Appellant, Steven Yerks, a teacher who was discharged primarily for failing to correct alleged performance deficiencies, appeals a final administrative order in which the Broward County School Board rejected certain factual findings of the Administrative Law Judge (“ALJ”) and rejected the ALJ’s recommendation that appellant be reinstated. We reverse.
This dispute arises out of appellant’s performance as a mathematics teacher at Boyd Anderson High School during the 2013-14 school year. At all relevant times, the School Board used a performance evaluation system known as iObservation, which is based on the methods of Dr. Robert Marzano.
The iObservation form consists of 60 strategies or criteria listed in four domains—classroom strategies, planning and preparation, reflecting on teaching, and professionalism. Observers score the criteria in the domains by issuing the following “datamarks” to indicate the teacher's level of performance—innovating (4 points), applying (3 points), developing (2.5 points), beginning (2 points), and not using (1 point).
In February 2014, appellant was placed on a Performance Development Plan. There were 18 observations of appellant’s classroom performance during the school year. Most of the observations were conducted by either the principal or the assis*846tant principal. Over 100 “datamarks” concerning appellant’s teaching performance were entered into the iObservation system during the observations. Appellant’s final score was unsatisfactory.
In June 2014, the Broward County Superintendent of Schools filed an Administrative Complaint recommending that áppellant be terminated from his employment with the school district. The case proceeded to a formal administrative hearing.
At the three-day hearing, the ALJ heard testimony from 16 witnesses and admitted multiple exhibits into evidence, including the 102 datamarks recorded in the performance evaluation system. Several of the Superintendent’s witnesses testified about 'appellant’s alleged misconduct, but the ALJ found that much of this testimony was not credible.1
The principal and assistant principal testified that appellant’s classroom performance was unsatisfactory based on their observations. No other witness conducted, or was'credentialed to conduct, an observation of appellant’s teaching under the Mar-zano method.
Appellant did, however, present the testimony of a liaison between the Teachers Union and the school district, who testified ■that she had concerns about the procedural fairness of the observations of appellant’s teaching. Although she was not a credentialed Marzano observer, she received training concerning the Marzano method and testified that some of the data-marks appellant received were contradictory to that training. For example, appellant received multiple “not.using” datamarks that failed to contain any feedback on how to improve. Likewise, for one observation, there was evidence that appellant had posted a learning goal, but appellant still received a “not using” datamark, which was contradictory to the Marzano evaluation system.
Appellant, who was also trained in the Marzano method but was not a credentialed observer, testified in his own defense and criticized how the observations were scored.2
The ALJ issued a lengthy recommended order, finding that the Superintendent failed to prove any of the charges against appellant. The main focus of the recommended order was on whether the Superintendent proved the allegations regarding appellant’s failure to correct performance deficiencies. On this issue, the ALJ found that the observations conducted by the principal and assistant principal “were tainted by incompetence, carelessness, and confirmation bias.”
The ALJ discarded or revised various datamarks on the following grounds: (1) the observers misapplied criteria requirements; (2) the observers disproportionately observed certain criteria; and (3) certain datamarks were not proven by the greater weight of the evidence.
Contrary to the dissent’s suggestion that the ALJ’s decision was based on nothing more than a conclusory impression that *847the observations were tainted, the ALJ undertook a painstaking review of each observation and discussed his findings at length in a 129-page, order. The following are but a few examples of the ALJ’s criticisms of the datamarks;
• The iObservation form requires observers to check items of “Teacher Evidence” that are present during an observation of a particular criterion. However, in the datamarks issued through February 2014, the assistant principal would often check items that he believed were missing rather than items that were present, which created an unreasonable likelihood of an unreliable observation. The ALJ thus observed: “For particularly inattentive or inept observers, checking what is missing may be intuitive to support relatively low datamarks for items of Teacher Evidence that are clearly illustrative, even though the absence of an illustrative item logically does not preclude a relatively high data-mark, if other evidence of compliance is present.”
• With respect to the datamarks issued by the assistant principal for domains 2 through 4, the ALJ found that there was “no evidence of any exchange of information or examination of planning materials,” and that “[t]he failure to inform these determinations in the manner provided by the iObservation form undermines the reliability of all of the datamarks for these criteria.”
• The ALJ found that there was an inconsistency among certain datamarks entered by the assistant principal as to the same criteria over the same time span, explaining that the datamarks “collectively make no sense” and should be discarded.
• With respect to a datamark of “hot using”' that appellant received for the criterion of “promoting positive interactions. with colleagues,” the ALJ found that the comment corresponding to that datamark reflected the assistant principal’s opinion of appellant’s behavior at a recent teacher meeting and nothing more. The criterion required an analysis of the extent to which a teacher interacts with other teachers to promote learning, which was not done. Thus, the ALJ discarded the datamark to avoid the unreasonable likelihood of an unreliable evaluation.
• The ALJ noted that appellant received a “beginning” datamark for a criterion concerning “seeking mentorship” based on the assistant principal’s comment that appellant did not show appreciation for the efforts- and assistance of his academic coaches, even though gratitude was not an element of the criterion at issue.
