997, the district court issued a partial final judgment — making a series of factual findings with respect to Joe’s employment practices. See Joe’s Stone Crab, 969 F.Supp. at 727-35.
To hire new food servers, Joe’s conducts a “roll call” every year on the second Tuesday in October. Although Joe’s rarely advertises, significantly, the district court found that the roll call is “widely known throughout the local food server community,” and typically attracts over 100 applicants for only a limited number of slots. Joe’s Stone Crab, 969 F.Supp. at 733. At a typical roll call, each applicant completes a written application and an individual interview. Selected applicants then enter a three-day training program where they shadow experienced servers. Upon successful completion of the program, they then become permanent hires. See id.
Until the EEOC’s charge, roll call interviews and hiring selections were handled exclusively by the daytime maitre d’ with occasional interview assistance from other staff members.2 Hiring decisions were made by the daytime maitre d’ on the basis of four subjective factors (appearance, articulation, attitude, and experience) and without upper management supervision or the benefit of instructive written or verbal policies. See id. After the EEOC’s discrimination charge in 1991, Joe’s changed its roll call format somewhat. All applicant interviews were conducted by three members of Joe’s management.3 In addition, each applicant was required to take and pass a “tray test,” which involved the lifting and carrying of a loaded serving tray, or else be automatically disqualified from a food server position. The district court found the tray test to be a “legitimate indicator of an individual’s ability to perform an essential component of a food server’s job at Joe’s,” id., and that “women have the physical strength to carry serving trays,” id. at 732.
In addition to its description of Joe’s hiring process, the district court also made several subsidiary findings relating to the historical operation of the roll call system. The district court observed that while “women have predominated as owner/managers,” “most of Joe’s female employees have worked in positions traditionally viewed as ‘women’s jobs,’ e.g., as cashiers or laundry workers. Food servers generally have been male.” Id. at 731. Although Joe’s hired female food servers during World War II, most of these positions “reverted to men at the’conclusion of the war.” Id. Further, the district court found that, “[f|rom 1950 on, the food serving staff has been almost exclusively male. Indeed, one striking exception proves the rule. Dotty Malone worked as a food server at Joe’s for seventeen years, and for most of this time she was the lone female on a serving staff that ranged between twenty-four and thirty-two.” Id.
In explaining this historical dearth of female food servers, the district court found that Joe’s maintained an “Old World” European tradition, in which the highest level of food service is performed *1270by men, in order to create an ambience of “fine dining” for its customers. Id. at 733. The district court elaborated:
The evidence presented at trial does not establish that Joe’s management had an express policy of excluding women from food server positions. To the contrary, the evidence portrays owner/managers who have been courageous in opposing overt discrimination. For example, Joe’s was picketed for two years when the owners insisted on hiring African-American employees who had been excluded from union membership because of race. What the evidence in this case does prove is that Joe’s management acquiesced in and gave silent approbation to the notion that male food servers were preferable to female food servers.
Id. at 731. As evidence, for this finding, the district court cited three pieces of witness testimony. First, the district court pointed to the testimony of Grace Weiss, Joe’s owner, who stated, “I cannot explain the predominance of male servers, but perhaps it has to do with the very heavy trays to be carried, the ambience of the restaurant, and the extremely low turnover in servers.” Id. at 731-32 (emphasis added by the district court). Second, the district court highlighted the testimony of Roy Garrett, a longtime maitre d’ of Joe’s with hiring authority, who explained that Joe’s had a “tradition” that food server positions were “a male server type of job”:
As I said before, we had very few female applicants over the years. It was sort of a tradition.... It was always tradition from the time I arrived there that it was a male server type of job. And until just recently when we became aware that we had to do other things, ... originally it was traditionally a male place. We always had women that were qualified women .... Traditionally, I mean, it’s just some restaurants, when you walk in, you know there are going to be women waitresses, other restaurants you know it is going to be male waiters.
Id. at 732 (emphasis added).4 Finally, the district court referred to the testimony of Joe’s own restaurant industry expert, Karen McNeil, for a historical explanation of the “male-only” server tradition.
It has been an attitude and standard, it comes from Europe. In all of Europe you will find in all of the grade three restaurants in Europe, there is an impression that service at that high level is the environment of men, and that it ought to be that way. And I think that that attitude a few decades ago came and was felt a little bit here in this country.... Those [European] opinions and those sensibilities, I think were in fact carried here by restauranteurs who hoped to create something serious. If you wanted to create a serious restaurant that would become known in the community, that would become one of the community’s great restaurants, you did what they did in Europe, you modeled yourself after them. I don’t think anybody thought about it. They said, well, men did it there. It tended to be men here, too, who had those skill.sets, and so men were [sic] automatically became the labor pool.
Id. The district court added that “Joe’s [had] sought to emulate Old World traditions by. creating an ambience in which tuxedo-clad men served its distinctive menu.” Id.
With this historical background in place, the district court then focused on Joe’s female hiring statistics for the relevant pre- and post-charge periods. For the pre-charge period of 1986-1990, the number of female food server applicants at Joe’s annual hiring roll calls was minuscule. While there is little available evidence as to the actual numbers of female applicants at these roll calls (because Joe’s *1271historically did not retain any employment data from its roll calls), the district court determined, and both parties agreed, that during this period, no more than two or three women per year (or, at most, 3% of the overall applicant class) actually attended the roll calls. - See id. at 733. In that same period, 108 new male food servers were hired while zero women were hired. See id. During the post-charge period (from 1991 to 1995), many more women (in all, 22% of the actual applicant pool) applied for food server positions. Of Joe’s 88 new food server hires during this period, 19 were women. These post-charge figures translate into a female hiring percentage of 21.7% — -a percentage almost exactly proportional to the percentage of females in the actual applicant pool. See id. at 733-34. Joe’s female applicant flow data for the post-charge period breaks down the following way:
Season Women applicants Women hired
1991-92 15.1% 20.0%
1992-93 21.9% 22.7%
1993-94 23.0% 10.5%
1994-95 26.8% 35.3%
Oct. — Dec.1995 23.3% 20.0%
22.02% 21.7% Average
Id. at 734.
However, in making its findings, the district court found this actual applicant flow data “unreliable because it is skewed.” Id. at 734. Relying on hearsay trial testimony from local female food servers, the district court found that Joe’s public reputation for not hiring women encouraged women to self-select out of the hiring process— thereby skewing the actual applicant flow.5 See id. at 733-34. The district court explained:
In the preceding findings, the court held that Joe’s reputation in the community, which reflected the restaurant’s historical hiring practice, led potential female applicants not to apply for server positions. Joe’s reputation, therefore, was largely responsible for the gender skew in the pool of applicants at the annual roll call. It is well-settled that an employer’s reputation for discriminatory hiring practices can lead to a self-selected applicant pool not reflective of the actually available labor pool. Quite irrespective of the intentions of the employer, a rational qualified female candidate is likely to self-select out of the application process, declining to make what she considers a “futile gesture.” Thus,-the existence of such a reputation is highly relevant to whether Joe’s actual applicant flow data reflects the available labor pool. Put more narrowly, evidence *1272of Joe’s reputation in.the food serving community was admitted as highly relevant to whether, how, and why would-be applicants were chilled from applying for traditionally male jobs. .While Joe’s vehemently. contested the admissibility of this reputation evidence, its objections fall wide of the mark. They presuppose that evidence of Joe’s reputation was offered as proof of conduct consistent with the reputation,, as proof of Joe’s hiring practices themselves, or as proof of bad character or intent to discriminate. None of these is the case. Evidence of Joe’s reputation was admitted solely to establish the existence of the reputation, and not for any other purpose.
Id. at 736 (internal citations omitted). Although the district court noted that female food server applications to Joe’s dramatically increased as a result of publicity about the EEOC charge, it still found Joe’s post-charge applicant pool data (depicting a female applicant pool of 22%) unreliable after comparing' it with hiring rates, between 30% and 40% female, for other area seafood restaurants.6 See id. at 734.
Having found the actual applicant pool data wholly unreliable, the district court discarded it and then set about selecting alternative non-applicant labor market data. The EEOC’s expert witness, a labor economist, suggested a qualified female labor pool of 44.1% based on 1990 census data for female food servers living and/or working in the Miami Beach area (a labor pool which included cocktail and buffet servers). See id. at 734-35. Not surprisingly, the district court rejected this figure in part because there was no demonstration that this female labor pool necessarily was qualified to work at Joe’s. Instead, the district court “refined” the relevant labor pool to include all female servers who lived or worked on Miami Beach and earned between $25,000 and $50,000 — thereby “using past earning capacity as a proxy for experience, and by extension, experience as a proxy for qualification.” Id. at 735. Solely based on this alternative methodology, the district court was able to find “that at all relevant times, 3119% of the available labor pool has been female.” Id.
