Appeal from the U.S. District Court, D.N.J. Magistrate Judge Cathy L. Waldor, No. 2:17-cv-07386

Before: KRAUSE, PHIPPS, and ROTH, Circuit Judges Argued Nov. 3, 2025; Decided June 11, 2026

KRAUSE, Circuit Judge . Class action counsel serve a valuable role in our legal system and deserve to be paid. But not twice. To ensure counsel are paid only reasonable attorneys’ fees, courts often calculate fees using the lodestar method—the product of class counsel’s reasonable hours on a case multiplied by their reasonable hourly rates. In certain cases, however, courts may increase the resulting lodestar using a “lodestar multiplier” to account for special circumstances. In which cases is that permissible? The Page 2Supreme Court has curtailed the use of lodestar multipliers to calculate reasonable attorneys' fees awarded under federal fee-shifting statutes, see Perdue v. Kenny A. ex rel. Winn , 559 U.S. 542 (2010), but it has not addressed the use of multipliers when fees are awarded under contractual fee-shifting provisions. That is the question presented in this case, where Appellant BMW appeals the District Court’s use of a lodestar multiplier in awarding fees pursuant to a settlement agreement. Because we conclude that the Supreme Court’s constraints on the use of lodestar multipliers in statutory fee-shifting cases also apply in contractual fee-shifting cases, we will vacate the District Court’s fee award and remand for further proceedings.

This appeal concerns the calculation of attorneys' fees following the settlement of a consumer class action. In September 2017, Plaintiffs sued BMW of North America and its German parent company in a putative class action for allegedly selling cars with defective timing chains. BMW filed a motion to dismiss that was granted in part, prompting Plaintiffs to file an amended complaint that asserted 20 federal and state causes of action on behalf of a nationwide class and twelve state-specific subclasses.¹ The parties then engaged in four months of paper discovery, totaling approximately 12,000 pages of documents, before reaching a settlement within one Page 3day of mediation. Though that settlement resolved the merits of the dispute, the parties could not agree on the amount of attorneys' fees. So they entered into another round of mediation and landed on a fee arrangement that was then incorporated into the final Settlement Agreement. Under that Agreement, Class Counsel would apply to the District Court for an award of “reasonable attorneys’ fees,” Supp. App. 104, to be paid by BMW 'separate and apart from any relief provided to the Settlement Class,” Supp. App. 130. Although the Settlement Agreement did not specify how the District Court was to calculate those reasonable fees, it did provide two guideposts for the District Court’s exercise of its discretion: Class Counsel agreed to request no more than $3.7 million, while BMW agreed not to oppose Class Counsel’s application for fees if it requested up to $1.5 million.

Not surprisingly, after the District Court preliminarily approved the class action settlement, Class Counsel applied for the full $3.7 million, which BMW opposed. The District Court then employed the lodestar method. It determined that Class Counsel had reasonably expended the 2,713 hours they claimed in litigating the case, which, multiplied by an average rate of $716 per hour, produced a baseline lodestar of $1.9 million. The District Court, however, viewed that lodestar as insufficient, so it proceeded to apply a lodestar multiplier. It considered the factors we identified in Gunter v. Ridgewood Energy Corp. as relevant in equitable common-fund cases when courts calculate fees using the percent-of-fund method, a circumstance in which 'the attorneys' fees and the clients'Page 4 award come from the same source and the fees are based on a percentage amount of the clients’ settlement award.” 223 F.3d 190, 195 n.1 (3d Cir. 2000). Those factors include the size of the fund, the skill and efficiency of the attorneys, the complexity and duration of the litigation, and the risk of nonpayment. Id.

Weighing these factors, the District Court settled on a multiplier of 1.94. Perhaps not coincidentally, that multiplier yielded a fee award of $3.7 million—which the District Court ordered be paid to Class Counsel. That order led to BMW’s first appeal and resulted in our vacatur of the fee award and remand to the District Court. See Gelis v. BMW of N. Am., LLC ( Gelis I ), 49 F.4th 371 (3d Cir. 2022).

