HARRELL, Judge.
Prabhjot S. Kohli, appellant, filed a complaint with the Maryland Commission on Human Relations (“the Commission”), alleging religious discrimination arising out of the failure of LOOC, Inc. d/b/a Domino’s Pizza (“LOOC”), and Domino’s Pizza, Inc. (“DPI”),1 appellees, to hire him due to his refusal to shave the beard he wore for religious purposes. An Administrative Law Judge (ALJ) from the Office of Administrative Hearings determined that appellees had failed to meet their burden of demonstrating that they would suffer undue hardship if they were required to accommodate appellant’s religious practice. The Appeal Board of the Commission reversed the Provisional Order of the ALJ and dismissed the complaint. The Circuit Court for Baltimore County affirmed the Appeal Board, prompting this appeal.
[698]*698
ISSUES
Appellant presents the following challenges to the Appeal Board’s decision:
I. Did the Appeal Board disregard arbitrarily, capriciously and erroneously as a matter of law the standard of review that it was legally bound to apply, when it set aside the administrative law judge’s conclusion that appellees had failed to prove that it would suffer undue hardship if it accommodated appellant by permitting him to wear a beard net or snood2?
II. Is the Appeal Board’s decision supported by competent, material and substantial evidence, in light of the entire record as submitted, insofar as it concluded that appellees had met their burden of demonstrating that they could not accommodate appellant’s religion without undue hardship on the conduct of. their business?
We answer appellant’s first question in the affirmative, that is that the Appeal Board, in overturning the decision of the Administrative Law Judge, disregarded its own standard of review as set forth in the rules of procedure of the Commission on Human Relations. Accordingly, we shall reverse without reaching appellant’s other query.
FACTS
On 14 December 1987, Prabhjot S. Kohli, appellant, applied for a managerial position with one of LOOC’s Domino’s Pizza stores in Baltimore County. Mr. Kohli was granted an interview, but was denied employment after he refused to shave off [699]*699the beard he wore for religious reasons.3 Appellant explained to his prospective employer that cutting one’s body hair is forbidden under the tenets of Sikhism, the religion of which he was a devout member.4 Nevertheless, the LOOC official conducting the interview informed Kohli that he could not be offered the position because of the company’s policy, promulgated by DPI, that did not permit the hiring of bearded, persons.
Domino’s has implemented strict grooming standards that, without exception since 1980, prohibit its employees from wearing beards. DPI’s stated reason for having a no-beard policy is based upon its concern that hair may get into food, as well as its belief that a customer may be reluctant to purchase from a food service establishment whose employees are not clean shaven. Compliance with the policy is considered especially important for Domino’s store managers, whose job requires constant contact with both food and the public. The manager is responsible for overseeing the store’s food-making, baking pizzas, oven tending, and delivery. In addition, managers answer telephones, deliver pizzas, and handle customer [700]*700complaints. The policy is further grounded in the belief that the corporate image will be best promoted through a uniformity of its market image as well as consistent standards for the professional appearance of its employees. The express concern is that to allow certain employees to maintain beards would interfere with the corporate reputation and hurt sales.
Appellant requested an accommodation from LOOC due to his religion. The interviewer responded that the grooming policy was strictly enforced and that no deviation was possible. Mr. Kohli thereafter telephoned DPI’s corporate headquarters office on 22 December to explain his situation, but received the same response. That office informed appellant that he could take up the matter with DPI’s legal department, which prompted Kohli to mail a letter reiterating his request for an accommodation. That request was also denied, despite his willingness to wear a beard snood or net while working, in lieu of having to shave off his facial hair. .
In February 1988, Mr. Kohli filed a complaint against Domino’s with the Maryland Commission on Human Relations. General counsel for the Commission filed a statement of charges on behalf of Mr. Kohli, alleging that Domino’s had engaged in unlawful employment discrimination in declining to hire appellant. Mr. Kohli was permitted to intervene, and was represented by his own counsel. Specifically, the complaint alleged that appellees’ refusal to accommodate appellant’s religious belief by permitting an exception to their grooming policy constituted an act of religious discrimination in violation of Md.Code Ann., Art. 49B, § 16(a). That section provides, in pertinent part, that “[i]t shall be an unlawful employment practice for. an employer: (1) To fail or refuse to hire ... any individual ... because of such individual’s ... religion ...” Section 15(f) of the same subtitle defines religion as “including] all aspects of religious observances and practice, as well as belief, except in those cases when the observance, practice, or belief cannot be reasonably accommodated by an employer without causing undue hardship on the conduct of the employer’s business.” In response, Domino’s contended that its no-beard policy was justified by business necessity, [701]*701and that Mr. Kohli could not be accommodated without causing undue hardship to the companies’ business.