With these findings in place, the district court then drew two pertinent conclusions of law. First, the district court summarily rejected the EEOC’s disparate treatment claims without analysis, stating only that “the court finds that the EEOC has not met its burden of proof under disparate treatment analysis.” Id. at 735. The only other mention of the disparate treatment claims is found in the introduction of the district court’s opinion. There, the district court unambiguously states: “[biased on an evaluation of the evidence, the court finds that the EEOC has not proven intentional discrimination.” Id. at 730. Second, however, the district court determined that Joe’s was liable for disparate impact discrimination.7 See id. Specifically, the district court found that “the challenged employment practice in this case ..., [Joe’s] undirected and undisciplined delegation of hiring authority to subordinate staff,” id. at 738, was responsible for the statistical disparity between the 31.9% female “available” labor pool and Joe’s female hiring rates in the pre-charge (0%) and post-charge (21.7%) periods, id. at 739-40. The district court then entered a partial judgment of liability in favor of the EEOC.
*1273On April 15, 1998, a bench trial was held on the remedies portion. The EEOC presented five female plaintiffs who unsuccessfully applied for food server positions at Joe’s in the 1990’s. They testified that they would have applied to Joe’s at an earlier juncture but for the fact that they knew applying was futile based on Joe’s male-only reputation. The. district court awarded four of them backpay relief plus prejudgment interest. The district court also ordered extensive injunctive relief through the year 2001 that required Joe’s to adopt a statement of non-discrimination in the hiring of food servers, comply "with the district court’s monitoring of Joe’s future hiring and recruiting practices (including its public advertising of hiring roll-calls), allow the supervision of each roll-call by a court-appointed monitor, permit the introduction of a standardized tray test at the roll-call, and provide mandatory training sessions with an industrial psychologist for all of Joe’s hiring decisionmakers.
II.
The first and central issue in this appeal is whether the district court erred in finding that the EEOC had established disparate impact discrimination.8 We review the district court’s conclusions of law de novo, and its factual findings for clear error. See Central State Transit & Leasing Corp. v. Jones Boat Yard, Inc., 206 F.3d 1373, 1376 (11th Cir.2000); Hill v. Seaboard Coast Line R.R. Co., 885 F.2d 804, 812 (11th Cir.1989) (citing Eastland v. Tennessee Valley Auth., 704 F.2d 613, 620 (11th Cir.1983)). In this case, where the bulk of the evidence came in the form of conflicting witness testimony, we allot even greater deference to the factfinder who is in a better position to assess the credibility of the witnesses. See Stano v. Butterworth, 51 F.3d 942, 944 (11th Cir.1995) (citing Anderson v. City of Bessemer City, 470 U.S. 564, 575, 105 S.Ct. 1504, 1512, 84 L.Ed.2d 518 (1985)).
That said, we have struggled on appeal to find the proper resolution of this case. As we explain in detail, we believe that the district court’s factual findings simply do not support a legal conclusion that Joe’s is liable for disparate impact discrimination. Based on the district court’s findings, no specific facially-neutral employment practice of Joe’s can be causally connected to the statistical disparity between the percentage of women in the qualified labor pool and the percentage of women hired as food servers by Joe’s.
A. Disparate Impact
Under Title VII of the Civil Rights Act of 1964, an employer may be found liable for unlawful sex discrimination under any one of three discrete theories: pattern and practice discrimination, disparate treatment discrimination, or disparate impact discrimination. Both pattern and practice and disparate treatment claims require proof of discriminatory intent;9 disparate impact claims do not. See In Re Employment Litig. Against the State of Ala., 198 F.3d 1305, 1310 n. 8 (11th Cir.1999). In order to show discriminatory intent, a plaintiff must demonstrate “ ‘that the decisionmaker ... selected or reaffirmed a particular course of action at least in part “because of,” not merely “in spite of,” its adverse effects on an identifiable group.’” Id. at 1321 (quoting Personnel Administrator of Mass. v. Feeney, 442 U.S. 256, 279, 99 S.Ct. 2282, 2296, 60 L.Ed.2d 870 (1979)). Therefore, in a dis*1274parate treatment case, the plaintiff bears the ultimate burden of proving that the employment action at issue was taken because of the plaintiffs sex. See Holifield v. Reno, 115 F.3d 1555, 1564-65 (11th Cir.1997). Likewise, in a pattern and practice case, the plaintiff must prove, normally through a combination of statistics and anecdotes, that discrimination is the company’s “standard operating procedure.” International Brotherhood of Teamsters v. United States, 431 U.S. 324, 335-36, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977); see also Mozee v. American Commercial Marine Service Co., 940 F.2d 1036, 1051 (7th Cir.1991).
In contrast, disparate impact theory prohibits neutral employment practices which, while non-discriminatory on their face, visit an adverse, disproportionate impact on a statutorily-protected group. See Griggs v. Duke Power Co., 401 U.S. 424, 431, 91 S.Ct. 849, 853, 28 L.Ed.2d 158 (1971) (explaining that Title VII “proscribes not only overt discrimination but also practices that are fair in form, but discriminatory in operation”); see also In re Employment, 198 F.3d at 1311; Fitzpatrick v. City of Atlanta, 2 F.3d 1112, 1117 (11th Cir.1993). The doctrine seeks the removal of employment obstacles, not required by business necessity, which create “ ‘built-in headwinds’ ” and freeze out protected groups from job opportunities and advancement. Griffin v. Carlin, 755 F.2d 1516, 1524 (11th Cir.1985) (quoting Griggs, 401 U.S. at 431-32, 91 S.Ct. 849). As the district court correctly identified, “[t]he premise of disparate impact theory is that some employment practices, adopted without a deliberately discriminatory motive, may be the. functional equivalent of intentional discrimination.” Joe’s Stone Crab, 969 F.Supp. at 735. In essence, disparate impact theory is a doctrinal surrogate for eliminating unprovable acts of intentional discrimination hidden innocuously behind facially-neutral policies or practices.
The disparate impact framework under Title VII by now is well-settled. “Since Griggs, Congress has codified the appropriate burdens of proof in a disparate impact case in 42 U.S.C. § 2000e-2(k) (1994), and a settled" jurisprudence has arisen to implement the methodology.” In re Employment; 198 F.3d at 1311. As correctly identified by the district court, a plaintiff in a sex discrimination suit must establish three" elements: first, that there is a significant statistical disparity between the proportion of women in the available labor pool and the proportion of women hired; second, that there is a specific, facially-neutral, employment practice which is the alleged cause of the disparity; and finally, and most critically in this case, that a causal nexus exists between the specific employment practice identified and the statistical disparity shown. Joe’s Stone Crab, 969 F.Supp. at 735. See generally MacPherson v. University of Montevallo, 922 F.2d 766, 771 (11th Cir.1991) (citing Wards Cove Packing Co. v. Atonio, 490 U.S. 642, 655-56, 109 S.Ct. 2115, 2124, 104 L.Ed.2d 733 (1989); Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 994-95, 108 S.Ct. 2777, 2789, 101 L.Ed.2d 827 (1988)).
According to Title VII, “[i]n the first stage of a disparate impact case, the ‘complaining party [must] demonstrate [ ] that a respondent uses a particular employment practice that causes a disparate impact on the basis of race, color, religion, sex, or national origin.’ ” In re Employment, 198 F.3d at 1311 (quoting 42 U.S.C. § 2000e-2(k)(1)(A)(i)). “To ‘demonstrate’ means to ,‘meet[ ] the burdens of production and persuasion.’ ” Id. (quoting 42 U.S.C. § 2000e(m) (1994)). “In other words, in order to surmount the first hurdle in a disparate impact race discrimination case, the plaintiff must make out a prima facie case ‘that [a] facially neutral employment practice ha[s] a significantly discriminatory impact.’ ” Id. (quoting Connecticut v. Teal, 457 U.S. 440, 446, 102 S.Ct. 2525, 2530, 73 L.Ed.2d 130 (1982)). As the Supreme Court explained in Watson, “the plaintiff must offer statistical evi*1275dence of a kind and degree sufficient to show that the practice in question has caused the exclusion of applicants for jobs or promotions because of their membership in a protected group.” Watson, 487 U.S. at 994, 108 S.Ct. 2777 (emphasis added); see also Edwards v. Wallace Community College, 49 F.3d 1517, 1520 (11th Cir.1995) (observing that “[a] plaintiff must identify a specific employment practice that leads to the disparate impact”); MacPherson, 922 F.2d at 771(noting that “ ‘a plaintiff must demonstrate that it is the application of a specific or particular employment practice that has created the disparate impact under attack’ ”) (internal citation omitted).
