This case involves step two of the familiar three-step analysis for Title VII disparate impact claims. The district court found first that appellant had satisfied his initial burden of identifying an employment practice with a discriminatory impact. The district court next found that appellee had avoided a finding of discrimination against it by satisfying its burden of demonstrating that the practice was a business necessity and had a manifest relationship to the employment in question. Because we hold that appellee has not satisfied this burden, we reverse without considering the third step in the analysis. We remand the case to the district court for the sole purpose of granting relief to the plaintiff as instructed in this opinion.
I.
Appellant Winston Nash visits this court a second time. The first round of proceedings is set out in detail in Nash v. City of Jacksonville, 763 F.2d 1393 (11th Cir.1985). Appellant, a black fire engineer in the City of Jacksonville fire department, applied for a promotion to combat lieutenant in 1976, 1978 and 1981. The City requires a score of 70 or higher on the (combat) promotion examination in order to be considered for promotion. Appellant scored an 84 on the test in 1975 and was placed on the promotion eligibility list. The list expired, however, before the City promoted him. He scored 69 and 69.072 in his next two attempts, both below the cutoff score of 70.
The test was prepared by a committee of two fire chiefs and one captain. Although the City states that supervisory ability is necessary for the position of combat lieutenant, the City admits that there are no questions on the examination relating to supervisory skills because such skills cannot be tested by written examination.1 Only three of the twenty blacks who had taken the test at the time of the trial had passed (none was promoted), while 57 of the 100 whites who had taken the test had passed. Because this violates the EEOC’s “four-fifths rule”, see 29 C.F.R. § 1607.4(D),2 and because appellant’s ex*1536pert produced additional statistical evidence, the district found that appellant had stated a prima facie case of discriminatory impact. See Nash, 763 F.2d at 1395.
The test questions themselves, however, never made their way into evidence. Neither appellant’s expert witness nor the district court ever saw the questions. Appellant’s expert concluded from indirect evidence that some of the test questions showed a disparate impact. Appellant’s expert also testified that, based on his statistical analysis, it was impossible to “even conclude that the examination selected people who best performed on the examination, much less who ... would best perform on the job.” The only witness who actually reviewed the test questions was the City’s former personnel examiner and current retirement pension manager, who also served as the City’s expert at trial. He refuted the conclusions of appellant’s expert on the grounds that it was necessary to consider the actual questions (which appellant’s expert was not allowed to do) in order to determine whether or not the questions were job-related. The district court credited the City’s evidence, discredited the testimony of appellant’s expert, and ruled for the City. The court, still without ever seeing the examination, found that it was content-valid and job-related.3
On appeal this court subtly observed that “[h]ad the burden of persuasion as to ‘content-validity’ been placed where it should have been, on defendant, we think defendant might have felt it incumbent to produce the examination questions.” 763 F.2d at 1398. This court then reversed the district court, “expressing] wonder that any court should attempt to determine whether a written examination was ‘job-related’ and ‘content-valid’ without having before it the questions asked of the applicants.” Id. Our wonder has now ripened into full incredulity, because on remand the district court ruled again for appellee, still without ever seeing the test questions. This appeal therefore forces us to abandon subtlety in applying the law under Title VII to the facts of this case.
II.
Once an employee has demonstrated that a facially neutral employment or promotion practice has a discriminatory impact, the burden shifts to the employer to demonstrate that the practice has a manifest relationship to the employment in question or is a business necessity. See Connecticut v. Teal, 457 U.S. 440, 446-47, 102 S.Ct. 2525, 2530, 73 L.Ed.2d 130 (1982); Griggs v. Duke Power Co., 401 U.S. 424, 432, 91 S.Ct. 849, 854, 28 L.Ed.2d 158 (1971); Craig v. Alabama State Univ., 804 F.2d 682, 685 (11th Cir.1986); Walker v. Jefferson County Home, 726 F.2d 1554, 1558 (11th Cir.1984). If the employer meets this burden, the employee may still prevail by establishing that the employer used the practice as a mere pretext for discrimination. Teal, 457 U.S. at 447, 102 S.Ct. at 2530; see Dothard v. Rawlinson, 433 U.S. 321, 329, 97 S.Ct. 2720, 2726, 53 L.Ed.2d 786 (1977); Albemarle Paper Co. v. Moody, 422 U.S. 405, 425, 95 S.Ct. 2362, 2375, 45 L.Ed.2d 280 (1975); Craig, 804 F.2d at 685. If the employer fails to meet this burden then discrimination has been proved, and there is no need to examine pretext or to consider the existence of less discriminatory alternative practices. The employee is entitled to relief. See Albemarle Paper, 422 U.S. at 436, 95 S.Ct. at 2380; Crawford v. Western Elec. Co., Inc., 745 F.2d 1373, 1386 (11th Cir.1984); Walker, 726 F.2d at 1559.