As relevant to the issues before us today, we held in Gelis I that the record was not then sufficient to support a $3.7 million award. We agreed with BMW that because Class Counsel submitted only three, single-page summary charts using vague language to describe each biller's hours, e.g., 'discovery activities,' we could not 'discern . . . whether certain hours [were] duplicative . . . or whether the total hours billed were reasonable for the work performed.' Id. at 376-77, 380 (citation modified). BMW also argued that the District Court had erred in applying a lodestar multiplier, but as we were vacating and remanding in any event, we “decline[d] to decide whether the District Court was bound by the strictures of Perdue ” and the federal statutory fee-shifting cases. Id. at 381. We did observe, however, that 'the parties' focus on the statutes under which named plaintiffs sued [was] misplaced”Page 5 because Class Counsel's fees were awarded 'pursuant to a contract—the settlement agreement—not pursuant to a statute.” Id. (quoting In re Home Depot Inc. , 931 F.3d 1065, 1082 (11th Cir. 2019)). We also noted, for guidance on remand, that the District Court's justification for using a multiplier lacked the detail necessary “to give us a sufficient basis to review [the] fee enhancement.' Id. (citation modified).

On remand, Class Counsel supplemented the record with detailed billing statements and again requested $3.7 million in fees, this time claiming 2,877 hours (including an additional 164 hours accrued between the initial fee request and the District Court's fairness hearing) at a pre-multiplier average rate of $726 per hour. Broken down by category, these hours were attributed to pre-litigation investigation (92 hours), drafting of complaints (262 hours), case development/administration (279 hours), motion practice/memoranda drafting/legal research (246 hours), negotiation/settlement process (172 hours), discovery (379 hours), court hearings/appearances (77 hours), post-filing investigation (250 hours), class claims administration (317 hours), communications with consultants/experts (37 hours), settlement (247 hours), and the final approval process (518 hours).

Despite finding that Class Counsel's hours for certain activities seemed “high,” the District Court ultimately approved them in full on the ground that the case was 'a complex class action” featuring claims of a “technical nature.” App. 17, 19. Class Counsel’s additional 164 hours and higher Page 6average billing rate added approximately $144,000 for a new baseline lodestar of about $2.1 million. As for the lodestar multiplier, the District Court again considered the factors discussed in Gunter along with those referenced in In re AT&T Corp. Securities Litigation —another equitable common-fund case calculating fees as a percent of the fund—including the value of benefits flowing to class members as a result of the settlement and the settlement’s “innovative terms.” 455 F.3d 160, 165 (3d Cir. 2006) (citation modified). This time, however, the District Court reduced the multiplier from 1.94 to 1.75 which-yet again-produced an award of exactly $3.7 million. BMW then timely filed this second appeal, to which we now turn.

The District Court had jurisdiction under the Class Action Fairness Act of 2005. 28 U.S.C. § 1332(d). We have jurisdiction under 28 U.S.C. § 1291.

We review de novo the legal standards used by a district court in calculating a fee award. Gelis I , 49 F.4th at 377. But “‘so long as [the district court] employs correct standards and procedures and makes findings of fact that are not clearly erroneous,’ a district court has discretion to decide the amount of an award.” Id. (citation modified) (quoting In re Rite Aid Corp. Sec. Litig. , 396 F.3d 294, 299 (3d Cir. 2005), as amended (Feb. 25, 2005)). A district court must also “clearly set forth [its] reasoning for fee awards so that we will have a sufficient basis to review for abuse of discretion.' Id. (citation modified).

Page 7BMW argues that the District Court erred in awarding Class Counsel $3.7 million in attorneys' fees for three reasons: (A) the Settlement Agreement did not authorize any lodestar multiplier; (B) even if a multiplier were authorized, no enhancement was justified in this case; and (C) the District Court miscalculated the baseline lodestar because it erred in finding all of Class Counsel’s requested hours reasonable. We consider these arguments in turn.

As we observed in Gelis I , the District Court 'awarded the attorney’s fees pursuant to a contract—the [S]ettlement [A]greement-not pursuant to a statute.' 49 F.4th at 381 (quoting In re Home Depot Inc. , 931 F.3d at 1082). We therefore look to the terms of that contract in the first instance.

