JORDAN, Circuit Judge:

Rachel Goode appeals the district court’s order affirming the denial of her application for benefits under the Social Security Act, 42 U.S.C. § 401 et seq. She argues that substantial evidence does not support the finding of the administrative law judge (ALJ) that she could perform a job that exists in significant numbers in the national economy.

Following a review of the record, and with the benefit of oral argument, we agree with Ms. Goode. The vocational expert’s testimony as to available jobs, on which the ALJ relied, was significantly and admittedly flawed. We therefore vacate the district court’s order and remand for another hearing before the ALJ.^[1]

I

Following a severe vehicle accident that prevented her from continuing to work as a nurse, Ms. Goode filed for disability benefits on April 21, 2011, alleging a disability onset date of January 30, 2011. When her claims were initially denied, Ms. Goode timely filed a request for an administrative hearing before an ALJ. Before discussing the testimony at that hearing, we briefly summarize the framework for disability determinations under the Social Security Act.

A

To make a disability determination, an ALJ must engage in a five-step sequential analysis. See 20 C.F.R. § 416.920(a)(1). A claimant bears the burden at the first four steps, but the burden shifts at step five, which asks whether the claimant can perform a job that exists in significant numbers in the national economy given her vocational profile. See Winschel v. Comm’r of Soc. Sec., 631 F.3d 1176, 1178 (11th Cir. 2011); 20 C.F.R. § 416.920(a)(4)(v).

At step five, which is at issue in this case, the Commissioner of Social Security must show “the existence of . . . jobs in the national economy which, given the claimant’s impairments, the claimant can perform.” Hale v. Bowen, 831 F.2d 1007, 1011 (11th Cir. 1987). If the Commissioner makes this showing, “the burden shifts back to the claimant to prove she is unable to perform the jobs suggested[.]” Id.

B

A vocational expert testified, via telephone, at the hearing before the ALJ. The ALJ asked the vocational expert to consider whether a person with Ms. Goode’s specific limitations could perform jobs that existed in significant numbers in the regional or national economy. The vocational expert testified that such a person could perform the unskilled job of bakery worker (Dictionary of Occupational Titles (DOT) No. 524.687-022), and said that there were 43,000 bakery worker jobs

nationally and 1,000 bakery worker jobs regionally. See Tr. 722 (“There’s over 1,000 jobs in the State of Florida. Nationally, there’s over 43,000.”). 2

Ms. Goode’s attorney questioned the vocational expert’s methodology in reaching those numbers. The vocational expert explained that he relied upon “the Occupation Employment Quarterly . . . to identify light line [INAUDIBLE] the DOT codes, and then what I use is it breaks down into jobs that are light, but only ones that are sedentary . . . and also unskilled.” Id. at 724. The vocational expert went on to describe his methodology:

It is a process of going through and identifying all of the codes, and the ones that are — the one that the baker helper goes into, and identifies the Occupational Employment Survey, the OES code, and it’s based upon employment data of 2014. And then you use that information, what we’ve got on that whole spreadsheet, and you identify . . . the ones that meet this hypothetical, in terms of the sedentary with an SVP-2, or light, in this case. And then you would identify by looking at all of the ones in that OES code, which has the DOT codes in it. I could identify those. And so then this sheet actually provides the numbers, based upon that information that I plug into it. Id. at 726–27. Finding this explanation unsatisfactory, Ms. Goode’s attorney pressed the vocational expert to explain how he arrived at his numbers. But the ALJ

interrupted the questioning of Ms. Goode’s attorney, and when the attorney explained that the vocational expert’s testimony was not clear, the ALJ said, “he explained it multiple times,” and expressed a need to move on. See id. at 727. Ms. Goode’s attorney maintained that the vocational expert’s testimony was not persuasive or credible. See id. at 727.