The Administrative Law Judge’s Decision
A public hearing was conducted before an ALJ on 10-13 June 1991. Thirteen witnesses testified at the hearing, and numerous exhibits were admitted into evidence. Domino’s did not challenge appellant’s protected status. The ALJ initially found, in her 30 March 1992, forty-seven page provisional order and opinion, that the Commission had established a prima facie case of religious discrimination by showing that (1) Kohli belonged to a protected class; (2) he had applied for and was qualified for the job sought; (3) he was rejected for the position despite his qualifications; (4) he held a bona fide religious belief, which he communicated to the employer; and (5) the employer continued to solicit applications from other qualified applicants following Kohli’s rejection.
In considering whether appellees had met their burden of demonstrating that the accommodation of appellant would have caused them undue hardship, the ALJ heard testimony from various expert and corporate witnesses that business would be lost if Domino’s were to provide an accommodation. In support of appellees’ position, DPI presented the results of two studies it had commissioned to determine the attitudes of customers toward employees wearing beards (“The Beard Perceptions Study”) and those wearing a beard snood (“The Beard Snood Perceptions Study”). William Denck, a market research consultant from MARC (the firm that conducted the studies), testified as the sole expert concerning the conclusions to be drawn from the surveys. The ALJ made the following pertinent findings of fact with respect to these studies (we shall use the ALJ’s number identification for each of her findings of fact discussed in our opinion):
67. The Beard Study was conducted by MARC, a marketing research company, through use of a questionnaire. The company selected a sample of the country, consisting of 600 males and females, aged 18-65, who had purchased pizza from a restaurant within three months. The margin of [702]*702error for the study is +3.9%. The participants in the study were asked to respond to questions which focused on three types of beards (2 inch beard, inch beard, and 1mm stubble).
68. Results of the study showed that customers felt that beards of varying lengths on pizza employees were troublesome. However, 55% of the participants stated that they would purchase pizza from a bearded employee. Forty-five percent of the participants stated that they would be more likely to purchase pizza from a restaurant where all the employees were clean shaven. These participants identified sanitation and cleanliness as factors affecting their future purchase intent.
71. In 1990, DPI commissioned another beard survey from MARC, known as the Beard Snood Perception Study (Snood Study). The purpose of the Snood Study was to determine the future purchase intent of customers if they saw an employee wearing a beard snood over a beard. The Snood Study was conducted at six shopping malls throughout the United States to achieve geographical diversity. Using a two-cell design, 150 people were exposed to a clean shaven Domino’s employee, while 150 people were shown a picture of the same employee with a beard and beard snood. The purpose of the cell with the clean shaven employee was to control the participants’ reaction to the employee’s appearance for reasons other than the beard snood. The results of the study showed that, at the 99% level of confidence, that [sic] 16% of the participants said that they definitely/probably would not buy pizza in the future if they saw a bearded employee wearing a beard snood. 7.3% of the participants shown the picture of a clear [sic] shaven employee said they definitely/probably would not buy pizza from the employee.
The ALJ also recognized Domino’s use of in-house programs to gauge its customer perceptions of the company’s operation:
73. DPI uses a mystery program comprised of customers to rate its stores. Each mystery customer receives a ques[703]*703tionnaire asking questions about the product and service. One section relates to grooming and appearance. Specifically, question 19 asks about neatness and cleanliness. During a recent period, 19% of the participants made negative comments on neatness and cleanliness, some of which included references to the fact that the driver was not clean shaven.
Despite the results of these investigations indicating that there might be some impact on future purchases by customers faced with employees wearing beards and beard-restraining devices, the ALJ concluded in her Provisional Order and Opinion that Domino’s had failed to show by a preponderance of the evidence that it would suffer undue hardship if it accommodated Mr. Kohli. As a basis for her decision, the ALJ found the Beard Perceptions Study to be of limited value because its focus was solely upon the condition of being shaven or unshaven—it did not address customer attitudes toward employees who wore beard restraints. According to the ALJ, the lack of such additional evidence prevented the study from logically supporting DPI’s contention that granting appellant an accommodation by allowing him to wear a beard net or snood would have resulted in lost business. She furthermore found the study to be without scientific validation, opining that DPI had neither validated the sampling technique used nor explained how reliability was ensured.
The ALJ minimized the significance of the Beard Snood Study as well, noting that the survey did not portray significant statistical differences between customer attitudes nationwide and those in the Baltimore area. The Provisional Order and Opinion also made reference to the following testimony of Mr. Denck:
Q: Can you tell me whether there was a statistically significant difference between the 25 people [in Baltimore] who were shown a picture of the beard with the snood over it, and the 25 who were not—who were shown a picture of the man without a beard?
[704]*704A: No, I’m not prepared to do that today, which was outside the scope of the study.
The ALJ therefore gave little weight to the survey as conducted because it did not compare the opinion of customers served by a snooded versus a clean shaven employee, and, in addition, because it did not inquire into customer reaction to an employee wearing a beard net as opposed to a snood.5 Similarly, the ALJ observed that DPI’s Mystery Customer Program failed to inquire into customer opinions toward employees who wore beard restraints.