After discarding and revising various da-tamarks, the ALJ recalculated appellant’s score and determined that his performance was in the effective range.
The ALJ determined that, .the Superintendent had “failed to prove any performance deficiency for the 2013-14 school year.” The ALJ found -that “[t]he evidence affirmatively proves that [appellant] worked faithfully and diligently for his students.” The ALJ concluded that “[r]egardless of whether the observation data were rejected in their entirety for these reasons or, as here, discarded and revised after analysis of each datamark, there is no performance deficiency to correct.” Accordingly, the ALJ recommended that the School Board. dismiss the Administrative Complaint and reinstate appellant with back pay.
In a final order, the School Board rejected'many of the ALJ’s factual findings and conclusions of law. Most important, the School Board rejected the ALJ’s Findings of Fact “to the extent that such findings interpret the evaluation rating methodology and/or modify or discard evaluation da-*848tamarks .... ” The School Board thus rejected the ALJ’s conclusions of law on the counts regarding performance deficiencies.3 The School Board determined that appellant’s employment would be terminated. This appeal from the School Board’s final order followed.
An administrative agency, such as a school board, may not reject or modify an ALJ’s findings of fact “unless the agency first determines from a review of the entire record, and states with particularity in the order, that the findings of fact were not based upon competent substantial evidence or that the proceedings on which the findings were based did not comply with essential requirements of law.” § 120.57(1)l), Fla. Stat. (2015). Stated another way, “neither an administrative agency nor a reviewing court may reject an administrative hearing officer’s findings of fact, as long as those findings are supported by competent, substantial evidence in the record.” Maynard v. Fla. Unemployment Appeals Comm’n, 609 So.2d 143, 145 (Fla. 4th DCA 1992). “Nor may the agency reject a finding that is substantially one of fact simply by treating it as a legal conclusion.” Abrams v. Seminole Cty. Sch. Bd., 73 So.3d 285, 294 (Fla. 5th DCA 2011).
If the ALJ’s factual findings “are supported by competent substantial evidence, the agency cannot reject them even to make alternate findings that are also supported by competent substantial evidence.” Resnick v. Flagler Cty. Sch. Bd., 46 So.3d 1110, 1112-13 (Fla. 5th DCA 2010). “Factual issues susceptible of ordinary methods of proof that are not infused with policy considerations are the prerogative of the hearing officer as the finder of fact.” Heifetz v. Dep’t of Bus. Regulation, Div. of Alcoholic Beverages & Tobacco, 475 So.2d 1277, 1281 (Fla. 1st DCA 1985). For example, the performance or competence of a professional is an issue susceptible to ordinary methods of proof and is thus an issue of fact to be determined by the ALJ. See Gross v. Dep’t of Health, 819 So.2d 997, 1003 (Fla. 5th DCA 2002) (the issue of whether a doctor deviatéd from the applicable standard of care is susceptible to ordinary methods of proof and is an issue of fact to be determined by the administrative law judge); Harac v. Dep’t of Prof'l Regulation, Bd. of Architecture, 484 So.2d 1333, 1337 (Fla. 3d DCA 1986) (agency could not reject hearing officer’s determination that candidate’s answer to architectural exam question demonstrated minimum competence: “Although the determination of design competency in architecture requires specialized knowledge and experience, it is not so unique so as to defy ordinary methods of proof in formal adversarial proceedings.”).
“The agency is not authorized to weigh the evidence presented, judge credibility of witnesses, or otherwise interpret the evidence to fit its desired ultimate conclusion.” Heifetz, 475 So.2d at 1281. In a fact-driven case where an employee’s conduct is at issue, “great weight is given to the findings of the administrative law judge, who has the opportunity to hear the witnesses’ testimony and evaluate their credibility.” Resnick, 46 So.3d at 1112. “An agency abuses its discretion when it improperly rejects an administrative law judge’s findings.” Id. at 1113.
If an agency rejects conclusions of law contained in a recommended order, “the agency must state with particularity its reasons for rejecting or modifying such conclusion of law or interpretation of administrative rule and must make a finding that its substituted conclusion of law or interpretation of administrative rule is as *849or more reasonable than that which was rejected or modified.” § 120.57(1)(l), Fla. Stat. (2015). An appellate court reviews an agency’s conclusions of law de novo, and “will defer to the agency’s conclusions of law unless they are clearly erroneous or contrary to law.” U.S. Blood Bank, Inc. v. Agency for Workforce Innovation, 85 So.3d 1139, 1142 (Fla. 3d DCA 2012).