There is no dispute that the district court correctly found that appellant proved a pri-ma facie case of discrimination. The district court found, however, that the City successfully carried its burden of proving that the promotion examination was job-related and a business necessity.4 Under this *1537burden the employer must show “that the challenged employment practice has a ‘manifest relationship’ to the employment in question, or in other words that it is supported by ‘business necessity.’ ” Craig, 804 F.2d at 685 (citing Griggs, 401 U.S. at 431-32, 91 S.Ct. at 853-54).
A.
The district court found that the test was job-related. This finding is reversible either if it is clearly erroneous or if the district court erroneously applied the law. Nash, 763 F.2d at 1396-97. The district court based its conclusion on the procedure used to construct the test and on the fact that appellant failed to produce evidence that the examination was not job-related. This conclusion, however, was based on both legal errors and clearly erroneous factual findings.
The district court first committed reversible error by repeating the same legal error it committed in its first ruling. In Nash this court instructed the district court that in Title VII disparate impact cases the burden is on the employer to prove job-relatedness. 763 F.2d at 1397. This court wrote that in an impact case the employer must do more than “articulate” non-discriminatory reasons for its employment practice. Id. This court noted that the district court repeatedly stated that the plaintiff “has failed to persuade the court ‘by a preponderance of the evidence.’ The idea seems to be that mere ‘articulation’ shifts the burden back to plaintiff to persuade that the test is discriminatory, but this is contrary to the authorities in this circuit as to ‘impact’ employment test cases.... ” Id. at 1398.
On remand, the district court committed the same error. Finding the test job-related, the district court wrote in its memorandum opinion: “Evidence that the examination was not job-related was noticeably lacking. Plaintiff’s expert never correlated the test with successful job performance....” As we thought we had made clear, once the plaintiff made out his prima facie case of disparate impact, it was the defendant’s burden to show that the test was job-related, not the plaintiff’s burden to show the test was not job-related.
The district court also committed reversible error by allowing the City’s evidence of how the test was prepared to satisfy its burden of proving that the content of the test was related to job performance. Just because test questions are drafted by qualified city employees does not mean that the questions are job-related. Even qualified supervisory personnel can construct test questions very unrelated to job performance. The fact that the City presented evidence relating to how the test was constructed is not legally sufficient by itself to show the questions were job-related.
Job-related test questions are those that are “shown, by professionally acceptable methods, to be ‘predictive of or significantly correlated with important elements of work behavior which comprise or are relevant to the job or jobs for which candidates are evaluated.’ ” Albemarle Paper, 422 U.S. at 431, 95 S.Ct. at 2378 (quoting 29 C.F.R. § 1607.4(c)); see Police Officers for Equal Rights v. City of Columbus, 644 F.Supp. 393, 414 (S.D.Ohio 1985). In this case the City presented no evidence at all that the test questions were job-related in any way. The City did not present the actual test questions and the City’s expert testified that he did not evaluate the test questions as to whether or not they were job-related.5 Job-relatedness can be established through the testimony *1538of expert witnesses supported by a validation study. See, e.g., Black Law Enforcement Officers Ass’n v. City of Akron, 824 F.2d 475, 480-81 (6th Cir.1987); Davis v. City of Dallas, 111 F.2d 205, 209-10 (5th Cir.1985), cert. denied, 476 U.S. 1116, 106 S.Ct. 1972, 90 L.Ed.2d 656 (1986); Gillespie v. Wisconsin, 111 F.2d 1035, 1041-44 (7th Cir.1985), cert. denied, 474 U.S. 1083, 106 S.Ct. 854, 88 L.Ed.2d 894 (1986); Contreras v. Los Angeles, 656 F.2d 1267, 1280-84 (9th Cir.1981), cert. denied, 455 U.S. 1021, 102 5.Ct. 1719, 72 L.Ed.2d 140 (1982); Brunet v. City of Columbus, 642 F.Supp. 1214, 1246-47 (S.D.Ohio 1986). In this case, however, there was no validation study conducted by an expert.6 Without the test itself and with no testimony presented by the defendant regarding the job-relatedness of the questions, there was no way for the district court to conclude that the questions were job-related.