Finally, the ALJ opined that Domino’s had failed to demonstrate that it would lose its competitive edge were it required to make an accommodation. The ALJ concluded that Domino’s had presented insufficient evidence to sustain its argument that competitive losses would have occurred if it were to make an exception to its grooming policy, especially in view of testimony that its chief competitor, Pizza Hut, had itself made such exceptions in the past.
Decision of the Appeal Board
Domino’s appealed the ALJ’s decision to an Appeal Board of the Human Relations Commission. In seeking review of the AL J’s conclusions, appellees did not dispute that appellant had established a prima facie case of discrimination. Rather, the pertinent question asserted before the Commission was as follows:
Assuming that Md.Code Ann., Art. 49B, § 16(g)(2) still requires religious accommodation, does the accommodation being sought here—waiver of the grooming standard to allow the employee to wear a beard snood or net—cause undue hardship?
Following oral argument on 24 September 1992 before the three Commissioners comprising the Appeal Board, the Appeal Board adopted and incorporated by reference all of the [705]*705findings of fact made by the ALJ, but issued a memorandum and order reversing her decision.
The Appeal Board took the position that the ALJ had reached the wrong conclusions of law based upon the evidence in the record. In overturning the provisional order, the Appeal Board identified several of the findings of fact that were proposed by Domino’s and accepted by the ALJ, but upon which the ALJ did not expressly rely in her discussion or conclusions, to wit:
31. Because some customers find beard snoods to be objectionable, Domino’s will lose customers if employees are allowed to have beards and wear beard snoods. DPI would be at a competitive disadvantage ...
33. There is no inherent difference between consumers in Baltimore and consumers in the rest of the country. If beard and beard snoods on pizza employees are a problem nationally, then they are a problem in Baltimore.
34. The average Domino’s customer spends $300 a year on Domino’s pizza. The loss of one customer at every store would equal a negative $1,500,000 a year. DPI will lose $230,000,000 in sales annually if it loses 10% of its customers.
40. If one customer at each of LOOC’s 25 stores stopped buying Domino’s pizza, LOOC would lose $7500 a year in sales. A LOOC employee wearing a beard snood could mean a loss in sales between $29,000 to $53,000 a year, if 16% of the customers stopped buying pizza there.
The Appeal Board also cited Mr. Denck’s testimony that a segment of pizza purchasers found bearded employees to be objectionable, which would cause Domino’s to “lose some customers” and foster a competitive disadvantage. It noted that no evidence was introduced to challenge the validity of the surveys or to contradict their results. The absence of such opposing evidence led the Board to consider any conclusion by the ALJ as to the studies’ lack of scientific validation to be without foundation, particularly in light of the record evidence indicating the manner in which they were conducted. [706]*706Reasoning that a net or snood does not shield a beard from view, the Board further disagreed with the ALJ’s conclusion that the Beard Perceptions Study had limited value because it failed to address customer attitudes toward employees with beard restraints. The Appeal Board therefore concluded that any record evidence of lost business attributable to bearded employees would have the same weight regardless of whether a restraining device were worn.
Finally, the Appeal Board disagreed with the ALJ respecting the Beard Snood Perception Study. Although the ALJ had decided that, based on that study, one could not conclude that there was any appreciable difference in attitudes among customers served by a clean shaven employee and those wearing a beard snood, the Board cited Mr. Denck’s testimony that, at the 99% level of confidence, there was a significantly different purchase intent shown between those two groups. It furthermore found no evidence in the record to support any showing that a customer would react differently to an employee wearing a net as opposed to a snood. For all of these reasons, the Appeal Board concluded that appellees had met their burden of proving by a preponderance of the evidence that they could not accommodate Kohli’s religious practice without undue hardship on the conduct of their business.
Circuit Court Proceeding
In his appeal before the circuit court, Kohli questioned, as he does before us, whether the Appeal Board erroneously disregarded the standard of review to which it was bound pursuant to the Commission’s rules of procedure,6 as well as [707]*707whether the decision of the Appeal Board was supported by competent, material, and substantial evidence, in light of the entire record as submitted. In affirming the Appeal Board, the court noted that the Commission was governed by the same standards in reviewing an ALJ’s determination as was the court itself sitting in review of the Commission’s decision. Accordingly, the court considered ostensibly whether there was substantial evidence to support the ALJ’s conclusions.
The circuit court initially concluded that the Board had improperly substituted its judgment for that of the ALJ with respect to the Beard Perception Study. According to the court, the record did not support the Board’s conclusion that any loss of business stemming from a bearded employee would occur regardless of whether a beard restraint were worn. On the other hand, the circuit court agreed with the Appeal Board’s rejection of the ALJ’s conclusions concerning the Beard Snood Perception Study. The court acknowledged Mr. Denck’s inability to differentiate statistically between the twenty-five people in Baltimore who were shown a photograph of a clean shaven man and the twenty-five people who evaluated pictures of a bearded man wearing a snood. Nonetheless, the lack of evidence to demonstrate an inherent difference between consumers in Baltimore and consumers nationwide, coupled with Denck’s testimony that such a statistical difference did exist in the attitudes of customers in different regions of the country, led the court to find the ALJ’s decision in that regard to be erroneous. Similarly, the court cited a lack of record evidence to support the conclusion that a beard net is [708]*708more attractive than a snood, and therefore concluded that the ALJ’s finding in this regard was speculative.