Here, we conclude that the School Board exceeded its authority by rejecting the ALJ’s factual findings that were supported by competent substantial evidence.To be sure, the ALJ should not have attempted to rescore appellant’s evaluation, given that the ALJ did not actually observe appellant teaching.4 Nonetheless, competent substantial evidence supported the ALJ’s conclusion that a considerable percentage of the administrators’ evaluation datamarks regarding appellant’s teaching performance were flawed. Put simply, the ALJ’s core findings regarding the flawed nature of appellant’s performance evaluation were supported by competent substantial evidence. '
The issue of appellant’s classroom performance was not so unique as' to defy ordinary methods of proof in formal adversarial proceedings. It was within the ALJ’s purview as a fact-finder to determine as a factual matter that the observations of appellant’s performance were “tainted by incompetence, carelessness, and confirmation bias.” The ALJ made numerous valid criticisms of the datamarks, many of which could be gleaned from the evaluation forms themselves. For example, the ALJ correctly pointed out that on several occasions the assistant principal improperly filled out the observation form by checking items that he believed were missing rather than items that were present The School Board has not shown with particularity how the ALJ’s criticisms of the datamarks were wrong.
The ALJ’s lack of credentials as a Mar-zano observer is not a valid basis for rejecting the ALJ’s findings. The ALJ did not need to be a credentialed Marzano observer—nor did he need to actually observe - appellant’s teaching—to conclude from the evidence that a substantial number of the school administrators’ data-marks were based “on their misreading of the requirements of the criteria that they were assessing,” thus signifying incompetence, .carelessness, ■ and confirmation bias. Just as a fact-finder is generally free to reject expert testimony, so. too may a fact-finder reject the validity of an evaluation from a credentialed Marzano observer.5
Accordingly, even if' the ALJ should not have rescored the observations of appellant’s performance, the ALJ was permitted to reject the validity of the evaluation as a whole.6 Thus, competent substantial evi*850dence supported the ALJ’s findings that the Superintendent failed to meet his burden as to all of the counts related to appellant’s alleged performance deficiencies.
Similarly, the School Board should not have rejected the ALJ’s cónclusions of law as to the misconduct counts. Although the ALJ rejected the clear majority of the factual allegations of misconduct against appellant, the ALJ found that the record established two isolated incidents of appellant’s inappropriate treatment of his teaching coaches in the presence of students. The School Board, relying on these incidents, found that its substituted conclusion of law as to the misconduct counts was “as or more reasonable [than] that which is being rejected.’’
An agency is hot, however, permitted to “reject a finding that is substantially one of fact simply by treating it as a legal conclusion.” Abrams, 73 So.3d at 294. Here, the School Board did just that. In light of the ALJ’s undisturbed findings of fact that appellant’s behavior could not be characterized as harassment, discriminatory, intimidating, oppressive, or otherwise capable of reducing his colleagues’ ability to perform their duties, the School Board’s conclusion of law regarding the misconduct counts simply is not . “as or more reasonable” than that which was rejected or modified. In any event, appellant’s two “inappropriate” statements, as found by the ALJ, were so minor that the ALJ’s conclusion of law on the misconduct counts is more reasonable than the School Board’s.
In the final analysis, the School Board has not shown that the ALJ’s key factual findings regarding the reliability of the datamarks were unsupported by competent substantial evidence. We therefore reverse and remand with, instructions for the School Board to adopt the ALJ’s recommended order to the extent the recommended order is consistent with the above analysis, and to implement the ALJ’s recommendation that appellant be-reinstated with back pay.
Reversed, and Remanded.
Conner,' J., concurs. Forst, J,, dissents with opinion., The dissent emphasizes that there was ample evidence of appellant’s disrespectful conduct. However, neither the School Board, nor this court, is authorized to reweigh the evidence or to judge the credibility of witnesses, And, in any event, the School Board rejected the Superintendent’s exceptions to the ALJ’s findings of fact regarding the misconduct ■ counts.
. The dissent asserts that appellant "only commented upon two of the eighteen observations.” This assertion requires clarification. After an observer submits a classroom observation on the iObservation system, a teacher has access to it and may write a comment. Here, appellant wrote comments in the iOb-servation system as to only two of the 18 ^ observations. But, in his testimony, he criticized far more than two of the observations,
. The School Board also rejected the ALJ’s conclusions of law as to the misconduct allegations of Counts I and II, but this determination appears inconsistent with the School Board’s acceptance of the ALJ's findings of fact related to those counts.
. While the ALJ should not have attempted to rescore appellant’s performance, there is no indication that the ALJ considered any evidence outside the record in making his findings. We therefore reject the School Board’s attempt to liken this case to Thorn v. Florida Real Estate Commission, 146 So.2d 907 (Fla. 2d DCA 1962), which stands for the unremarkable proposition that it is improper to base a decision on facts outside the record. At most, the AU might have made some findings that were unsupported by competent substantial evidence.
.. Indeed, if an ALJ could never reject the validity of a Marzano evaluation, then what would be the point of a formal hearing? ' Judges are frequently called upon to scrutinize the reliability of expert opinions in areas requiring specialized knowledge and experience, even though judges may not be experts themselves.
. The School Board and the dissent each suggest that the ALJ harbored a "disdain” for the iObservation evaluation process. While the ALJ did suggest there were flaws in the iOb-servation system, the ALJ was primarily concerned with the reliability of the datamarks *850themselves and how the tool was used in this case to justify appellant's termination.