The best evidence on job-relatedness of the questions came from the testimony of appellant’s expert, who testified that from his statistical analysis of the test results and his study of the materials he received from the City, the test was “not capable of being validated, because there’s too many items that cannot be validated.”7 Indeed the City admitted through the testimony of its current personnel examiner that the examination did not test the one aspect of job performance that differentiated the job of firefighter engineer from fire lieutenant (combat): supervisory skills. An officer’s job in a fire department involves “complex behaviors, good interpersonal skills, the ability to make decisions under tremendous pressure, and a host of other abilities— none of which is easily measured by a written, multiple choice test.” Firefighters Inst. for Racial Equality v. City of St. Louis, 616 F.2d 350, 359 (8th Cir.1980), cert. denied, 452 U.S. 938, 101 S.Ct. 3079, 69 L.Ed.2d 951 (1981). It is a position “which involves tasks and characteristics which cannot be tested exclusively by such tests.” Vulcan Pioneers, Inc. v. New Jersey Dep’t of Civil Serv., 625 F.Supp. 527, 547 (D.N.J.1985), aff'd 832 F.2d 811 (3d Cir.1987). The best that can be said of the City’s test based on the evidence at trial was that it may have been valid with respect to reading materials provided to the *1539applicants. This is immaterial, however, to whether the content of the questions related to the performance of the job. See Vulcan Pioneers, 832 F.2d at 815-16; see also 29 C.F.R. § 1607.14(C)(4) (a test that purports to sample work behavior should closely approximate the work situation). The district court’s finding that the examination relates to job performance is therefore clearly erroneous.
B.
The district court also found that the City’s promotion examination was a business necessity. “The business necessity doctrine is ‘very narrow,’ and the employer bears the burden of not only articulating but also proving business necessity through the evidence.” Craig 804 F.2d at 689 (quoting Parson v. Kaiser Aluminum & Chem. Corp., 575 F.2d 1374, 1389 (5th Cir.1978), cert. denied, 441 U.S. 968, 99 S.Ct. 2417, 60 L.Ed.2d 1073 (1979)) (citation omitted) (emphasis in original). This court stated the test most recently in Craig:
The test is whether there exists an overriding legitimate business purpose such that the practice is necessary to the safe and efficient operation of the business. Thus, the business purpose must be sufficiently compelling to override any racial impact; the challenged practice must effectively carry out the business purpose it is alleged to serve; and there must be available no acceptable alternative policies or practices which would better accomplish the business purpose advanced or accomplish it equally well with lesser differential racial impact.
804 F.2d at 689 (quoting Pettway v. American Cast Iron Pipe Co., 494 F.2d 211, 245 (5th Cir.1974), cert. denied, 439 U.S. 1115, 99 S.Ct. 1020, 59 L.Ed.2d 74 (1979)); see Crawford, 745 F.2d at 1384-85.
The district court found that the City had “an obligation to demand and pursue excellence in its Public Safety Department.” While this statement is correct, it is not a basis for finding that the test at issue was a business necessity. The City’s expert admitted that 27 of the 97 questions on the 1981 examination had an adverse impact on black applicants. Thus while the City has a business necessity in ensuring that it promotes only qualified firefighters to lieutenant, it has no business purpose at all for using an improper means to do so. The City presented no evidence on why a racially discriminatory test is a business necessity. See Crawford, 745 F.2d at 1385-86 (no business necessity for the mechanics of employer’s employment practice, even though theoretically the practice related to safety and efficiency). The district court’s finding that the test was a business necessity is therefore clearly erroneous.
III.