The circuit court also decided that the Appeal Board’s decision was supported by competent, material, and substantial record evidence. In view of the court’s determination that the Appeal Board should not have substituted its judgment for that of the ALJ with respect to the Beard Perception Study, it resolved this question by determining that the Snood Study alone was sufficient to satisfy appellees’ burden. To reach its conclusion that LOOC and DPI had made the requisite showing of undue hardship, the court pointed to the results of the sampling of customers interviewed in Baltimore as part of the Snood Study. Based upon the evidence that one Baltimore customer in twenty-five indicated that he or she would not purchase pizza from such a bearded employee with a snood, the study could be interpreted to support the offered proposition that 4% of Domino’s Maryland customers would be lost if its franchisees permitted the accommodation requested by appellant. This consideration, coupled with the record evidence that: 1) the average Domino’s customer spends three hundred dollars per year at Domino’s outlets nationwide; and 2) approximately 66,666 Maryland customers are served annually, enabled the court to conclude that appellees had shown that they would likely suffer at least a de minimus loss.
STANDARD OF REVIEW
Judicial review of the decision of an administrative agency is narrow. United Parcel Service, Inc. v. People’s Counsel, 336 Md. 569, 576, 650 A.2d 226 (1994). The criteria of the standard of review that applies to the actions of covered State agencies (of which the Commission is one), whether the reviewing court be a circuit court or an appellate court, Fort Washington Care Center v. Department of Health and Mental Hygiene, 80 Md.App. 205, 213, 560 A.2d 613 (1989), is set forth in Md.State Gov’t Code Ann. § 10-222(h) (Supp.1994):
Decision—In a proceeding under this section, the court may:
[709]*709(1) remand the case for further proceedings;
(2) affirm the final decision; or
(3) reverse or modify the decision if any substantial right of the petitioner may have been prejudiced because a finding, conclusion, or decision:
(i) is unconstitutional;
(ii) exceeds the statutory authority or jurisdiction of the final decision maker;
(iii) results from an unlawful procedure;
(iv) is unsupported by competent, material, and substantial evidence in light of the entire record as submitted; or
(v) is arbitrary or capricious.
Ordinarily, the task of a reviewing court is to determine whether there is substantial evidence in the record as a whole to support the final decision of the agency. Human Relations Comm’n v. Baltimore, 86 Md.App. 167, 172-73, 586 A.2d 37, cert. denied, 323 Md. 309, 593 A.2d 668 (1991). “Substantial evidence” to uphold the factual findings of an administrative agency has been traditionally defined as “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion.” Anderson v. Department of Public Safety, 330 Md. 187, 213, 623 A.2d 198 (1993) (quoting Bulluck v. Pelham Wood Apartments, 283 Md. 505, 512, 390 A.2d 1119 (1978). Correspondingly, with regard to factual determinations, the scope of judicial review is limited to whether a reasoning mind reasonably could have reached the factual conclusion the agency reached. Id. In applying the substantial evidence test, the reviewing body does not substitute its judgment for the expertise of the agency. Maryland State Police v. Lindsey, 318 Md. 325, 333, 568 A.2d 29 (1990). Rather, the test is a deferential one, requiring restrained judicial judgment so as not to interfere with the factual conclusions of the agency that are adequately supported by the record. State Election Bd. v. Billhimer, 314 Md. 46, 58, 548 A.2d 819 (1988); cert. denied, 490 U.S. 1007, 109 S.Ct. 1644, 104 L.Ed.2d 159 (1989).
[710]*710Although the factual determinations of an administrative agency are entitled to deferential review, a different standard applies to its purely legal interpretations. Strother v. Board of Educ., 96 Md.App. 99, 108, 623 A.2d 717 (1993). To guide the reviewing court in applying the appropriate standard of review to the various conclusions of the agency, our cases have articulated a threefold analysis:
1. First, the reviewing court must determine whether the agency recognized and applied the correct principles of law governing the case. The reviewing court is not constrained to affirm the agency where its order “is premised solely upon an erroneous conclusion of law.”
2. Once it is determined that the agency did not err in its determination or interpretation of the applicable law, the reviewing court next examines the agency’s factual findings to determine if they are supported by substantial evidence, i.e., by such relevant evidence as a reasonable mind might accept as adequate to support a conclusion. At this juncture, ... “[i]t is the agency’s province to resolve conflicting evidence, and, where inconsistent inferences can be drawn from the same evidence, it is for the agency to draw the inference.”
3. Finally the reviewing Court must examine how the agency applied the law to the facts. This of course is a judgmental process involving a mixed question of law and fact, and great deference must be accorded to the agency. The test of appellate review of this function is “whether, ... a reasoning mind could reasonably have reached the conclusion reached by the [agency], consistent with a proper application of the [controlling legal principles].”