Once each of these three elements are shown, a plaintiff has established a prima facie case of disparate impact discrimination. See Fitzpatrick, 2 F.3d at 1117; MacPherson, 922 F.2d at 771. The burden of production then shifts to the defendant to establish that the challenged employment practice serves a legitimate, non-diseriminatory business objective. See Fitzpatrick, 2 F.3d at 1117. However, even if the defendant satisfies this burden, a plaintiff may still prevail by proving that an alternative, non-discriminatory practice would have served the defendant’s stated objective equally as well. See id. at 1118.
As for the first prong of the analysis, it is critical to observe that no statistically-significant disparity exists between the percentage of women who actually applied to Joe’s and the percentage of women who were hired as servers by Joe’s. The record indicates that for the precharge period (October 1986 to June 1991) very few female food servers applied to Joe’s, “perhaps 3% of [all] applicants,” Joe’s Stone Crab, 969 F.Supp. at 734, out of an actual applicant pool of between 80 and 120 people a year.10 In this five-year time period, 108 male food servers were hired and no women were hired. Despite the fact that no women were hired during this period, Joe’s pre-charge hiring rate demonstrated no significant statistical disparity because so few women actually applied for food server positions.11 For the post-charge period (July 1991 to December 1995), the district court found that, on average, 22.02% of Joe’s food server applicants were women and that Joe’s hired roughly 21.7% women for these positions. Both parties admit (as they must) that, based on this record, there is no statistieally-significant hiring disparity when the actual number of female applicants is compared to the actual number of female hires for either period.12 In other words, Joe’s hiring system did not produce a significant statistical disparity between the actual *1276percentage of women who applied to Joe’s for server positions and the percentage of women actually hired for these positions.
This insight is important for disparate impact analysis because the mere fact that Joe’s hired no women in the pre-charge period is not, alone, sufficient to impose upon Joe’s Title VII liability. To hold otherwise would be to impose liability upon Joe’s based on “bottom line” reasoning which the Supreme Court has expressly forbade. In Watson, the Supreme Court made clear that Title VII liability could not be based solely on “bottom line” statistical imbalances in an employer’s workforce. See Watson, 487 U.S. at 992, 108 S.Ct. 2777 (explaining that it is “unrealistic to suppose that employers can eliminate, or discover and explain, the myriad of innocent causes that may lead to statistical imbalances in the composition of their workforces”). The Supreme Court then further explained in Wards Cove:
Just as an employer cannot escape liability under Title VII by demonstrating that, “at the bottom line,” his work force is racially balanced (where particular, hiring practices may operate to deprive minorities of employment opportunities), a Title VII plaintiff does not make out a case of disparate impact simply by showing that, “at the bottom line,” there is racial imbalance in the work force. As a general matter, a plaintiff must demonstrate that it is the application of a specific or particular employment practice that has created the disparate impact under attack. Such a showing is an integral part of the plaintiffs prima facie case in a disparate-impact suit under Title VII.
Id., 490 U.S., at 656-57, 109 S.Ct. 2115 (internal citation omitted) (emphasis added); see also MacPherson, 922 F.2d at 771.
This disdain for “bottom line” reasoning reflects the belief that holding employers liable for statistical imbalances 'per se is inconsistent with Title VII’s plain language and statutory purpose. Section 708(3) of Title VII, 42 U.S.C.2000e-2(j), in fact, explicitly rejects the notion that employers must adopt numerical hiring quotas or “grant preferential treatment ... on account of an imbalance which may exist with respect to the total number or percentage of persons ... in comparison with' the total number or percentage ... in any community.” Based on this statutory language, the Supreme Court has interpreted this provision of Title VII to mean that employers possess no affirmative duty to redress workforce imbalances not attributable to their own corporate conduct. See Watson, 487 U.S. at 993, 108 S.Ct. 2777 (finding that employers have no duty under Title VII to ameliorate uncaused workforce imbalances because such a legal rule is “ ‘far from the intent of Title VII’ ”) (quoting Albemarle Paper Co. v. Moody, 422 U.S. 405, 449, 95 S.Ct. 2362, 45 L.Ed.2d 280 (1975) (Blackmun, J., concurring)); Furnco Construction Corp. v. Waters, 438 U.S. 567, 577-78, 98 S.Ct. 2943, 2950, 57 L.Ed.2d 957 (1978) (finding that employers are under no affirmative duty to impose hiring quotas to reflect demographic percentages). Indeed, if employers could be held liable for an unlawful disparate impact on account of statistical workforce imbalances per se, then they would be forced to use numerical quotas and other forms of preferential treatment in their hiring and promotion policies, in express contravention of Title VII,, in order to insulate themselves from the potential legal liability that would arise if their workforce demographics did not closely mirror the demographics of their surrounding community or local competitors. As a result, a plaintiff must do more than simply identify a workforce imbalance to establish a prima facie disparate impact case; it must causally connect a facially-neutral employment practice to the identified disparity.
In this case, the district court could create a statistically-significant disparity only by first throwing out the actual applicant data as a point of comparison and instead comparing the percentage of women hired for server positions at Joe’s with the percentage of women in the “qualified” labor pool. The district court recognized *1277that the number of women who actually applied for server positions at Joe’s was disproportionately low when compared with the number of women in the Miami Beach area who were seemingly qualified for such positions. There was, in fact, a significant statistical disparity between the percentage of female applicants to Joe’s during the pre- and post-charge periods and the percentage of female applicants to comparable area restaurants. Joe’s female applicant percentage of 3% and 22.02% for the pre-and post-charge periods respectively varied sharply from the female applicant percentage of area restaurants which ranged from 29.5% to 42.1%. See Joe’s Stone Crab, 969 F.Supp. at 734. As a result of these findings, the- district court found the actual applicant flow data to be “unreliable because it is skewed.” See id. It concluded that the data was skewed because of the pronounced self-selection of women out of Joe’s hiring process. See id. The district court then expressly rejected the actual applicant flow data in favor of an alternative labor pool consisting of those local food servers who were theoretically “available” and “qualified” to work at Joe’s.13 After hearing testimony from the expert witnesses of both parties, the district court arrived at an “eligible” labor pool, based on 1990 census data refined for qualification/experience on the basis of past earning capacity, which was 31.9% female. The district court- then used this alternative labor data and compared it to Joe’s actual hire statistics. By comparing Joe’s pre-charge female hiring percentage (0%) with the percentage of women in the qualified labor market (31.9%), the district court created a legally-cognizable statistical disparity.14
*1278Assuming this substitution of data was appropriate, in order to establish disparate impact discrimination, the EEOC still was required to show a causal link between some facially-neutral employment practice of Joe’s and the statistical disparity.15 In other words, the EEOC was required to prove that at least one facially neutral employment practice proximately caused the disparity. This finding is essential to avoid the potential conflation of disparate treatment and disparate impact claims. As we have noted, the central difference between disparate treatment and disparate impact claims is that disparate treatment requires a showing of discriminatory intent and disparate impact does not. See In re Employment, 198 F.3d at 1310 n. 8. In fact, the judicial doctrine of disparate impact was created in Griggs specifically to redress facially-neutral policies or practices which visited disproportionate effects on groups protected by Title VII. The Supreme Court has explained that, “[u]n-der the act, practices, procedures, or tests neutral on their face, and even neutral in terms of intent, cannot be maintained if they operate to ‘freeze’ the status quo of prior discriminatory employment practices.” Griggs, 401 U.S. at 430, 91 S.Ct. 849 (emphasis added).
Since Griggs, we are aware of no case in which a facially-discriminatory practice has been challenged successfully under a disparate impact theory. Simply put, disparate impact theory is available for the challenge of facially-neutral employment practices. See, e.g., Lanning v. Southeastern Pennsylvania Transp. Auth., 181 F.3d 478, 485 (3rd Cir.1999) (finding that “plaintiffs estabhsh a prima facie case of disparate impact by demonstrating that application of a facially neutral standard has resulted in a significantly discriminatory hiring pattern”) (emphasis added). Indeed, the district court properly recognized that “[s]ex discrimination under the theory of disparate impact occurs when a facially neutral rule or practice of the employer has a disproportionate impact on one sex.... To establish a prima facie case of disparate impact sex discrimination, the plaintiff must show that a facially neutral practice of the employer has a disproportionate impact on one sex.” Joe’s, 969 F.Supp. at 735 (emphasis added).