The Settlement Agreement provides for the award of “reasonable attorneys’ fees” without an accompanying definition or direction as to methodology. Supp. App. 104. But it does direct us to the relevant body of law to interpret the term. While the Agreement states as a general matter that it is to “be interpreted and enforced pursuant to New Jersey law,” it provides specifically that 'Class Counsel's motion for attorneys’ fees and expenses” is governed by “[f]ederal law.” Supp. App. 137. Consistent with that choice-of-law provision, both parties extensively rely on federal law in their briefing before this Court, implicitly acknowledging that this appeal is Page 8governed by federal law.² And under federal law, 'reasonable attorneys' fees' is a term of art in the context of statutory fee-shifting cases that has historically included the potential for lodestar enhancements.

That history goes back decades. By way of example, 42 U.S.C. § 1988(b) entitles prevailing parties to 'a reasonable attorney's fee,' that we and other courts have interpreted to occasionally permit a lodestar multiplier. See, e.g. , Institutionalized Juvs. v. Sec’y of Pub. Welfare , 758 F.2d 897, 921 (3d Cir. 1985); Perotti v. Seiter , 935 F.2d 761, 765 (6th Cir. 1991). We have construed the Clean Water Act's authorization of “reasonable attorney and expert witness fees” for prevailing parties in citizen suits to include the possibility of a multiplier. 33 U.S.C. § 1365(d); see Student Pub. Int. Rsch. Grp. of N.J., Inc. v. AT & T Bell Lab’ys , 842 F.2d 1436, 1452-54 (3d Cir. 1988). The same is true of many other federal fee-shifting statutes that award prevailing parties reasonable fees. See City of Burlington v. Dague , 505 U.S. 557, 562 (1992) (noting that “case law construing what is a ‘reasonable’ fee applies uniformly to all” similarly worded federal statutes);Page 9 Perdue , 559 U.S. at 553 (concluding that lodestar awards may be enhanced in some situations); see also 42 U.S.C. § 7604(d) (Clean Air Act); 42 U.S.C. § 2000e-5(k) (Title VII).

In the face of that precedent, there is no traction to BMW’s contention that “reasonable attorneys’ fees” is an ambiguous term and that this ambiguity, along with the Settlement Agreement’s silence on lodestar multipliers, precludes them. To the contrary, because the Agreement specified that federal law governs Class Counsel's motion for attorneys' fees and expenses, and federal law has long permitted the use of lodestar multipliers to calculate reasonable attorneys’ fees, the District Court was authorized under the Agreement to consider a multiplier.

We next address whether the District Court applied the proper legal standard in awarding a 1.75 lodestar multiplier, considering: (1) whether the Settlement Agreement contains a fee-shifting provision; (2) Perdue 's guidance on the use of multipliers in statutory fee-shifting cases; (3) whether the logic of Perdue applies with equal force in contractual fee-shifting cases governed by federal law; and (4) the implications of our analysis for this case.

We begin by classifying the fee provision in the Settlement Agreement. “The general rule in our legal system is that each Page 10party must pay its own attorney’s fees and expenses,” Perdue , 559 U.S. at 550, leaving each side to negotiate a fee arrangement with its own counsel. This “bedrock principle” is known as the American Rule. Hardt v. Reliance Standard Life Ins. Co. , 560 U.S. 242, 253 (2010). But that Rule has exceptions.

One is fee shifting. Under a fee-shifting regime, an attorney’s fees are paid not by her client, but by the opposing party. Fee shifting may occur “(1) when a statute grants courts the authority to direct the losing party to pay attorney’s fees; (2) when the parties agree in a contract that one party will pay attorney’s fees; and (3) when a court orders one party to pay attorney's fees for acting in bad faith.' In re Home Depot Inc. , 931 F.3d at 1078.

Another exception is the “equitable common-fund” doctrine. A common fund arises, typically in class actions, when litigation generates economic value to a large and defined class of beneficiaries without the promise of fees from a fee arrangement. See Boeing Co. v. Van Gemert , 444 U.S. 472, 478-80 (1980). Where an agreement or court order creates a common fund, the prevailing party's attorney is generally 'entitled to a reasonable attorney's fee from the fund as a whole' even though the attorney has not negotiated a fee arrangement with the client.³ Id. at 478. This payment Page 11mechanism 'reflects the traditional practice in courts of equity' that 'persons who obtain the benefit of a lawsuit without contributing to its cost [should not be] unjustly enriched at the successful litigant’s expense.” Id.