Strother, supra, 96 Md.App. at 109-10, 623 A.2d 717 (quoting Comptroller of the Treasury v. World Book Childcraft Int’l, Inc., 67 Md.App. 424, 508 A.2d 148, cert. denied, 307 Md. 260, 513 A.2d 314 (1986); see also Maryland Aviation Admin. v. Newsome, 99 Md.App. 269, 275-76, 637 A.2d 469 (1994), rev’d on other grounds, 337 Md. 163, 652 A.2d 116 (1995)).
[711]*711Thus, where the issue before the appellate court is one of law, the court’s review is expansive. Gray v. Anne Arundel County, 73 Md.App. 301, 307-09, 533 A.2d 1325 (1987). That is, where the agency’s decision is predicated solely upon an error of law, no deference is appropriate and the reviewing court may substitute its judgment for that of the administrative agency. Washington Nat’l Arena v. Comptroller of Treasury, 308 Md. 370, 378-79, 519 A.2d 1277 (1987). Where, however, the issue involves review of the agency’s factual determinations or a mixed question of law and fact, the more restricted standard of review applies—determining whether there is substantial evidence in the record to support the factual findings by ascertaining whether, applying the appropriate legal principles, a reasoning mind could have reached the conclusion reached by the agency. In other words, “[t]he order of an administrative agency ... must be upheld on review if it is not premised upon an error of law and if the agency’s conclusions on questions of fact or on mixed questions of law and fact are supported by substantial evidence presented to it.” Enviro-Gro v. Bockelmann, 88 Md.App. 323, 329, 594 A.2d 1190 (citing Harford County v. McDonough, 74 Md.App. 119, 122, 536 A.2d 724 (1988)), cert. denied, 325 Md. 94, 599 A.2d 447 (1991).
RULES OF PROCEDURE OF THE COMMISSION ON HUMAN RELATIONS
The threshold question in this appeal concerns not what standard of review the courts apply, but, rather, what is the standard of review that the final agency decision maker in this case must apply in reviewing the intermediate administrative decision of the ALJ. To answer this question, we must examine the rules of procedure of the Human Relations Commission.
In Maryland, administrative agencies are authorized to fashion their own rules by virtue of express or implied legislative delegations. See Department of Natural Resources v. Linchester Sand & Gravel Corp., 274 Md. 211, 218, 334 A.2d 514 (1975) (tracing history of right of legislature to delegate [712]*712powers to administrative agencies). The authority of the Human Relations Commission to promulgate rules and regulations flows from those sources. See Md.Ann.Code of 1957, Art. 49B, § 3; Md.Code Ann., State Gov’t § 10-206(b) (Supp.1994); see also Weathersby v. Kentucky Fried Chicken Nat’l Mgt. Co., 86 Md.App. 533, 587 A.2d 569, rev’d on other grounds, 326 Md. 663, 607 A.2d 8 (1992). Under Maryland’s current system of State administrative procedures, set forth in The Administrative Procedure Act (“the APA”), Md.Code Ann., State Gov’t § 10-101 et seq. (1993 Repl.Vol. & Supp.1994), the head of a covered agency has the option, under § 10-205 of the APA, of either allowing the agency itself to conduct the hearing in a contested case, or delegating such authority to the Office of Administrative Hearings (“the OAH”), which designates an administrative law judge to perform that function. In the event that an agency elects to have the OAH play a role in the hearing process, administrative law judges are generally employed not to render a final decision as a result of the hearing, but rather to develop a record and to make a recommendation to the agency head, which may either be adopted, modified, or rejected at the agency’s discretion. See Honorable John W. Hardwicke, The Central Hearing Agency: Theory and Implementation in Maryland, 14 J.Nat’l Ass’n Admin.L.J. 5, 13 (Spring 1994). The APA, however, does permit an agency to delegate to the OAH, if it so chooses, the authority to issue (1) proposed or final findings of fact; (2) proposed or final conclusions of law; (3) proposed or final findings of fact and conclusions of law; (4) proposed or final orders or orders under Article 49B of the Code; or (5) the final administrative decision of an agency in a contested case. Md.Code Ann., State Gov’t § 10-205(b)(l-5) (emphasis supplied). Thus, the agency is responsible for defining the scope of and weight given to the ALJ’s efforts.