The central problem in this case, however, is that the district court has identified no facially-neutral employment practice responsible for the gender disparity in Joe’s food server population, and we can find none. The EEOC and the district court have identified, at most, two neutral employment practices on which to ground a disparate impact’ analysis: first, Joe’s word' of mouth recruiting, and second, Joe’s “undirected and undisciplined delegation of hiring authority to subordinate staff,” Joe’s Stone Crab, 969 F.Supp. at 738, resulting in its subjective “roll call” hiring process. Disparate impact analysis fails in this case because neither neutral practice can be causally connected to the gender disparity.
*1279First, there is no evidence that Joe’s word of mouth recruiting method caused any disparity between the percentage of women in the qualified labor pool and the percentage of women actually hired by Joe’s as servers. Notably, this is not a case where Joe’s formal recruiting practices or its informal word-of-mouth recruiting network kept women from learning about available jobs at Joe’s. Compare United States v. Georgia Power Co., 474 F.2d 906, 925 (5th Cir.1973) (finding that word-of-mouth recruiting system can operate as a “‘built-in headwind’” isolating blacks from “web of information” relating to job openings). Rather, the district court specifically found quite the opposite, namely that local female food servers knew about the availability of positions at Joe’s and the logistical details of Joe’s hiring roll calls. Indeed, it observed that although the hiring roll calls were “rarely advertised,” they were “widely known [about] throughout the local food server community.” Joe’s Stone Crab, 969 F.Supp. at 733. No woman testified that, she failed to apply for a position at Joe’s because she was unaware of Joe’s roll call method for filling openings. Plainly, the disparity between the percentage of women in the qualified labor pool and the percentage of women actually hired as servers by Joe’s cannot be causally linked to Joe’s word-of-mouth recruiting process because this practice in no way prevented women from applying to or being hired by Joe’s.
Nor is there any evidence that Joe’s facially-neutral, albeit undisciplined and subjective, hiring practices caused the disparity the district court found between the percentage of women in the qualified labor pool and the percentage of women actually hired as servers by Joe’s. There is no evidence that Joe’s subjective hiring criteria either caused women not to apply to Joe’s or caused those who applied not to be hired. Joe’s hiring roll call decisions were made through a subjective hiring process in which Joe’s hiring maitre d’ relied on short applicant interviews to assess an applicant’s qualification based on a range of subjective factors, including “appearance, attitude, articulation, and experience.” Joe’s Stone Crab, 969 F.Supp. at 733. No witnesses testified and no evidence was presented into the record indicating that any women failed to apply to Joe’s because its hiring criteria included specific judgments about an applicant’s appearance, attitude, articulation, or experience. Nor was any evidence presented showing that women who did apply for server positions at Joe’s were disadvantaged by these specific hiring criteria. Indeed, as we have stated previously, there is in fact no disparity between the percentage of women who actually applied to Joe’s for server positions and the percentage of women hired. Plainly, therefore, the subjective hiring criteria did not harm women once they entered the application process.16
The district court, recognizing that it could not causally connect Joe’s neutral, albeit subjective, recruiting and hiring practices with the disparity between the percentage of women in the qualified labor pool and the percentage of women actually hired as servers by Joe’s, identified Joe’s reputation as a discriminator against women as the causal agent for the disparity. See Joe’s Stone Crab, 969 F.Supp. at 740. For the district court, Joe’s reputation for not hiring female food servers acted as the essential bridge connecting the neutral practices to the statistical disparity. In other words, according to the district court’s own reasoning, it was not Joe’s neutral recruiting or hiring practices that *1280caused the disparity, but rather Joe’s reputation as a discriminator against women. Because of Joe’s reputation for discriminating, the district court essentially found, women did not apply to Joe’s and therefore were not hired as servers.
We conclude that the district court’s .use of reputation was, on the face of this record, both problematic and inadequate for several independent reasons. First, reputation itself is neither a specific act or a practice. It is far more amorphous. Reputation is “ ‘a prevalent or common belief, a general name, the opinion of a number of persons.’ ” United States v. North Carolina Nat'l Bank, 336 F.2d 248, 253 (4th Cir.1964) (quoting United States v. C.I.T. Corp., 93 F.2d 469, 471 (2nd Cir.1937)). Reputation is the community “picture” of an individual or corporate entity formed over a number of years. See generally Michelson v. United States, 335 U.S. 469, 477, 69 S.Ct. 213, 93 L.Ed. 168 (1948). Reputation has never been used, as far as we can tell, as a facially-neutral employment act or practice for disparate impact purposes. In the intentional discrimination context, some cases have considered reputation evidence for the limited purpose of defining the parameters of Title VII remedial relief where intentional discrimination either has been conceded or proven and there is evidence that an employer’s discriminatory practices prevented qualified applicants from applying for new jobs. See Morrow v. Crisler, 491 F.2d 1053, 1055-57 (5th Cir.1974) (en banc) (instructing district court on remand to consider the role of Mississippi Highway Patrol Department’s entrenched reputation for race discrimination—a reputation based on their historical practice of intentional race discrimination—in discouraging black applicants when shaping remedial recruiting policies for the Department); see also EEOC v. Rath Packing Co., 787 F.2d 318, 337 (8th Cir.1986) (explaining that reputation evidence could be considered in determining the relevant labor market for the computation of a Title VII class backpay award given the employer’s well-known historical practice of intentional sex discrimination) (citing Teamsters v. United States, 431 U.S. 324, 365, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977)). We have been pointed to no case, however, and can find none that has treated an employer’s reputation as a discriminator as- itself an act or practice for the purposes of establishing a prima facie case under a theory of disparate impact. Indeed, no case has ever used reputation as a bridge connecting a neutral hiring practice to a statistical disparity in order to establish disparate impact liability where the neutral employment practices alone did not cause the disparity.
In addition, even if reputation could somehow be used in theory as a causal bridge, in this case there is no logical or factual connection between any facially-neutral component of Joe’s employment practices and Joe’s reputation as a discriminator. Nothing in this record indicates that Joe’s recruitment by “word-of-mouth” rather than through other recruiting mechanisms such as print or television advertising contributed in any way to Joe’s reputation for discrimination. Nor is there any evidence that the use of appearance, articulation, attitude, and experience as hiring criteria contributed to Joe’s reputation for discrimination. Indeed, there is no suggestion from either party that these hiring criteria are themselves somehow illegitimate or discriminatory.17 Instead, the suggestion from the EEOC is that these criteria are simply a cover or a smokescreen and do not reflect the real bases for Joe’s hiring decisions. Rather than hiring on the bases of appearance,' articulation, attitude and experience, the EEOC contends, Joe’s really hired servers on the -basis of sex and it is this *1281discriminatory hiring, rather than the use of subjective hiring criteria, that arguably led to Joe’s reputation as a discriminator. See Appellee Brief at 30-31. But, where Joe’s neutral hiring and recruiting practices did not cause its reputation, we think it is wholly inappropriate to use reputation as the causal bridge connecting neutral practices to a statistical disparity for the purposes of establishing Joe’s disparate impact liability. See Lewis v. Tobacco Workers’ Int’l Union, 577 F.2d 1135, 1143 (4th Cir.1978) (holding that an employer cannot be found liable under Title VII simply because potential minority applicants subjectively believe the company will not hire them because of their race where this belief is not attributable, to the employer’s conduct); EEOC v. Sheet Metal Workers, Int’l Assoc., 463 F.Supp. 388, 425 (D.Md.1978) (finding that an employer “must itself do something to violate Title VII; it does not violate Title VII because someone else thinks it has violated Title VII”).
Finally, we observe that the district court expressly admitted evidence of Joe’s reputation as a discriminator not for the truth of the matter asserted but only to show the state of mind of the women who failed to apply for server positions at Joe’s. According to the district court, Joe’s reputation was not entered into evidence “as proof of conduct consistent with the reputation, as proof of Joe’s hiring practices themselves, or as proof of bad character or intent to discriminate” but “was admitted solely to establish the existence of the reputation.” Joe’s Stone Crab, 969 F.Supp. at 736. The district court thereby expressly refused to find any direfet causal connection between any of Joe’s neutral employment practices and its reputation as a discriminator.
