Before Higginbotham, Duncan, and Engelhardt, Circuit Judges. Stuart Kyle Duncan, Circuit Judge: The Restaurant Law Center and the Texas Restaurant Association (“Plaintiffs”) challenge a Department of Labor regulation that refines how the federal minimum wage applies to tipped employees. The district court denied Plaintiffs a preliminary injunction on the sole ground that they failed to establish irreparable harm from complying with the new rule. We disagree. Because Plaintiffs sufficiently showed irreparable harm in unrecoverable compliance costs, we reverse and remand for further proceedings. I. The federal minimum wage is currently $7.25 per hour. 29 U.S.C. §§ 206(a)(1)(C), 213(a). There is an exception for “tipped employee[s],” meaning “any employee engaged in an occupation in which he customarily and regularly receives more than $30 a month in tips.” Id. § 203(t). Tipped employees may be paid as low as $2.13 per hour, provided their tips fill out the rest of the minimum wage. Id. § 203(m)(2)(A). This is known as the “tip credit.” Over the past decades, the Department of Labor has fleshed out the contours of the tip-credit provision through regulations and other guidance.[1] In late 2021, the Department revised and added to a regulation about when an employee works in a “tipped occupation” under § 203(t). See 29 C.F.R. § 531.56(e), (f) (2021). In relevant part, the new rule permits an employer to take a tip credit, not only for an employee’s tip-producing work, but also for other work that “directly supports tip-producing work, provided that the employee does not perform that work for a substantial amount of time.” 29 C.F.R. § 531.56(f)(4). In turn, a “substantial amount of time” exists when: (i) The directly supporting work exceeds a 20 percent workweek tolerance, which is calculated by determining 20 percent of the hours in the workweek for which the employer has taken a tip credit. The employer cannot take a tip credit for any time spent on directly supporting work that exceeds the 20 percent tolerance. Time for which an employer does not take a tip credit is excluded in calculating the 20 percent tolerance; or (ii) For any continuous period of time, the directly supporting work exceeds 30 minutes. If a tipped employee performs directly supporting work for a continuous period of time that exceeds 30 minutes, the employer cannot take a tip credit for any time that exceeds 30 minutes. Time in excess of the 30 minutes, for which an employer may not take a tip credit, is excluded in calculating the 20 percent tolerance in paragraph (f)(4)(i) of this section. Ibid. The 20% rule in subpart (i) essentially codifies the “80/20 guidance” that had appeared in various Department documents over the past three and a half decades. See 86 Fed. Reg. at 60,116–17 (discussing development in Wage and Hour Division opinion letters and Field Operations Handbook of “80/20 guidance”). The “continuous 30-minute” rule in subpart (ii) is entirely new, however. See id. at 60,115–20 (providing no historical precursor on the 30-minute limitation). Finally, the new rule carries forward the Department’s longstanding “dual jobs” regulation, which recognizes that, for employees employed in both tipped and non-tipped occupations, employers may claim the tip credit only for the time those employees spend in the tipped occupation. See id. at 60,116; see also 29 C.F.R. § 531.56(e). In December 2021, Plaintiffs challenged these amendments to § 531.56 in federal court. They alleged the rule violated the Fair Labor Standards Act, the Administrative Procedure Act, and the Constitution’s separation of powers. On December 17, 2021, Plaintiffs moved for a preliminary injunction. The district court held an evidentiary hearing on February 9, 2022. On February 22, 2022, the district court denied a preliminary injunction. The court did not reach the merits of Plaintiffs’ claims. Rather, the court assumed Plaintiffs were likely to succeed on the merits but concluded they had failed to show they were irreparably harmed by the costs of complying with the new rule. Compliance costs, the court reasoned, “should have already been incurred” because the rule had been in place a month before Plaintiffs sued. The court found any “remaining” costs to be “unspecific,” “purely speculative,” and “overstate[d].” For instance, the court emphasized that the new rule “does not require the level of detailed monitoring of which Plaintiffs warn,” and that it is “similar[]” to the longstanding 80/20 Rule. The court also criticized Plaintiffs’ witnesses for making “ only rough generalizations” about compliance costs and, in one instance, “wholly uncredible” claims about those costs. The court concluded that the “regulations may be costly, but that does not make them unlawful.” Plaintiffs timely appealed. We have jurisdiction under 28 U.S.C. § 1292(a)(1). II. “We review a preliminary injunction for abuse of discretion, reviewing findings of fact for clear error and conclusions of law de novo.” Tex. All. for Retired Ams. v. Scott, 