Title 14 of COMAR sets forth the procedure adopted by the Commission on Human Relations in a claim alleging discriminatory employment practice under Md.Ann.Code of 1957, Art. 49B. The Commission has elected in its procedural rules to delegate to the OAH the responsibility for conducting' the [713]*713initial hearing in a contested case. Md.Regs.Code tit. 14, § 03.01.08(B)(1). In the absence of a timely appeal to an Appeal Board of the Commission, the written findings of fact, conclusions of law, and decision issued by the AL J, styled as a “provisional order,” become the final order of the Commission from which no further administrative appeal may be taken. Id., § 14.03.01.09(H)(5). The regulations provide, however, that an appeal to an Appeal Board of the Commission (which, in the event of an appeal, is responsible for issuing the final order of the agency) may be taken by filing with the ALJ a written notice of appeal within thirty days from the date the AL J’s provisional order is delivered to the party who wishes to appeal. Id., § 14.03.01.10(A)(1). As noted in footnote 6 supra, “[t]he Appeal Board may affirm, reverse, or modify the administrative law judge’s decision in accordance -with the standards as set forth in State Government Article, § 10-222(h), Annotated Code of Maryland.” Id., § 14.03.01.10(F)(1). Thus, although § 10-222(h) codifies the standards by which a court is bound in reviewing the final administrative decision in a contested case, the Commission has elected to bind its Appeal Board, in cases involving discriminatory employment practices, to those same standards in reviewing the ALJ’s decision in the instant case.
ANALYSIS
The parties do not dispute that appellees were permitted to deny Mr. Kohli employment on the basis of his religion pursuant to Article 49B, § 16B if accommodating his religious practice would result in undue hardship in the conduct of Domino’s business. In addition, appellees concede' that Kohli has met his initial burden of establishing a prima facie case of religious discrimination—that he 1) was a member of a protected group; 2) applied for the position in question; 3) was qualified for the position; 4) was rejected for the position in favor of someone not a member of the protected group under circumstances giving rise to an inference of unlawful discrimination. See, e.g., Alvarado v. Board of Trustees, 928 F.2d 118 (4th Cir.1991); Wright v. Nat’l Archives & Records, 609 F.2d [714]*714702 (4th Cir.1979);7 see also Stanley Mazaroff, Maryland Employment Law 506 (1990). Appellees further acknowledge that once appellant advanced a prima facie case of religious discrimination, the burden shifted to them to demonstrate that they could not accommodate Mr. Kohli’s religious practice without undue hardship. See, e.g., Maryland Comm’n on Human Relations v. Mayor of Baltimore, 86 Md.App. 167, 586 A.2d 37, cert. denied, 323 Md. 309, 593 A.2d 668 (1991) (discussing burdens of proof in the context of a discrimination claim involving a handicapped individual). Although the parties agree on the applicable law, they part company over the scope of the Appeal Board’s authority to reverse the ALJ’s provisional order, in which the ALJ found that LOOC and DPI had failed to meet their burden of proving undue hardship.
Pursuant to the Commission’s Rules of Procedure in CO-MAR 14.03.01.10(F), appellant maintains that the incorporation by reference of the criteria in Md.Code Ann., State Gov’t § 10-222(h), binds the Appeal Board to the judicial standard of review. Thus, according to appellant, the Appeal Board must exercise the same disciplined restraint as a reviewing court in analyzing the ALJ’s provisional order. In support of his position that the Commission improperly disregarded its own standard of review, appellant initially points out that the Appeal Board made no mention of the applicable standard of review within the text of its decision and order. Furthermore, appellant highlights several instances in the Commission’s final decision to illustrate that the Appeal Board [715]*715rejected the ALJ’s decision because it, substituting its judgment, disagreed with certain of her conclusions, rather than on the basis that no reasoning mind reasonably could have reached the factual conclusions she reached. In holding that the ALJ erroneously reached certain conclusions regarding the consumer surveys offered by DPI, the Appeal Board explained itself as follows:
First, the Board disagrees that the Beard Perceptions Study ‘has limited value because it focused only on the condition of not being shaven ... (and) did not address customer attitudes towards employees who wore beard restraints.’ The employee wearing a net or snood is clearly bearded to the observer as the restraint does not cover the beard from view; therefore, any loss of business shown by the study to result from an employee not being clean-shaven would apply regardless of whether a net or snood is being worn.
Second, the Board disagrees with the Administrative Law Judge’s conclusion that the Beard Perceptions Study ‘has limited value because of the lack of scientific validation.’ The Commission and Kohli have not offered any evidence to challenge the validity or reliability of the Study ... There is substantial evidence in the record showing the manner in which the Study was conducted, and the testimony of Denck, an expert in market research, supports the validity of the results.
The Appeal Board also disagrees with the Administrative Law Judge’s conclusions with regard to the Beard Snood Perception Study____ The ALJ ... reach[ed] the wrong conclusion regarding the result of the Study. The Fact that the Study surveyed consumer attitudes about bearded employees wearing snoods rather than testing for responses to bearded employees wearing nets does not serve to discredit the findings of the survey. The two types of coverings are substantially similar, and there is no evidence in the record showing that a consumer’s reaction to a bearded employee wearing a snood would be any different from the response to the same employee wearing a net.
[716]*716In appellant’s view, the above rationale demonstrates that the Appeal Board improperly failed to subject the ALJ’s conclusions to the substantial evidence test, and chose instead to substitute its judgment for hers where no pure question of law was involved.