We therefore reverse the district court’s judgment for the City and hold that because the City failed to meet its burden appellant is entitled to judgment. Where the plaintiff has established a prima facie case, and the defendant fails to show the job-relatedness of an examination with discriminatory impact, the issues are not open for redetermination on remand. Bunch v. Bullard, 795 F.2d 384, 395 (5th Cir.1986); see Crawford, 745 F.2d at 1386. We therefore remand the case to the district court for the entry of judgment on liability.
Plaintiff is entitled to the position of fire lieutenant (combat), with the seniority and benefits accorded a lieutenant as of the date the first lieutenant was selected for promotion from the eligibility list for the 1981 examination. We remand for judgment to that effect, for a determination of the amount of back pay and for such other relief as may be appropriate.
REVERSED and REMANDED with instructions.
. The development of the promotion examination is set out in detail in Nash, 763 F.2d at 1395-96.
. 29 C.F.R. § 1607.4(D) provides in part:
Adverse impact and the “four-jijths rule. ” A selection rate for any race, sex, or ethnic group which is less than four-fifths (Vs) (or eighty percent) of the rate for the group with the highest rate will generally be regarded by the Federal enforcement agencies as evidence of adverse impact....
. The district court completed the third part of the analysis and found that there were no less discriminatory ways for the City to select fire engineers for promotion.
. At the first trial, the district court only required the City to "articulate” some legitimate, non-discriminating reasons, rather than carry the burden of persuasion. This is contrary to *1537the law of disparate impact cases. See Nash, 763 at 1397-98.
. The transcript of the trial reads:
Q (counsel for defendant): Did you evaluate the examination as to whether or not it related to job performance?
A (defendant’s expert): As far as job performance, no.
The defendant’s expert also testified that he was not familiar with the principles of validation and the use of personnel selection procedures set forth by the American Psychological Association. The EEOC’s guidelines, 29 C.F.R. § 1607, are based on these guidelines. Albemarle, 422 U.S. at 430-31 & n. 29, 95 S.Ct. at 2378 & n. 29.
. The City’s personnel examiner assigned to the 1981 examination claims to have completed a content-validity study on the examination. There is no evidence, however, and the district made no finding, that the City's promotion test was validated under the EEOC guidelines. The EEOC guidelines, though not substantive regulations promulgated pursuant to procedures established by Congress, "are 'entitled to great deference.’” Albemarle, 422 U.S. at 431, 95 S.Ct. at 2378 (quoting Griggs v. Duke Power Co., 401 U.S. 424, 433-34, 91 S.Ct. 849, 854, 28 L.Ed.2d 158 (1971)). Several courts have held that certain fire department examinations must be validated in conformity with the guidelines in order to prove job-relatedness. See, e.g., Brunet, 642 F.Supp. at 1245.
Appellant presented evidence that the City is in violation of several requirements for EEOC validation. See, e.g., 29 C.F.R. §§ 1607.5(G), 1607.14(C)(9) (use of selection procedure on a ranking basis); id. at §§ 1607.5(H), 1607.-15(C)(7) (use of ranking system and cutoff scores); id. at § 1607.15(C) (documentation of content validity); id. at § 1607.5(1) (use of selection procedures for promotion). In any event, we believe that an analysis of the job-relatedness of the City’s test conducted by the City itself was not legally sufficient to carry the City’s burden. See Albemarle, 422 U.S. at 433 n. 32, 95 S.Ct. at 2379 n. 32. The City's expert, who testified that he had not conducted a content-validity study, was also a City employee. Had he evaluated the job-relatedness of the test questions, his testimony would be subject to examination with great care. Id.
. The transcript of the trial reads:
Q (counsel for plaintiff): From your analysis of all the documentation and material that was supplied to you, ... what is your opinion concerning the validity of the selected procedure used? The test, in other words?
A (plaintiff’s expert): From the documents that were given me, it appeared that there was no reason why one would think that they could have constructed a valid test, given the way they went about doing the job analysis. And given the fact that the test was constructed by listing reading material and selecting questions out of the reading material, rather than directly constructing the test to reflect the job, itself — and that was very clear in the materials, that that was the method of making the test — so that even at that point it appeared that what the test was, was some kind of quasi-academic hurdle. [The test did not] measur[e] performance on the job and us[e] performance on the job as the predictor of who had the most knowledge and information and ability and skill to perform the job....