Later, in describing how he concluded that bakery worker would be an appropriate job for Ms. Goode, the vocational expert said the following:

[T]he statistical group, the OES code for [baker helper] . . . is 51-3099. And that is food processing workers as a group. Of that, there’s 65 different DOT codes in that group. One of them happens to be the baker helper. . . . So, for instance, I said that there’s 43,000 in this OES group. Of that, it goes down to the ones that have the baker helper has [sic] light, 2 numbers.

Id. at 730. He did not provide any further explanation.

The ALJ, considering the vocational expert to be “a specialist in the occupational field whose vocational opinions were not contradicted,” expressly adopted his testimony. See Notice of Decision at 12. The ALJ found that that there were 43,000 bakery worker jobs nationally and 1,000 regionally, concluded that Ms. Goode was not disabled because she could perform a job that existed in significant numbers in the national economy, and denied her claims. See id. at 13.

Ms. Goode timely requested review of the ALJ’s decision, which the Appeals Council denied. She then filed suit in the district court, seeking review of the

Commissioner’s final decision. The district court affirmed, and Ms. Goode appealed.

II

The question for us is whether substantial evidence supported the ALJ’s conclusion that there are a significant number of bakery worker jobs in the national economy suitable for Ms. Goode to perform. See 42 U.S.C. § 405(g). “Substantial evidence is more than a scintilla and is such relevant evidence as a reasonable person would accept as adequate to support a conclusion.” Crawford v. Comm’r of Soc. Sec., 363 F.3d 1155, 1158 (11th Cir. 2004).

“We review de novo the district court’s decision on whether substantial evidence supports the ALJ’s decision.” Wilson v. Barnhart, 284 F.3d 1219, 1221 (11th Cir. 2002). “If the [ALJ’s] decision is supported by substantial evidence[,] we must affirm.” Phillips v. Barnhart, 357 F.3d 1232, 1240 n.8 (11th Cir. 2004). “We may not decide the facts anew, reweigh the evidence, or substitute our judgment for that of the [ALJ].” Id. Nevertheless, “[r]emand for further factual development of the record before the ALJ is appropriate where the record reveals evidentiary gaps which result in unfairness or clear prejudice.” Henry v. Comm’r of Soc. Sec., 802 F.3d 1264, 1267 (11th Cir. 2015) (quotation omitted).

III

Ms. Goode challenges the ALJ’s finding at step five that there exist a significant number of bakery worker jobs in the national economy. Because the ALJ adopted the vocational expert’s testimony, and that testimony contained several significant mistakes, we agree with Ms. Goode that a remand is required.

A

Work which “exists in the national economy” means “work which exists in significant numbers either in the region where [the] individual lives or in several regions of the country.” 42 U.S.C. § 423(d)(2)(A). See also 20 C.F.R. § 416.966(a) (same). The Social Security Administration, however, “does not tally the number of job openings at a given time, but rather approximates the number of positions that exist, whether vacant or filled, and without regard to the location of the work and a claimant’s likelihood of being hired.” Chavez v. Berryhill, 895 F.3d 962, 964 (7th Cir. 2018). ALJs may therefore use vocational experts and other specialists to assist with step five determinations. See 20 C.F.R. § 416.966(e). See also Washington, 906 F.3d at 1360 (“Put simply, the critical inquiry at step five is whether jobs exist in the national economy in significant numbers that the claimant could perform in spite of his impairments, and the ALJ can consider both jobs data drawn from the DOT as well as from the testimony of the [vocational expert] in making this determination.”).

To obtain a job-numbers estimate, a vocational expert may consult a publication like the DOT. See 20 C.F.R. § 416.966(d)(1). The DOT groups jobs into “occupations” based on their similarities and assigns each occupation a code number. Unfortunately, the Department of Labor has not revised the DOT since 1991. Although the DOT is outdated—some would say obsolete—it remains one of the vocational expert’s primary tools. See Chavez, 895 F.3d at 964–65 (discussing the shortcomings of the DOT and noting an upcoming system that aims to replace it in 2020); Purdy v. Berryhill, 887 F.3d 7, 14 n.10 (1st Cir. 2018) (same).