While a company may be held liable for a discriminatory reputation if there is evidence it caused or perpetuated that reputation through some intentional affirmative act, see Morrow, 491 F.2d at 1055-57; Rath Packing, 787 F.2d at 337, we know of no federal circuit that has found an employer liable under Title VII on the basis of a reputation for discrimination it did not cause. See Lewis, 577 F.2d at 1143. Nor are we prepared to impose on an employer an affirmative duty under Title VII to ameliorate a public reputation not attributable to its own employment conduct. See id. (observing that “[w]e do not think a failure of the company to announce innocence is a violation of Title VII”); see also Sheet Metal Workers, Int’l Assoc., 463 F.Supp. at 425. In fact, we are unaware of any case that requires a Title VII employer to affirmatively dispel a negative public image not of its own making or else be subject to a finding of Title VII discrimination.
That said, the record extant and some of the district court’s findings of fact can be read to support the alternate conclusion that Joe’s -management intentionally excluded women from food serving positions in order to provide its customers with an “Old World,” fine-dining ambience. Thus, for example, the district court found that “Joe’s management acquiesced in and gave silent- approbation to the notion that male food servers were preferable to female food servers.” Joe’s Stone Crab, 969 F.Supp. at 731. At another point in its findings, the district court observed that “Joe’s sought to emulate Old World traditions by creating an ambience in which tuxedo-clad men served its distinctive menu.” Id. at 732-33. Moreover, the district court apparently;-also credited the testimony of one of Joe’s 'former hiring maitre d’, Roy Garrett, who' explained that Joe’s was “a male server type of job” by tradition. Id. at 732. As a result, the district court said that “women have systematically been excluded from the most lucrative entry level position, that of server.” Id. at 740. Finally, the district court found that this historical practice of hiring only men was responsible for Joe’s “male-only” reputation. The district court held that “Joe’s reputation in the community, which reflected the restaurant’s historical hiring practice, led potential female applicants not to apply for server posi*1282tions. Joe’s reputation, therefore, was largely responsible for the gender skew in the pool of applicants at the annual roll call.” Id. at 736.
But, these factual findings do not mesh easily with a disparate impact theory because they suggest that Joe’s hiring system was not in practice facially-neutral, but rather was facially-discriminatory on the basis of gender. They suggest the conclusion that in fact Joe’s had a desired preference for male servers and that this preference influenced the hiring decisions of Joe’s decisionmakers, resulting in the deliberate and systematic exclusion of women as food servers. If this were true, Joe’s could be found liable for intentional discrimination in violation of Title VII. We emphasize that this is not a case like Griggs, where there was a pronounced history of intentional discrimination followed by a facially-neutral employment practice which perpetuates the effects of an employer’s previous discrimination. See also Rowe v. General Motors Corp., 457 F.2d 348, 356 (5th Cir.1972) (holding that GM’s promotion/transfer standards “freeze” into effect the racial disparity in salaried jobs created by the company’s prior policy of explicit discrimination); Senter v. General Motors Corp., 532 F.2d 511, 526, 530 (6th Cir.1976) (affirming the district court’s finding that the employer’s subjective promotions procedures had the effect of “locking” minorities into the hourly ranks and out of the supervisory ranks). The district court’s findings and the record evidence indicate that Joe’s hiring methodology and practices have remained relatively constant throughout the relevant time periods. Therefore, if Joe’s was guilty of intentionally discriminating against women in hiring servers, it would be liable for intentional discrimination throughout the entire pre-charge period since there is absolutely no evidence that Joe’s adopted new facially-neutral hiring requirements until, at best, the post-charge period when it implemented an objective tray test and started to use a three-person interview panel.
Having said all this, we reiterate that nothing in this record supports a disparate impact theory of liability. Rather, much of the district court’s findings (as well as the credited record evidence), may be read to support the conclusion that Joe’s employment practices in hiring servers were really permeated with an unlawful intention to discriminate. None of the district court’s findings support the conclusion that a facially-neutral practice or policy of Joe’s caused its reputation, and there is not a scintilla of evidence in the record to support this notion. In short, under the district court’s findings, it is not the formal mechanics of Joe’s roll-call system or the criteria embedded in its subjective hiring practices, nor its formal delegation of hiring authority to its maitre d’s which kept women from applying to and being hired by Joe’s during the pre- and post-charge periods.
At bottom then, this case really centers around the theory that women refrained from making the “futile gesture,” Teamsters, 431 U.S. at 365-366, 97 S.Ct. 1843, of applying to Joe’s when they knew that Joe’s only hired men as food servers. If Joe’s reputation came from anything causally attributable to the restaurant, it emanated from Joe’s own purportedly discriminatory hiring practices, not from the specific facially neutral practices identified by the district court. While we agree that in some situations evidence of prior historical discrimination may provide relevant background to a contemporary disparate impact challenge,18 the facts of this *1283case may be read to suggest something quite different; he. that Joe’s hiring decisionmakers systematically excluded female applicants from consideration, that over time this male-only preference became common knowledge, and that eventually most potential, qualified, female applicants self-selected out of Joe’s hiring process precisely because of its reputation for intentional sex discrimination. Indeed, the subsidiary factual findings in this case could be read in simple syllogistic form: first, “Old World” fine-dining meant hiring only tuxedo-clad male servers; second, Joe’s sought to emulate “Old World” fine-dining; and finally, Joe’s therefore only hired male servers. If this is what the district court meant to find, it is indicative of something quite different from the theory of disparate impact. But we cannot affirm a disparate impact judgment where the case centers entirely around allegations and evidence of intentional discrimination. The record does not support it, and to do so would unwisely conflate the distinct theories of disparate impact and disparate treatment.
B. Remand
We are left then with two unattractive choices on appeal: first, we can affirm the liability judgment on an alternate theory of Title VII liability such as disparate treatment or pattern or practice discrimination, as the EEOC suggests, or we can remand so that the district court may reconsider its factual findings and conclusions of law. Although the district court’s findings may be read to suggest a pattern or practice on the part of Joe’s to intentionally discriminate on the basis of sex in its hiring of food servers, we are not prepared to draw this conclusion in the face of the district court’s having expressly rejected this theory; rather we think a remand to the district court is the wiser choice.
We reach this conclusion for three principal reasons. First, we are deeply troubled by and unable to easily square the fundamental inconsistency between the district court’s express rejection of the EEOC’s intentional discrimination claim and several of its subsidiary-factual findings that Joe’s hired male servers only in order to create an “Old World” fine dining ambience. At trial, the EEOC primarily argued an intentional discrimination theory of liability. However, as noted, the district court summarily rejected this theory without - analysis. It- unambiguously stated in the opening paragraph of its partial final judgment order that “[b]ased on an evaluation of the evidence, the court finds that the EEOC has not proved-intentional discrimination.” Joe’s Stone Crab, 969 F.Supp. at 730. Later, in its conclusion of law section, the district court reiterated this conclusion observing that “[t]he court finds that the EEOC has not met its burden of proof under disparate treatment analysis.” Id. at 735.
Second, after carefully reading the trial transcript, we believe the district court’s conclusion that the EEOC has not met its burden of proving intentional discrimination may have been based on an erroneous view of Title VII case law. When “ ‘a district court has failed to make a finding because of an erroneous view of the law, the usual rule is that there should be a remand for further proceedings to permit the trial court to make the necessary findings.’ ” Perryman v. Johnson Products Co., Inc., 698 F.2d 1138, 1144 n. 11 (11th Cir.1983) (quoting Pullman-Standard v. Swint, 456 U.S. 273, 291, 102 S.Ct. 1781, 72 L.Ed.2d 66 (1982)). In light of the district court’s seemingly unambiguous findings that “Joe’s has been a ‘male server type’ establishment for the better part of the century” and that “women have systematically been excluded from the most lucrative entry level position, that of server,” Joe’s Stone Crab, 969 F.Supp. at 740, we,emphasize that a finding of disparate treatment requires no more than a finding that women were intentionally treated differently by Joe’s because of or on account of their gender. To prove the discriminatory intent necessary for a disparate .treatment .or pattern or practice claim, a plaintiff need not prove that a *1284defendant harbored some special “animus” or “malice” towards the protected group to which she belongs.19 In the race discrimination context, we recently have explained that “ill will, enmity, or hostility are not prerequisites of intentional discrimination.” Ferrill v. Parker Group, Inc., 168 F.3d 468, 473 n. 7 (11th Cir.1999). In Ferrill, for example, we held a defendant, who acted without racial animus but consciously and intentionally made job assignments based on racial stereotypes, liable for intentional discrimination. See id. The Supreme Court reached a similar conclusion in Goodman v. Lukens Steel Co., 482 U.S. 656, 107 S.Ct. 2617, 96 L.Ed.2d 572 (1987). In Goodman, union members sued their union under 42 U.S.C. § 1981 for intentionally failing to assert race discrimination claims against their employer. In its analysis, the Supreme Court explained that though “there was no suggestion [ ] that the [defendant] held any racial animus against or denigrated blacks generally,” id. at 668, 107 S.Ct. at 2625, the union still could be held liable for racial discrimination “ ‘regardless of whether, as a subjective matter, its leaders were favorably disposed toward minorities.’ ” Goodman, 482 U.S. at 669, 107 S.Ct. at 2625 (citing Goodman v. Lukens Steel Co., 580 F.Supp. 1114, 1160 (E.D.Pa.1984)). Based on this reasoning, we stated in Ferrill, “[t]he Goodman Court clearly held that liability for intentional discrimination under § 1981 requires only that decisions be premised on race, not that decisions be motivated by invidious hostility or animus.” Id., 168 F.3d at 473. While Goodman and Ferrill both involved § 1981 claims, there is no difference in the substantive doctrine of intentional discrimination under Title VII and § 1981.