28 F.4th 669, 671 (5th Cir. 2022) (citation omitted). To obtain the “extraordinary remedy” of a preliminary injunction, the movant must show he is likely to prevail on the merits and also “demonstrate a substantial threat of irreparable injury if the injunction is not granted; the threatened injury outweighs any harm that will result to the non-movant if the injunction is granted; and the injunction will not disserve the public interest.” Atchafalaya Basinkeeper v. U.S. Army Corps of Eng’rs, 894 F.3d 692, 696 (5th Cir. 2018) (citation omitted). III. On appeal, Plaintiffs argue the district court erred in its irreparable harm analysis. They also urge us to reach the other prongs of the preliminary injunction test, but, like the district court, which simply assumed a likelihood of success on the merits, we confine ourselves to irreparable harm. See, e.g., Stringer v. Town of Jonesboro, 986 F.3d 502, 509 (5th Cir. 2021) (“As we have repeatedly observed, we are a court of review, not first view.” (citation omitted)). Under our precedent, the nonrecoverable costs of complying with a putatively invalid regulation typically constitute irreparable harm. See Louisiana v. Biden, 55 F.4th 1017, 1034 (5th Cir. 2022) (“[C]omplying with a regulation later held invalid almost always produces the irreparable harm of nonrecoverable compliance costs[.]” (citing Texas v. EPA, 829 F.3d 405, 433 (5th Cir. 2016))).[2] To be sure, such costs must be based on more than “speculat[ion]” or “ unfounded fears.” Louisiana v. Biden, 55 F.4th at 1034 (quoting Holland Am. Ins. Co. v. Succession of Roy, 777 F.2d 992, 997 (5th Cir. 1985)). In determining whether costs are irreparable, the key inquiry is “not so much the magnitude but the irreparability.” Texas v. EPA, 829 F.3d at 433–34 (quoting Enter. Int’l Inc. v. Corporacion Estatal Petrolera Ecuatoriana, 762 F.2d 464, 472 (5th Cir. 1985)). Even purely economic costs may count as irreparable harm “where they cannot be recovered ‘in the ordinary course of litigation.’” Id. at 434 & n.41 (quoting Wis. Gas Co. v. FERC, 758 F.2d 669, 674 (D.C. Cir. 1985)); see also, e.g., In re NTE Conn., LLC, 26 F.4th 980, 990– 91 (D.C. Cir. 2022) (“We have recognized that financial injury can be irreparable where no adequate compensatory or other corrective relief will be available at a later date, in the ordinary course of litigation.” (cleaned up)). Relying on these principles, Plaintiffs point out that the Department concedes that some businesses will incur ongoing costs to comply with the rule. That is correct. In the rule, the Department explains that “some employers may incur ongoing management costs . . . to ensure that tipped employees are not spending more than 20 percent of their time on directly supporting work per workweek, or more than 30 minutes continuously performing such duties.” 86 Fed. Reg. at 60,142.[3] Additionally, Plaintiffs argue they produced uncontested evidence that their member businesses (all of whom want to continue claiming the tip credit) project precisely those kinds of ongoing management costs—costs, moreover, that will far exceed the Department’s rosy estimate of “10 minutes per week.” Further, Plaintiffs also introduced evidence that their members would have to institute costly measures to track employee time to comply with the rule and to defend against investigations or lawsuits. Curiously, the district court did not acknowledge the Department’s concession that some businesses will incur ongoing costs to ensure they can continue to claim a tip credit. See 86 Fed. Reg. at 60,142 (“The Department . . . believes that some employers may incur ongoing management costs [to ensure compliance with the 20-percent and 30-minute rules].”). Nor did the court cite our precedent teaching that nonrecoverable compliance costs are usually irreparable harm. See, e.g., Texas v. EPA, 829 F.3d at 433. Those omissions are striking, given that Plaintiffs assert that their members will incur exactly the kinds of continuing compliance costs predicted by the Department itself. And, of course, no one claims that those costs could be recovered if the rule were held invalid. See Wages & White Invs., 16 F.4th at 1142 (observing agency did not claim recoverable compliance costs “probably because federal agencies generally enjoy sovereign immunity for any monetary damages”). The district court’s order mentions none of this, despite the fact that Plaintiffs argued these points in support of a preliminary injunction. Instead, the district court emphasized the weakness of Plaintiffs’ evidence, as the Department now does on appeal. For instance, the court found Plaintiffs’ claimed ongoing costs “to be overstate[d]” because the rule does not require “ the level of detailed monitoring of which Plaintiffs warn.” Similarly, the Department claims that the “rule expressly states that it has ‘no recordkeeping requirement.’” DOL Br. at 20 (citing 86 Fed. Reg. at 60,140). Both points are meritless. To claim the tip credit, employers must “ensure that tipped employees are not spending more than 20 percent of their time on directly supporting work, or more than 30 minutes continuously performing such duties.” 