Aside from being three decades old, the DOT presents other difficulties. DOT codes, for example, do not provide statistical information about the number of jobs available in the national economy. Instead, the vocational expert must look to other sources like the Occupational Employment Quarterly (OEQ), which is compiled by a private organization called U.S. Publishing, to find employment statistics. See Herrmann v. Colvin, 772 F.3d 1110, 1113 (7th Cir. 2014); Brault v. Soc. Sec. Adm., 683 F.3d 443, 446 (2d Cir. 2012). The OEQ database, however, does not compile data by DOT codes, but rather through the Standard Occupational Classification (SOC) system. See Brault, 683 F.3d at 446; Occupational Employment Statistics, Bureau of Labor Statistics, https://www.bls.gov/oes/ (last visited April 30, 2020).

The SOC system groups together detailed occupations with similar job duties. As a result, a single SOC group may contain multiple DOT occupations. See

Herrmann, 772 F.3d at 1113–14. In the words of the Seventh Circuit, “[t]he use of one system to supply the job titles and another [system] to provide the number of jobs creates a matching problem: a one-to-one correlation does not exist.” Chavez, 895 F.3d at 965. See also Brault, 683 F.3d at 447 n.4 (noting the “information loss” that results from the “many-to-one mapping” between the DOT and SOC codes, and explaining that any estimate of existing jobs “may deviate significantly from the actual number of existing positions”). To provide some assistance in this regard, the Bureau of Labor Statistics has published a crosswalk which provides the corresponding SOC group code for each DOT occupation. See DOT Crosswalk Search, O*NET OnLine, https://www.onetonline.org/crosswalk/DOT/ (last visited April 30, 2020).

B

The Supreme Court recently provided some guidance on assessing the testimony of vocational experts at Social Security hearings, explaining that “[t]he inquiry, as is usually true in determining the substantiality of evidence, is case-by- case.” Biestek v. Berryhill, 139 S. Ct. 1148, 1157 (2019). A case-by-case approach “takes into account all features of the vocational expert’s testimony, as well as the rest of the administrative record.” Id. (rejecting a categorical rule that would have required vocational experts to always disclose the underlying data that they used to reach their opinions). Biestek appears to be consistent with our precedent, which

teaches that we “must scrutinize the record as a whole to determine if the [ALJ’s] decision is reasonable and supported by substantial evidence.” Hale, 831 F.2d at 1010–11.

Keeping Biestek in mind, we hold that the vocational expert’s testimony here was significantly and fatally flawed for two reasons. First, the vocational expert used the wrong SOC group code to determine whether there are a significant number of jobs in the national economy that Ms. Goode could perform. Second, even if we overlooked this foundational problem, the numbers that the vocational expert cited from the wrong SOC group code substantially overstated the number of available bakery worker jobs that Ms. Goode could perform.

The vocational expert’s first mistake was incorrectly converting the bakery worker DOT code into a SOC group code. The bakery worker DOT code, which the vocational expert correctly cited as No. 524.687-022, corresponds to the SOC group for “Production Workers, All Other,” at SOC No. 51-9199. See DOT Crosswalk Search for the Occupation of “Bakery Worker,” https://www.onetonline.org/crosswalk/DOT?s=524.687-022&g=Go (last visited April 30, 2020); Br. for Appellant at Exh. B. But the vocational expert erroneously testified that the job of a bakery worker corresponded to SOC No. 51-3099 for “Food Processing Workers, All Other,” and it is from that group that the expert derived his

figure of 43,000 jobs in the national economy and 1,000 jobs in the regional economy. See Tr. 730.