Simply put, Title VII prohibits “the entire spectrum of disparate treatment of men and women resulting from sex stereotypes,” Los Angeles Dept. of Water & Power v. Manhart, 435 U.S. 702, 708 n. 13, 98 S.Ct. 1370, 55 L.Ed.2d 657 (1978) (quoting Sprogis v. United Air Lines Inc., 444 F.2d 1194, 1198 (7th Cir.1971)), even where the stereotypes are benign or not grounded in group animus. Therefore, if Joe’s deliberately and systematically excluded women from food server positions based on a sexual stereotype which simply associated “fine-dining ambience” with all-male food service, it then could be found liable under Title VII for intentional discrimination regardless of whether it also was motivated by ill-will or malice toward women.20
*1285Moreover, in light of the district court’s findings that “Joe’s management acquiesced in and gave silent approbation to the notion that male food servers were preferable to female food servers,” Joe’s Stone Crab, 969 F.Supp. at 731, and that “what prevailed at Joe’s, albeit not mandated by written policy or verbal direction, was the ethos that female food servers were not to be hired,” Id. at 732, we also emphasize that under our controlling case law, either under a disparate treatment or a pattern or practice theory, Plaintiff need not show that hiring decisions were made pursuant to an express policy or directive from Joe’s owners. It is enough to show in a disparate treatment case that a particular employment decision was made because of sex and in a pattern or practice case that employment decisions were generally made deliberately because of sex, regardless of whether in either context a formal or express policy of discrimination existed from the employer. See Teamsters, 431 U.S. at 336, 97 S.Ct. 1843; Harris v. Shelby County Bd. of Educ., 99 F.3d 1078, 1083 (11th Cir.1996).
The third reason favoring remand is that almost all of the evidence of intentional discrimination came in the form of conflicting witness testimony- subject to lengthy cross-examination. It is clear from the trial record, for example, that several plaintiff witnesses provided testimony, which if credited by the trial court, could support a finding of intentional discrimination. Several witnesses testified that Joe’s management actively discouraged women from applying. Specifically, a former telephone clerk at Joe’s, Cathy Evans, testified that she was told by General Manager Robert Moorehead, among others, to inform women who called about server positions that the restaurant did not hire female servers. In addition, former take-out cook Cassandra Williams testified that she was told by management that the restaurant only hired women to work in the take-out section, that she was told by a waiter that Joe’s did not hire female servers, and that she overheard Roy Garret state that no women were hired to work in the, main dining room. Finally, Barbara Mommsen testified that when she applied for a server position in 1987, she was told by owner Joanne Bass that Joe’s did not hire female servers. It is, equally clear from the trial record that Joe’s owners and- key management personnel, including Joanne Bass and Robert Moorehead, vigorously denied these specific allegations at trial. The district court made no specific findings on the credibility of these witnesses, and did not spécifically resolve these credibility conflicts. We are riot in a position on appellate review to sort through this conflicting witness testimony in regard to Plaintiffs intentional discrimination claims. Under our caselaw, we allot substantial deference to the fact-finder, in this case, the district court, in reaching credibility determinations with respect to witness testimony. See Stano, 51 F.3d at 944 (holding that we defer even *1286beyond clear error review to trial court findings relating to witness credibility determinations).
Since the state of this record is replete with conflicting witness testimony and conflicting conclusions drawn by the district court, the wisest approach, we think, is to remand the case to the factfinder for more detailed findings on the EEOC’s intentional discrimination claims. Only in this way, can we be assured of reaching an outcome truly consonant with the factfinder’s view of the evidence. We therefore abide by the general rule of law that “ ‘a remand is the proper course unless the record permits only one resolution of the factual issue.’ ” Cooper-Houston v. Southern Ry. Co., 37 F.3d 603, 604 (11th Cir.1994) (quoting Kelley v. Southern Pacific Co., 419 U.S. 318, 331-32, 95 S.Ct. 472, 42 L.Ed.2d 498 (1974)); see also DeMarco v. United States, 415 U.S. 449, 450, 94 S.Ct. 1185, 39 L.Ed.2d 501 (1974) (stating that “factfinding is the basic responsibility of district courts, rather than appellate courts”).
Finally, before we remand, we take a moment to explicate in more detail settled law concerning the requirements of Title VII liability based on a finding of intentional discrimination. There are two theories of intentional discrimination under Title VII: disparate treatment and pattern or practice discrimination. Disparate treatment claims require proof of discriminatory intent either through direct or circumstantial evidence. See Harris, 99 F.3d at 1083 (observing that a “ ‘plaintiff must, by either direct or circumstantial evidence, demonstrate by a preponderance of the evidence that the employer had a discriminatory intent’ ” to prove a disparate treatment claim) (quoting Batey v. Stone, 24 F.3d 1330, 1334 (11th Cir.1994)). “Direct evidence is evidence that establishes the existence of discriminatory intent behind the employment decision without any inference or presumption.” Standard v. A.B.E.L. Servs., Inc., 161 F.3d 1318, 1330 (11th Cir.1998) (citing Carter v. City of Miami, 870 F.2d 578, 580-81 (11th Cir.1989)). Absent direct evidence, a plaintiff may prove intentional discrimination through the familiar McDonnell Douglas paradigm for circumstantial evidence claims. To establish a prima facie ease of disparate treatment under this rubric, a plaintiff “must show: (1) she is a member of a protected class; (2) she was subjected to adverse employment action; (3) her employer treated similarly situated male employees more favorably; and (4) she was qualified to do the job.” Maniccia v. Brown, 171 F.3d 1364, 1368 (11th Cir.1999). Once these elements are established, a defendant has the burden of producing “legitimate, non-discriminatory reasons for its employment action.” Holifield v. Reno, 115 F.3d 1555, 1564 (11th Cir.1997) (citing Texas Dep’t of Community Affairs v. Burdine, 450 U.S. 248, 101 S.Ct. 1089, 67 L.Ed.2d 207 (1981)). If such a reason is produced, a plaintiff then has the ultimate burden of proving the reason to be a pretext for unlawful discrimination. See Holifield, 115 F.3d at 1565.21
In contrast, a pattern and practice claim either may be brought by the EEOC if there is “reasonable cause to believe that any person or group of persons is engaged in a pattern or practice” of discrimination, 42 U.S.C. § 2000e-6(a) (1994); see also In re Employment, 198 F.3d at 1310 n. 8, or by a class of private plaintiffs under 42 U.S.C. § 2000e, et. seq., see Cox v. American Cast Iron Pipe Co., 784 F.2d 1546, 1549 (11th Cir.1986). In such suits, the plaintiffs must establish “ ‘that [sex] discrimination was the company’s standard operating procedure.’ ” Cox, 784 F.2d at 1559 (quoting Teamsters, 431 U.S. at 336, 97 S.Ct. 1843); see also Franks v. Bowman Transportation Co., 424 U.S. 747, 772, 96 S.Ct. 1251, 47 L.Ed.2d 444 (1976). To meet this burden of proof, a plaintiff must “prove more than *1287the mere occurrence of isolated or accidental or sporadic discriminatory' acts. It ha[s] to establish by a preponderance of the evidence that [ ] discrimination [is] the company’s standard operating procedure— the regular rather than unusual practice.” Teamsters, 431 U.S. at 336, 97 S.Ct. 1843 (footnote and internal quotation marks omitted). While “pattern or practice cases are a variant of the disparate treatment theory and thus ‘[p]roof of discriminatory motive is critical,’ ” statistical evidence often is used to establish the existence of a pattern or practice. Lujan v. Franklin County Bd. of Educ., 766 F.2d 917, 929 (6th Cir.1985) (quoting Teamsters, 431 U.S. at 335 n. 15, 97 S.Ct. 1843). A plaintiff may establish a pattern or practice claim “through a combination of strong statistical evidence of. disparate impact coupled with anecdotal evidence of the employer’s intent to treat the protected class unequally.” Mozee v. American Commercial Marine Service Co., 940 F.2d 1036, 1051 (7th Cir.1991) We also point out that “direct evidence of an intent to discriminate” may be used to establish a pattern or practice claim. Lujan, 766 F.2d at 929 n. 15. Finally, we observe that in determining pattern or practice liability, the government is not required to prove that any particular employee was a victim of the pattern or practice; it need only establish a prima facie case that such a policy existed.22
Having offered these observations about Title VII law, we remand to the sound discretion of the district court, which has labored so long and diligently, so that it, may make such factual findings and draw-such conclusions of law about the .EEOC’s intentional discrimination claims as it may deem appropriate. We have struggled mightily to avoid a remand. We know full well that much time and expense already has been spent on this case both before the district court and on appeal. However, in the end, we believe that the most just result is to remand to the district court for reconsideration in light of this opinion. In this way, the district court, which has heard all of the witness testimony firsthand, may conduct the relevant factfinding requisite for determining liability on the EEOC’s intentional discrimination claims.23
Ill;
Accordingly, we vacate the district court’s judgment of liability as to the EEOC’s disparate impact claims, and we remand to the district court so that it may reconsider its factual findings and conclusions of law on the EEOC’s intentional discrimination claims in light of this opinion. Because of our holding, we have.no occasion to reach the various issues raised on appeal regarding the propriety, of the-remedies awarded by the district court.