86 Fed. Reg. at 60,142; see 29 C.F.R. § 531.56(f)(4)(i), (ii). We cannot fathom how an employer could honor these specific constraints without recording employee time. What if an employer is investigated by the Department or sued by an employee for wrongly claiming the tip credit? Without time records, how could an employer defend itself? See Rafferty v. Denny’s Inc., 13 F.4th 1166, 1190–91 (11th Cir. 2021) (emphasizing the employer’s duty to create time records where a plaintiff claimed to perform non-tipped duties for more than 20 percent of her time). The rule itself confirms all this. Contrary to the Department’s claim that the rule has “no recordkeeping requirement,” DOL Br. at 20, here is what the rule actually says: “[T]he Department did not propose new records requirements and the final rule does not contain a revision to current recordkeeping requirements nor does it enact new recordkeeping requirements.” 86 Fed. Reg. at 60,140 (emphases added). Translated into English: the rule still has a recordkeeping requirement, and now it includes the new 30-minute limitation. In the same vein, the Department also claims that “employers need not engage in ‘minute to minute’ tracking of an employee’s time in order to ensure that they qualify for the tip credit.” DOL Br. at 20. No evidence is given for this assertion. The Department merely cites a sentence from the rule that baldly states, “the minute-to-minute tracking discussed by commenters is not required by the rule.” See 86 Fed. Reg. at 60,154. No explanation is given (nor can we imagine one) why an employer would not have to track employee minutes to comply with a rule premised on the exact number of consecutive minutes an employee works. The Department also assures us that a “30-minute uninterrupted block of time . . . can be readily distinguished from the work that surrounds it.” DOL Br. at 21 (citing 86 Fed. Reg. at 60,137). Maybe so, but that does not remove an employer’s need to account for blocks of employee time, especially if an employer is accused of violating the rule. Next, the district court doubted that compliance would be “unworkably burdensome” given the rule’s “similarity to the 80/20 guidance, which has governed the industry for decades.” The Department echoes this argument. Both miss a key point, however. Even assuming the new rule’s 20 percent threshold is exactly like the previous 80/20 rule, the 30-minute limit is, as everyone agrees, brand-new.[4] And it is an independent constraint on an employer’s taking the tip credit. Quite apart from the 20-percent rule (which concerns the workweek as a whole), an employer cannot take the credit for discrete periods where directly supporting work “exceeds 30 minutes.” 29 C.F.R. § 531.56(f)(4)(ii). Moreover, the 30-minute limit affects the 20-percent standard: time beyond 30 continuous minutes is excluded from the 20-percent workweek calculation. Ibid. In other words, the 30-minute limitation is a new constraint on the tip credit that both requires distinct recordkeeping and affects the existing 20-percent standard. Neither the district court nor the Department explains why this new requirement would not impose new costs. To the contrary, the rule itself confirms that employers who want to continue claiming the tip credit—like Plaintiffs’ members—will “incur ongoing management costs” to ensure employees do not spend more than 30 minutes continuously performing directly supporting work. 86 Fed. Reg. at 60,142. Finally, the district court faulted Plaintiffs for providing “only rough generalizations” about their members’ ongoing compliance costs and failing to “provide concrete evidence, or even rough estimates of the costs themselves.” The Department presses these points on appeal. We are again unpersuaded. For example, Plaintiffs’ witnesses offered specific estimates of the additional time that managers would incur to comply with the rule: “at least 8 hours a week,” said one, “at least 10 hours,” said another—all far exceeding the Department’s own 10-minute estimate. Plaintiffs also noted the need to “hire additional managers to perform ongoing monitoring of tasks, audits, and correct back pay when servers, bartenders, and bussers do not clock in and out correctly.” Further contrary to the district court’s view, this evidence is not “[s]peculative.” See Holland Am. Ins. Co., 777 F.2d at 997. The Department itself estimated that businesses would incur $177 million each year in compliance costs mirroring the ones Plaintiffs’ members claim. See 86 Fed. Reg. at 60,142–43. And even this figure was based on the Department’s “estimate” that businesses would only spend an average of “10 minutes per week on management costs.” Id. at 60,142 But the Department does not explain how it arrived at this “estimate,” and in any event the Department also believes the new rule requires no recordkeeping— an assumption we have already rejected. Nor does it matter that Plaintiffs did not convert each allegation of harm into a specific dollar amount. Our precedent requires only that alleged compliance costs must be “more than de minimis.” Louisiana v. Biden, 55 F.4th at 1035 (quoting Enter. Int‘l, 762 F.2d at 472). Stringently insisting on a precise dollar figure reflects an exactitude our law does not require. Under the proper inquiry for irreparable harm, Plaintiffs have provided sufficient evidence for a finding in their favor.