The mistake, though maybe understandable given the lack of one-to-one matching between the DOT and SOC classifications, is nevertheless an error within the vocational expert’s chosen methodology, and a significant one at that. The vocational expert was wrong, and clearly so, when he testified that the job of bakery worker was within SOC No. 51-3099. His testimony was therefore insufficient for the Commissioner to sustain his burden of proof at step five. See generally Carolyn Kubitscheck & Jon Dubin, Social Security Disability: Law and Procedure in Federal Court § 3:106 (2020 ed.) (“One is hard-pressed to understand how vocational expert testimony lacking even minimal methodological support, when questioned, meets the ‘substantiality’ threshold, particularly at step five of the proceedings where the burden of demonstrating the claimant’s ability to make a work adjustment to a significant number of jobs in the economy . . . has shifted from the claimant to the agency.”).

This mistake, moreover, was not harmless. First, the vocational expert never provided the relevant job numbers in the national and regional economies from the correct SOC group corresponding to bakery worker under DOT No. 524.687-022. Second, we cannot find any such numbers in the record. Third, the Commissioner has not provided us those numbers in his brief or asked us to take judicial notice of

anything that might contain those numbers. Under the circumstances, we cannot say that substantial evidence supports the ALJ’s finding that there are a significant number of jobs in the national economy that Ms. Goode could perform. In a case like this one, a finding based on such unreliable vocational expert testimony “is equivalent to a finding that is not supported by substantial evidence and must be vacated.” Chavez, 895 F.3d at 968 (quoting Britton v. Astrue, 521 F.3d 799, 803 (7th Cir. 2008)). See also McIntyre v. Colvin, 758 F.3d 146, 152 (2d Cir. 2014) (explaining that a vocational expert may testify about available jobs that the claimant could perform based upon his “professional experience and clinical judgment” as long as the testimony is “not undermined by . . . evidence in the record”); Brault, 683 F.3d at 450 (agreeing with the Seventh Circuit that “evidence cannot be substantial if it is ‘conjured out of whole cloth’”) (citation omitted).

We reject the Commissioner’s argument that Ms. Goode is grasping at straws. At the hearing, the vocational expert based his testimony about available jobs on a specific DOT code and a mismatched SOC group code. The vocational expert’s testimony was internally inconsistent, and created a critical void in the record. It is not our job or duty to speculate on what the testimony would have been without the error. Cf. Buck v. Berryhill, 869 F.3d 1040, 1046–47, 1052 (9th Cir. 2017) (remanding to the Commissioner in part because the claimant and the vocational

expert, using the same source, came to vastly different job numbers and the ALJ did not address the discrepancy).

In closing our discussion on this first point, we do not suggest (and certainly do not hold) that vocational experts at Social Security hearings must satisfy Rule 702 and Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993). See, e.g., Brault, 683 F.3d at 448–50 (criticizing the Seventh Circuit for imposing Daubert-like requirements on vocational experts). Indeed, 42 U.S.C. § 405 provides that evidence may be received at Social Security hearings “even though inadmissible under rules of evidence applicable to court procedure.” But here the mistake—the use of the wrong SOC group—at the beginning of the analysis is so substantively fundamental and serious that it tainted the testimony that followed. The fact that vocational expert testimony is admissible does not necessarily mean that it constitutes substantial evidence. See Biestek, 139 S. Ct. at 1162 (Gorsuch, J., dissenting) (“Some courts have . . . conflated the substantial evidence standard—a substantive standard governing what’s needed to sustain a judgment as a matter of law—with procedural rules governing the admission of evidence. These courts have mistakenly suggested that, because the Federal Rules of Evidence don’t apply in Social Security proceedings, anything an expert says will suffice to meet the agency’s burden of proof.”).

The second mistake committed by the vocational expert concerns the job numbers that he testified about at the hearing. Even if we overlook the DOT/SOC matching problem described above, and use SOC No. 51-3099 as the baseline, the numbers provided by the vocational expert are overstated in a very significant way.