Conversely, Domino’s insists that it was within the province of the Appeal Board to overturn the ALJ’s decision and to reach a contrary result. Citing Anderson v. Department of Public Safety, supra, and this Court’s recent decision in Department of Health and Mental Hygiene v. Shrieves, 100 Md.App. 283, 641 A.2d 899 (1994), appellees initially maintain that it is the final order of the agency that is entitled to the deferential standard of review, and, pursuant to the regulations of the Commission, the Appeal Board retains the authority to issue the final agency decision. In addition, they point to the Board’s holding that the ALJ “reached the wrong conclusion of law” on the record. By characterizing appellant’s challenge to Domino’s capacity to accommodate him without undue hardship as purely a question of law, appellees argue that the Appeal Board was permitted to apply the accepted findings of fact, or any of them, to the requisite legal standard and to overturn the ALJ’s order, without regard to the restrictions imposed by the substantial evidence test.
Both of appellees’ contentions are off the mark. First, appellees incorrectly portray the ALJ’s determination that Domino’s had not met its burden of demonstrating undue hardship as a question solely of law. As we discussed above, a question that is decisively one of law is not entitled to deference by the reviewing body. Washington Nat’l Arena, supra, 308 Md. at 378-79, 519 A.2d 1277. Therefore, if the Appeal Board had substituted a correct statement of the law for an erroneous one relied upon by the ALJ, and a different conclusion were obtained as a result, such action by the Board would have been proper under its rules. In the case sub judice, however, it is apparent that the Appeal Board took issue only with the ALJ’s application of the law to the facts or the inferences to be drawn therefrom. The Appeal Board affirmed and adopted each of the ALJ’s 103 findings of fact and [717]*717her rulings on rejected findings submitted by the parties, except as otherwise indicated by the Board’s decision and order. We have examined carefully the Board’s decision and order, and especially the part that follows the Board’s suggestion that it may have made additional or supplemental findings of fact that the ALJ had rejected in the submissions of either DPI or LOOC or that the record may have supported but which the ALJ simply had not included in her Provisional Order. Our review leads us to the conclusion that the Board made no additional or supplemental findings of fact than were expressed in the ALJ’s Provisional Order. That is to say, the Board did not adopt any of the proposed findings of fact offered by LOOC that were rejected by the ALJ (the ALJ accepted all of DPI’s proposed factual findings). Of the proposed findings tendered by LOOC that were rejected by the ALJ, none were material to the surveys and what was to be made of the survey results. Moreover, the Board did not delve into the record to render additional factual findings that had not been considered by the ALJ in formulating her Provisional Order. Despite appellees’ attempt to construe the ALJ’s decision as one premised on an incorrect conclusion of law, nowhere in its decision did the Board declare that it was reversing the ALJ’s decision because she had improperly stated the law governing a religious discrimination employment matter. The determination of whether Domino’s would suffer undue hardship by accommodating appellant is in all respects a mixed question of law and fact, to which the more restricted standard of review would apply, if that is what the agency rule requires.
With regard to Domino’s other argument, that the Commission’s final order is entitled to deference, it is true that Anderson and Shrieves stand for the proposition that the substantial evidence test normally applies to the final decision of an administrative agency. Ordinarily, only if the final order is not based upon substantial evidence can it be reversed by a court. Shrieves, supra, 100 Md.App. at 297, 641 A.2d 899. Throughout the Maryland appellate decisions, however, this principle of administrative law has generally been applied only [718]*718to circumstances where the agency has, within its regulatory scheme, reserved the capacity to substitute its judgment for that of the entity that has conducted the evidentiary hearing and rendered a recommended decision in a contested case. In determining the proper standard of review to apply in each case, however, of paramount importance is a consideration of the procedural rules that the agency itself has promulgated. Under the “Aceardi doctrine,” an agency of the government must ordinarily and scrupulously observe the rules, regulations, and procedures that it has established. United States ex rel Accardi v. Shaughnessy, 347 U.S. 260, 268, 74 S.Ct. 499, 503, 98 L.Ed. 681 (1954); Board of School Comm’rs v. James, 96 Md.App. 401, 421, 625 A.2d 361, cert. denied, 332 Md. 381 and 382, 631 A.2d 451 and 452 (1993). When the agency fails to follow its own rules, its action is invalid and courts will strike it down. James, 96 Md.App. at 421, 625 A.2d 361. The purpose of the doctrine is “to prevent the arbitrariness which is inherently characteristic of an agency’s violation of its own procedures.” United States v. Heffner, 420 F.2d 809 (1970). Although the reasoning in Accardi applies to federal administrative agencies, this Court has embraced the doctrine on several occasions. See, e.g., Board of Educ. v. Ballard, 67 Md.App. 235, 240-41, 507 A.2d 192 (1986); Maryland Nat’l Cap. Park and Planning Comm’n v. Friendship Heights, 57 Md.App. 69, 81, 468 A.2d 1353, cert. denied, 300 Md. 89, 475 A.2d 1200 (1984); Williams v. McHugh, 51 Md.App. 570, 444 A.2d 475 (1982); Board of Educ. v. Barbano, 45 Md.App. 27, 35-36, 411 A.2d 124 (1980); Hopkins v. Maryland Inmate Grievance Comm’n, 40 Md.App. 329, 335, 391 A.2d 1213 (1978).