[5] * * * Because the district court abused its discretion in finding no evidence of irreparable harm, we REVERSE the order denying a preliminary injunction and REMAND for further proceedings consistent with this opinion. We are confident that the district court will proceed expeditiously to consider the remaining prongs of the preliminary injunction analysis. Patrick E. Higginbotham, Circuit Judge, dissenting: With respect to my able brethren, I must dissent. I. The recitation of the facts is accurate, but the majority falls short in accounting for the demanding standard of review of a denial of a preliminary injunction, an “extraordinary remedy.”[6] Plaintiffs must establish, inter alia, that they are “likely to suffer irreparable harm”[7] that is “more than ‘speculative;’ ‘there must be more than an unfounded fear on the part of the applicant.’”[8] We review a “grant or denial of a preliminary injunction for abuse of discretion, with any underlying legal determinations reviewed de novo and factual findings for clear error,”[9] “giving ‘due regard to the trial court’s opportunity to judge the witnesses’ credibility.”[10] “Clear error review follows from a candid appraisal of the comparative advantages of trial courts and appellate courts.”[11] “In ‘applying this standard to the findings of a district court sitting without a jury, appellate courts must constantly have in mind that their function is not to decide factual issues de novo.’”[12] Thus, “[a] finding that is ‘plausible’ in light of the full record—even if another is equally or more so—must govern.”[13] It follows that “[e]ven if we disagree with the district court’s analysis in some places, ‘we may not simply . . . substitute our judgment for the trial court’s, else that court’s announced discretion would be meaningless.’”[14] In practice, then, “[o]nly under ‘extraordinary circumstances‘ will we reverse the denial of a preliminary injunction.”[15] II. I would not set aside the able United States District Judge’s assessment of the record evidence: “Plaintiffs’ arguments and evidence of irreparable harm amount only to speculative concerns, conclusory claims, and uncredible assertions.” Whether I would agree upon a de novo review aside, the record supports its finding. As the evidence shows that while the new rule may have irreparably harmed some set of restaurants within the United States or would, in the future, engender harm, Plaintiffs did not show that they themselves have suffered or would suffer harm.[16] First, it is true, as the majority writes, that “the Department concedes that some businesses will incur ongoing costs to comply with the rule.”[17] But neither the majority nor Plaintiffs speak to the specifics of that concession. The Department’s analysis, as published in the Federal Register, concedes the rule could engender three categories of costs—(1) regulatory familiarization costs, (2), adjustment costs, and (3) management costs.[18] Consider the regulatory familiarization costs. The Department only conceded that some businesses would incur a cost to familiarize themselves with the rule, requiring on average 1 hour of time to get up to speed.[19] But even that concession was qualified, concluding that the estimate “represents an average of employers who would spend less than 1 hour or no time reviewing, and others who would spend more time.”[20] Similarly, the Department estimated that adjusting the scheduling of staff would, on average, require a single hour of work,[21] and the Department then qualified this concession: “the Department believes that many employers likely would not need to make any adjustments at all, because either they do not have any tipped employees, do not take a tip credit, or the work that their tipped employees perform complies with the requirements set forth in this rule.”[22] Stepping back, because familiarization and adjustment precede rule implementation, there is good reason to believe that restaurants—including those represented by Plaintiffs and others—incurred these upfront costs prior to the district court’s evidentiary hearing. As the record evidence does not clearly and unmistakably show otherwise, one cannot say record evidence fails to lend plausible support for the district court’s finding of fact that “these costs should have already been incurred” with regard to the first two categories of costs. Implicitly making this point, Plaintiffs attack the district court’s conclusion with evidence regarding ongoing costs. Similarly, the majority focuses its attention to purported “ongoing management costs.”