The vocational expert, as noted, used SOC No. 51-3099 for “Food Processing Workers, All Others.” From that starting point, the vocational expert testified that there were 43,000 bakery worker jobs nationally and 1,000 regionally. See Tr. 730. The problem is that those numbers from SOC No. 51-3099 are aggregate numbers for 65 separate DOT codes, and Ms. Goode was only capable of performing the job for only one of those DOT codes—that of bakery worker.

Simple common sense makes it is impossible to say, based on the existence of 43,000 jobs nationally and 1,000 jobs regionally for a group of 65 occupations, that all of those jobs exist for just one of those occupations. If the vocational expert were right, that would mean that there are no available jobs (nationally or regionally) for the other 64 occupations in SOC No. 51-3099. Assuming such an improbable state of affairs were statistically or realistically possible, the vocational expert did not provide any support (empirical or otherwise) to justify that quantum leap.

Our sister circuits have recognized that a vocational expert, after figuring out the total number of jobs in an SOC group, needs to take an additional step to

approximate how many of those are the specific job or jobs that the claimant could perform. See Chavez, 895 F.3d at 965–66; Purdy, 887 F.3d at 14–15; Brault, 683 F.3d at 446–47. In other words, the vocational expert “must use some method for associating SOC-based employment numbers to DOT-based job types.” Brault, 683 F.3d at 446. Here the vocational expert did not take any further steps, or use any other methodology, to try to figure out how many of the 43,000 national or 1,000 regional jobs in SOC No. 51-3099 should be allocated to the occupation of bakery worker. Not surprisingly, at oral argument the Commissioner conceded that the vocational expert mistakenly testified regarding the number of bakery worker jobs available.^[3]

As far as we can tell from the cases and the literature, there are various methods available to a vocational expert, to be used in conjunction with his knowledge and expertise. We set out two of them below so that they will be on the

table when the matter returns to the ALJ. In doing so we do not mean for our short list to be exclusive, and we express no view on the merits of any particular approach.

One method is called the equal distribution method. This method assumes that the total number of jobs that exist for a given SOC group are distributed equally among the number of DOT occupations within that SOC group. See Chavez, 895 F.3d at 966, 969–70 (explaining the method but strongly criticizing it as “illogical” because it is based on “an assumption about the relative distribution of jobs within a broader grouping that lacks any empirical footing”).

A second method is the occupational density method. It approximates job numbers using a software program known as JobBrowser Pro from SkillTRAN, which interprets available data. See Purdy, 887 F.3d at 14–15. In some cases, the use of the occupational density method will lead to “substantially different estimates” than the equal distribution method. See, e.g., Chavez, 895 F.3d at 966 (noting that the occupational density method estimated that there were 800 bench assembler positions in the national economy while the equal distribution method put the number at 108,000).

When the matter returns to the ALJ, the vocational expert will need to ensure that he or she begins by using the correct SOC group code—SOC No. 51-9199—for the position of bakery worker, DOT No. 524.687-022. Once that is done, the vocational expert should figure out the total number of jobs available in SOC No.

51-9199, and determine how many DOT codes are indeed in that SOC group. From there, the vocational expert must estimate the number of available jobs for bakery workers, and provide some explanation for how he arrived at that latter number.

IV

We recognize that the Social Security Act and its regulations “do not mandate a precise count of job numbers.” Chavez, 895 F.3d at 968 (citing 20 C.F.R. § 416.960(c)(2)). And we do not expect vocational experts to “formulate opinions with more confidence than imperfect data allows.” Id. at 970. But when an ALJ at step five relies on the testimony of a vocational expert for the number of jobs in the national and regional economies that a claimant could perform, that testimony cannot be both internally inconsistent and incomplete. Because of the errors we have identified, the vocational expert’s testimony about available jobs lacked a baseline of reliability. We reverse the district court’s order affirming the Commissioner’s decision, and remand with instructions that the matter be returned to the ALJ for further proceedings consistent with our opinion.

REVERSED AND REMANDED.