VACATED AND REMANDED.
. The EEOC filed its original complaint on June 8, 1993. Joe's filed a motion for a more definite statement. In response, the EEOC amended its complaint on September 20, 1993.
. In 1986, the daytime maitre d’, Raymond Damiano, was responsible for interviewing and hiring new food servers, with occasional assistance from dinner maitre d' Roy Garret. Beginning in 1987, Anthony Arneson became the daytime maitre d’ and was given sole responsibility for hiring new food servers, although various staff members occasionally assisted him in the interview process. Specifically, Arneson testified that from 1987 to 1990, Night Captain Dennis Sutton, Night Captain George Silas, Lunch Captain Kevin Murphy, and Maitre d' Roy Garrett all sat in on the interview process at various times and offered him their opinions on the prospective servers interviewed.
. After the charge was filed in 1991, management directed Arneson to conduct interviews with a panel composed of Night Captain Dennis Sutton and General Manager Robert Moorehead. Id. at 733. In subsequent years, both male and female employees served on the panel.
. Similarly, Anthony Arneson, the maitre d’ in charge of hiring beginning in 1987, testified that gender was never mentioned by Joe’s managers or employees because of a "perception that people didn’t even think about ... that many fine dining establishments throughout the world have an all male staff.”
. At the liability trial, thirteen witnesses testified that Joe’s had a reputation in the Miami server community for hiring only men as servers. For example, Catherine Stratford testified that she moved from New Jersey to Miami Beach to attend Joe's "roll call” in 1990. After her move, she did not apply because her roommate told her that Joe’s did not hire female servers. In addition, Barbara Mommsen testified that "it was common knowledge that they [Joe’s] didn’t hire women at the time” and that "Joe’s had a reputation of being one of the best restaurant jobs to get if you were a man. As far as being a woman, you need not apply.” The district court later concluded:
As the court detailed in its findings of fact, Joe’s has been a "male server type” establishment for the better part of this century. While women have predominated among Joe's owner/managers, as well as among the laundry, cashiering, and take away staff, women have systematically been excluded from the most lucrative entry level position, that of server. So accepted was this practice that Joe's general manager, Robert Moorehead, candidly admitted that it never occurred to him that something might be wrong when 108 positions were filled sequentially with male applicants between 1986 and 1990. Many witnesses testified that, as a result of this known practice, Joe’s created a reputation for not hiring women servers and thereby induced qualified women to self-select out of the applicant pool at roll call. As a result, the roll call became skewed and remains skewed to the present day.
By failing to address its entrenched reputation and the resulting effect on the roll call, Joe's leaves undisturbed a significant factor causing the ongoing exclusion of women from server positions. Unless and until Joe’s takes reasonable steps to confront the reputation it has created, it will continue to experience skewed applicant pools at its annual roll call.
Id. at 740.
. Specifically, the district court found that, as of 1991, the Rusty Pelican had a 35.9% female wail staff; the 94th Aerosquadron had a 42.1% female wait staff; and Southpointe Seafood had a 29.5% female wait staff. We note, however, that unlike Joe’s, two of these restaurants, the Rusty Pelican and the 94th Aerosquadron, are not located on Miami Beach.
. Although not originally included in the amended complaint, at the conclusion of the trial, the EEOC moved to amend its complaint to include an allegation that the subjective interviewing process had an adverse effect on women. The district court granted the motion and denied Joe’s corresponding motion to strike this claim. Joe’s has not specifically challenged these rulings on appeal.
. Defendant Joe’s raises thirteen separate issues on appeal. These issues can be divided into three broad categories: (1) whether the district court erred in failing to dismiss this action or limit the scope of the judgment based upon procedural defects; (2) whether the district court erred in finding for the plaintiff EEOC on the merits of its disparate impact claims; and (3) whether the district court erred in its calculation of damages and prejudgment interest. We find that Joe's procedural arguments clearly lack merit. Because we conclude that the district court erred in finding Joe’s liable on the merits of its disparate impact claims, we need not address Joe’s damages arguments.
. The features of these two intentional discrimination claims are outlined in detail in Part II, Section B.
. The only real evidence on this point comes from Joe's Chief Financial Officer, Arnold Meyerson, who estimated, that between one and three women applied each year out of a total applicant pool ranging from 80 to 120 people a year. At trial, both sides stipulated to this figure and the district court relied on this stipulation for its findings on actual applicant flow data. See Joe’s Stone Crab, 969 F.Supp. at 734.
. While it is true that during this period Joe’s hired 108 men and zero women, this zero hiring percentage is deceptive. So few women applied to Joe's during this time period (perhaps one to three a year or, at most, fifteen in total as compared to 80 to 120 men or, at most, 600 men in total) that Joe's zero hiring rate is not significantly deviant from what Joe’s female hiring rate ought to be according to laws of random probability (around 1.5% at best). "The mere absence of minority employees in [] [particular] positions does not suffice to prove a prima facie case of discrimination...." Carter v. Ball, 33 F.3d 450, 456 (4th Cir.1994) (citing Moore v. Hughes Helicopters, Inc., 708 F.2d 475, 484 (9th Cir.1983)).
.While we adopt a flexible approach for determining whether a particular statistical deviation is "significant'' for disparate impact analysis in light of all the facts and circumstances, see Kilgo v. Bowman Transp. Inc., 789 F.2d 859, 872-73 (11th Cir.1986), on appeal, the EEOC concedes that there is no legally significant statistical disparity in either relevant time period when actual applicant flow data is used. Both parties’ statistical experts agreed that women needed to comprise five percent of the actual applicant class (they comprised between one and three percent) to have generated a significant disparate impact in Joe's pre-charge actual hiring pool.