Pursuant to Article 49B Md.Code Ann., State Gov’t § 10-206(b), the Human Relations Commission has developed its own rules of procedure to govern all proceedings before the agency. Under the plain language of those regulations as set forth in subsection 14.03.01.10(F)(1) of COMAR, the Commission has chosen to limit the Appeal Board’s authority to reverse the ALJ’s decision principally to those instances in which the ALJ’s conclusions are unsupported by competent, [719]*719material, and substantial evidence within the record submitted, or is otherwise arbitrary or capricious.8 Although the Commission was in no way obligated to impose upon itself such a rigorous standard of review, having done so it cannot [720]*720proceed without regard to them. Having discerned that the scope of review of the Appeals Board is restricted by COMAR § 14.03.01.10(F)(1), we now examine the decisions below to determine whether the Appeal Board erred in reversing the ALJ for the reasons it gave in its Decision and Order.
In reviewing the decision of the circuit court in an appeal from a decision of an administrative agency, the role of the Court of Special Appeals, as we noted earlier, “is essentially to repeat the task of the circuit court; that is, to be certain the circuit court did not err in its review.” Mortimer v. Howard Research, 83 Md.App. 432, 442, 575 A.2d 750, cert. denied, 321 Md. 164, 582 A.2d 499 (1990). Prior to its analysis of the Appeal Board’s reversal of the ALJ’s findings, the circuit court recognized that the “Board is governed by the same standards in reviewing an ALJ’s determinations as [it] is in reviewing the Board’s decision.” As it examined the record that was before the ALJ, however, it is apparent that the court, in part, did not give proper regard to the scope of the Appeal Board’s review.
With respect to the Beard Perception Study, the circuit court noted that the Appeal Board, by concluding that any loss of business resulting from a bearded employee would occur regardless of whether a net or snood is worn, merely disagreed with the ALJ. The ALJ had rejected the study because of its failure to address customer attitudes toward employees wearing beard restraints. Recognizing that the Appeal Board had disregarded the standard of review, the court ultimately declared to be improper the Board’s substitution of its assessment of the record evidence in place of the ALJ’s. In contrast, the court’s evaluation of the Beard Snood Study led it to conclude that the Appeal Board was correct in reversing the ALJ insofar as that survey was concerned. The court found the ALJ’s assertions, that the study neither portrayed a statistical difference in the attitudes of customers in the Baltimore area nor took into account the effect of a beard net as opposed to a snood, to be erroneous and not supported by substantial evidence.
[721]*721Despite its conclusion that the Appeal Board’s action concerning the Beard Perception Study was inappropriate, the circuit court affirmed the Board’s decision. The court should have recognized, however, that the Appeal Board reached its entire decision without regard to applying the substantial evidence test. Reference to § 10-222(h) would have necessarily required the Board to apply a different analysis to the Provisional Order of the ALJ. Because the Appeal Board may not have reached the same result had it applied the correct standard in reviewing the factual inferences drawn and conclusions reached by the ALJ from the studies presented by DPI, and the cross-examination to which they were subjected through the person of Mr. Denck, the court’s order of affirmance was erroneous.
It is evident that the Appeal Board, in reversing the ALJ, did not consider whether a reasoning mind could have reasonably reached the conclusions that she did. This notion is confirmed not only by want of a reference by the Board to the standards of § 10-222(h), but also by the fact that it based its decision upon a series of holdings that merely concluded that the “Administrative Law Judge reached the wrong conclusion of law from the facts in the record.” The Board never determined that the ALJ could not have adequately supported her conclusions based upon the evidence presented. Under these circumstances, where legal error is present in both the circuit court’s and the Appeal Board’s decisions, the appropriate remedy is to remand the case to the agency for a determination of whether the ALJ’s decision is supported by substantial evidence. This result is based upon our policy of according respect to the agency’s findings. See, e.g., Anderson, supra, 330 Md. at 217, 623 A.2d 198. But cf. Shrieves, supra, 100 Md.App. at 303-04, 641 A.2d 899 (proceeding to review the administrative decision directly upon finding error by lower court).
We perceive that the situation presented by the case sub judice may be an anomaly among State administrative agencies. We are unaware of any other State agency that has [722]*722opted to impose upon itself the standards of § 10-222(h) as the criteria by which it must review a decision of an ALJ.
JUDGMENT OF THE CIRCUIT COURT FOR BALTIMORE COUNTY REVERSED; CASE REMANDED TO THAT COURT WITH DIRECTION TO VACATE THE DECISION AND ORDER OF THE HUMAN RELATIONS COMMISSION AND TO REMAND THE CASE TO THE APPEAL BOARD OF THE COMMISSION FOR FURTHER PROCEEDINGS CONSISTENT WITH THIS OPINION.
COSTS TO BE PAID BY APPELLEES.