[23] True again, the Department concedes in the published rule that “some employers may incur ongoing management costs . . . to ensure that tipped employees are not spending more than 20 percent of their time on directly supporting work per workweek, or more than 30 minutes continuously performing such duties.”[24] Frustration with the district court’s failure to explicitly acknowledge this concession misses the mark. The question is solely if the district court’s finding of fact—that Plaintiffs failed to establish that they would be irreparably harmed in the form of ongoing management costs as a result of the rule—is plausible based on the attendant record. It is. And it is for the precise reason that the concession does not connect these costs to the specific plaintiffs in this action. For example, the district court was within its right to credit the Department’s evidence that many employers would not spend time on ensuring compliance with the rule because their businesses are already set up to comply with it,[25] but for the ones who would, compliance would likely require ten minutes per week.[26] Reframed, the concession comes clear: the Department cannot promise that no business will incur a cost, which does not reach as far as Plaintiffs wish—or the majority accepts. So, too, was the district court within its right to discredit vague and unsupported testimony by self-interested witnesses. Consider the testimony of Angelo Amador, the Executive Director of RLC. Amador did not name a single specific member restaurant that incurred additional compliance costs, presumed that restaurant owners were likely paying attorneys absent a single invoice, and thought that the time for such training “will be like 10 hours, not 10 minutes.” Further, Amador said, “I don’t see anybody taking 10 minutes in training for this regulation.” The same vagueness permeates the testimony of Emily Knight, the President and CEO of the Texas Restaurant Association. Knight attested to the fact that the rule would cause a “massive financial hit,” but the greatest specificity she could give was that monitoring costs would “be in the thousands of dollars” without any evidence to substantiate that number. And yet again, the declaration of one restaurateur similarly averred that bringing five restaurants into compliance with the rule would cost one million dollars per year without any documentation to support this assertion. As the majority writes—albeit only with regard to the rule—”[n]o evidence is given for th[e]s[e] assertion[s],” and “[n]o explanation is given.”[27] The attestants’ omnipotence is enviable, but “[s]peculation and suspicion are just not any evidence at all”;[28] to establish a concrete harm sufficient for injunctive relief, “[s]peculation is not enough.”[29] Indeed, the district court’s treatment of Amador’s statements is entitled to even more weight, as he made those statements at an evidentiary hearing wherein the trial court had the opportunity to assess his command of the issue firsthand. At bottom, the majority writes: “We cannot fathom how an employer could honor these specific constraints”[30] without incurring costs pursuant to the rule. Of course, and few would quarrel with the idea that no employer will come out of this rule’s implementation unscathed; rather, I stress only that the district court was entitled to insist on a far more concrete presentation of harms than these unsubstantiated observations—sound though they may be. Perhaps the harms are there, but we ought not fault a veteran District Judge for demanding more specificity and concrete evidence, or to at least be wary of its absence, particularly with such a far-reaching rule and proportionally far-reaching relief at stake. * * * * * The district court refused to issue a preliminary injunction—again, “an extraordinary remedy that may only be awarded upon a clear showing that the plaintiff is entitled to such relief”[31]—because Plaintiffs failed to make a clear showing that they were harmed. To my eyes, the majority yields to the temptation to insert its own logic to fill the void, as shown by the questions hypothetically and rhetorically posed. While they are powerful tools of the trade, the effect here is to supplant the district court’s judgment for its own,[32] reasoning that because some employers will be harmed by the Rule’s wide net, Plaintiffs via their member restaurants will inevitably be caught in the seine. “Where ‘the district court’s account of the evidence is plausible in light of the record viewed in its entirety, the court of appeals may not reverse it even though convinced that had it been sitting as the trier of fact, it would have weighed the evidence differently.’”[33] The district court’s factual findings that Plaintiffs have failed to make a clear showing that they will be harmed is plausibly supported by the record, viz. an absence of evidence connecting their restaurants to the Rule’s costs. With respect to my colleagues, I must DISSENT.[34]