. On appeal, Joe's challenges the district court’s use of this alternative labor data as clear error. See Kilgo, 789 F.2d at 869-70 & n. 10 (suggesting that the decision to reject actual applicant market data for alternative labor market data is necessarily fact-intensive and reviewed for clear error)., Joe’s contends that the district court improperly relied on hearsay reputation evidence in finding the actual applicant flow data unreliable. Joe’s also argues that the district court erred in using "refined” census data to define the relevant labor market for Miami Beach food servers qualified to work at Joe’s. The district court narrowed the labor pool to those Miami Beach food servers who made between $25,000 and $50,000 — a salary range comparable to what Joe’s food servers earn. Joe’s claims that the district court’s methodology of using income as a proxy for job qualifications in no way assures that the labor pool actually consists only of those food servers who have the special food-serving skills requisite to work at Joe’s. In the disparate impact context, we have explained that the "[djefinition of a qualified applicant pool will shift with the nature of the job or job benefit, and the nature of the challenged employment practice at issue.” In re Employment, 198 F.3d at 1312. We also have observed that" '[w]hen special qualifications are required to fill particular jobs,’ ” the use of certain statistics such as general population figures "becomes troublesome.” Id. at 1313 (quoting Hazelwood, 433 U.S. at 308 n. 13, 97 S.Ct. 2736); see also Alexander v. Fulton County, Ga., 207 F.3d 1303, 1327-28 (11th Cir.2000) (noting that class-based disparate treatment statistics showing that a given minority is underrepresented in the work force by comparison with the general population is generally useful only for claims involving jobs with low skill levels where the applicant pool can be considered roughly coextensive with the general population); Peightal v. Metropolitan Dade County, 26 F.3d 1545, 1554 (11th Cir.1994) (stating that "for positions requiring minimal training or for certain entry level positions, statistical comparison to the racial composition of the relevant population suffices, whereas positions requiring special skills necessitate a determination of the number of minorities qualified to undertake the particular task”). We have had no prior occasion to determine specifically whether salary is an adequate proxy for food server job qualifications at a fine dining establishment. Because we conclude infra that the district court erred in finding that the EEOC established disparate impact discrimination, even if we accept the alternative labor pool data, we need not address whether the alternative labor pool selected by the district court was comprised of "qualified” potential applicants.
. When this same 31.9% figure was applied to Joe’s post-charge hiring statistics, a slight disparity, bordering on the significant, was found. The statistical variation was between 1.96 to 2.07 under "standard deviation” analysis. However, no particular numerical deviation is required to establish a- prima facie case; instead courts employ a case-by-case approach dependent on the particularized case facts. See Watson, 487 U.S. at 995 n. 3, *1278108 S.Ct. 2777. In addition, our caselaw has recognized that post-charge hiring behavior is less probative than pre-charge conduct because a business may be improving its hiring practices to avoid liability or large damages in their pending discrimination case. See James v. Stockham Valves & Fittings Co., 559 F.2d 310, 325 n. 18 (5th Cir.1977); Rowe v. General Motors Corp., 457 F.2d 348, 359 (5th Cir.1972).
. In addition, because of the district court’s own causation reasoning we must focus our inquiry on the alleged causes of the gender disparity in both Joe’s applicant and hire pool. According to the district court, both disparities necessarily share the same cause since it was the disparity in Joe's applicant pool which directly led to the disparity in Joe’s hire pool. Under the district court’s factual findings, Joe’s public reputation as a sex discriminator caused (1) a gender disparity in Joe’s applicant pool which, in turn, created (2) a gender disparity in Joe's hire pool because there were so few women who actually applied, to Joe’s as a food server. Therefore, in order to understand the causes of Joe’s hiring pool disparity, it is essential to determine the’ causes of Joe's applicant pool disparity.
. This case differs significantly from the paradigmatic disparate impact case in which the plaintiffs show direct causation between an objective hiring requirement and the statistical disparity at issue. In Griggs, for example, the plaintiffs showed that the objective and facially neutral requirements of possessing a high school diploma and passing a general intelligence test in order to be hired or transferred to the company's more desirable departments had a disproportionate effect on white and black applicants. See Griggs, 401 U.S. at 430-431, 91 S.Ct. 849.
. We have made clear that employment decisions may legitimately be based on subjective criteria as long as the criteria are capable of objective evaluation and are stated with a sufficient degree of particularity. See Conner V. Fort Gordon Bus Co., 761 F.2d 1495, 1500 (11th Cir.1985); Woody v. St. Clair County Comm'n, 885 F.2d 1557, 1562-63 (11th Cir.1989).
. For example in Griggs the Supreme Court made clear that Title VII prohibited an employer from using neutral hiring and promotion practices to “freeze'' in place a status quo achieved through prior decades of intentional discrimination. See Griggs, 401 U.S. at 430, 91 S.Ct. at 853. In Castaneda v. Pickard, the former Fifth Circuit explained that it is appropriate to look beyond actual data to labor force data in order to establish disparate impact liability where the employer has used discriminatory recruiting practices which can "skew the ethnic composition of the applicant pool.” Id., 648 F.2d 989, 1003 (5th Cir. Unit A 1981).
. Several ambiguous phrases in the district court's opinion suggest that the district court may have been operating under this erroneous view. For instance, in finding that Joe’s did not have an express policy of excluding women from food server positions, the district court observed that the evidence only proved that "management acquiesced in and gave silent approbation to the notion that male food servers were preferable to female food servers.” Joe's Stone Crab, 969 F.Supp. at 731. Of course, the district court’s summary dispatch of the EEOC’s intentional discrimination claims precludes us from knowing for certain whether it thought the EEOC had to prove that Joe's hiring policies emanated from some special "animus” or "malice” towards women.
. Perhaps the seminal case in this context is Price Waterhouse v. Hopkins, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (1989). In Price Waterhouse, the district court found that the employer, an accounting firm, had discriminated against Ann Hopkins by permitting stereotypical attitudes about women (Hopkins had been found "too acerbic” in her tone for a woman) to play a role in its decision to deny her partnership in the firm. See Hopkins v. Price Waterhouse, 618 F.Supp. 1109, 1118-19 (D.D.C.1985). On appeal, the employer contended that it could not be liable for disparate treatment because "Hopkins did not prove 'intentional' discrimination on the part of the [decision-making] Board, but only unconscious’ sexual stereotyping by unidentified partners who participated in the selection process.” Hopkins v. Price Waterhouse, 825 F.2d 458, 468 (D.C.Cir.1987), aff'g in part and rev’g in part, Hopkins, 618 F.Supp. at 1113-21. This argument was squarely rejected by the court of appeals:
In keeping with [Title VII's remedial] purpose, the Supreme Court has never applied the concept of intent so as to excuse an artificial, gender-based employment barrier simply because the employer involved did not harbor the requisite degree of ill-will towards the person in question. As the *1285evidentiary framework established in McDonnell Douglas makes clear, the , requirement [ ] of discriminatory motive in disparate treatment cases does not function as a “state of mind” element, but as a method of ensuring that only those arbitrary or artificial employment barriers that are related to an employee or applicant’s race, sex, religion, or national origin are eliminated.
Id. at 468-69 (footnotes omitted). This reasoning was affirmed on appeal by a plurality opinion of the Supreme Court. Price Waterhouse, 490 U.S. at 250-52, 109 S.Ct. 1775 (plurality opinion); id. at 259, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (White, J., concurring in the judgment); id. at 261, 272, 277-78, 490 U.S. 228, 109 S.Ct. 1775, 104 L.Ed.2d 268 (O’Connor, J., concurring in the judgment); see also Hopkins v. Price Waterhouse, 920 F.2d 967, 969 (D.C.Cir.1990) (affirming district court decision after remand from Supreme Court and noting that this portion of Hopkins, 825 F.2d at 468-69, was upheld by the Supreme Court). Notably, the Price Waterhouse plurality explained that "[i]n the specific context of sex stereotyping, an employer who acts on the basis of a belief that a woman cannot be aggressive, or that she must not be, has acted on the basis of gender," and therefore has committed intentional discrimination under Title VII. Price Waterhouse, 490 U.S. at 249, 109 S.Ct. 1775 (emphasis added).
. In Reeves v. Sanderson Plumbing Products, Inc., — U.S. —, 120 S.Ct. 2097, 2108-09, 147 L.Ed.2d 105 (2000), the Supreme Court explained that "a plaintiffs prima facie case, combined with sufficient evidence to find that the employer’s asserted justification is false, may permit the trier of fact to conclude that the employer unlawfully discriminated.”
. Once this pattern or practice is established, the burden of proof then shifts to the employer to demonstrate that the government's showing of a pattern or practice of discrimination is either inaccurate or insignificant. If the employer fails to rebut the government’s prima facie case, the resulting finding of a discriminatory pattern or practice may give rise to an inference that all persons subject to the policy were its victims and are entitled to appropriate remedies. As we have explained previously, "once a pattern and practice of discrimination is established, a rebuttable presumption that [the] plaintiff was discriminated against because of her sex and is entitled to recoveiy obtains. The employer may overcome this presumption only with clear and convincing evidence that job decisions made when the discriminatory policy was in force were not made in pursuit of that policy.” Cox, 784 F.2d at 1559 (citing Teamsters, 431 U.S. at 362, 97 S.Ct. 1843).
. We stress that our remand in no way obligates the district court to hear additional evidence or argument in the case. Of course, the district court may choose to reopen trial proceedings for a limited purpose or conduct further evidentiary hearings if it feels such measures are necessary for the resolution of the case. . We leave these procedural decisions to the sound discretion of the district court.