There is also a third, hybrid exception, known as a “constructive common fund,” whereby courts sometimes treat fee-shifting arrangements under the rules applicable to equitable common funds. See, e.g. , In re Gen. Motors Corp. Pick-Up Truck Fuel Tank Prods. Liab. Litig. , 55 F.3d 768, 820 (3d Cir. 1995); In re Home Depot Inc. , 931 F.3d at 1080; In re Bluetooth Headset Prods. Liab. Litig. , 654 F.3d 935, 943 (9th Cir. 2011); Johnston v. Comerica Mortg. Corp. , 83 F.3d 241, 245-46 (8th Cir. 1996). The constructive common-fund doctrine recognizes the “practical realit[y]” that defendants are concerned only with their aggregate liability, In re Gen. Motors Corp. , 55 F.3d at 819, and that, absent court supervision, a defendant may “pay class counsel excessive fees and costs in exchange for counsel accepting an unfair settlement,' Staton v. Boeing Co. , 327 F.3d 938, 965 (9th Cir. 2003) (citation Page 12modified); see also In re Bluetooth Headset Prods. Liab. Litig. , 654 F.3d at 943. In such circumstances, courts may treat contractual fee-shifting awards as common funds—and therefore exercise greater scrutiny of attorneys' fees-to protect a class’s recovery.

There are important distinctions between common-fund and fee-shifting cases. The “key distinction” is whether “the attorney’s fees are paid by the client (as in common-fund cases) or by the other party (as in fee-shifting cases).” In re Home Depot Inc. , 931 F.3d at 1079; see also Staton , 327 F.3d at 967. Another difference is the presumptive methodology for calculating attorneys' fees, with the percent-of-fund method preferred for common-fund cases and the lodestar method preferred for fee-shifting cases. In re Prudential Ins. Co. Am. Sales Prac. Litig. Agent Actions , 148 F.3d 283, 333 (3d Cir. 1998); In re Rite Aid Corp. Sec. Litig. , 396 F.3d at 300. At the same time, these correlations are not hard and fast. Rather, we, like most of our sister circuits, allow district courts to determine common-fund fees either as a percentage of the fund or by performing a lodestar calculation. See, e.g. , In re Wawa, Inc. Data Sec. Litig. , 85 F.4th 712, 722 & n.18 (3d Cir. 2023).⁴

Page 13So how do we categorize the Settlement Agreement? By its terms, 'all attorneys' fees and costs . . . will be paid by [BMW],” and all fees “will be paid separate and apart from any relief provided to the Settlement Class.” Supp. App. 129-30. In other words, the Settlement Agreement does not force the 'winning party [to] pay[] his or her own attorneys' fees.' Staton , 327 F.3d at 967. Nor does it direct that fees be drawn from the class settlement fund. See Van Gemert , 444 U.S. at 478. Instead, as part of the class action settlement, it shifts the onus to BMW to “cover[] the bill.” Staton , 327 F.3d at 967; see In re Home Depot Inc. , 931 F.3d at 1079-80 (construing nearly identical language as a fee-shifting provision). The Settlement Agreement thus constitutes a fee-shifting arrangement that arises in the context of a contractual case.⁵

Having established that the Settlement Agreement sets out a fee-shifting arrangement and mindful that the Supreme Court has constricted the use of multipliers in the statutory fee-shifting context, we now briefly review the evolution of that case law and the Supreme Court’s reasoning in Perdue to determine whether Perdue applies in the context of contractual fee-shifting cases.

We start with a brief background on the lodestar methodology and the role of a multiplier. The concept of the baseline lodestar to calculate attorneys’ fees—the product of an attorney’s (reasonable) hours times her reasonable hourly Page 15billing rates-was 'pioneered' by this Circuit in 1973, Perdue , 559 U.S. at 551 (citing Lindy Bros. Builders, Inc. of Phila. v. Am. Radiator & Standard Sanitary Corp. , 487 F.2d 161 (3d Cir. 1973)), and has since “achieved dominance in the federal courts” in statutory fee-shifting cases, Gisbrecht v. Barnhart , 535 U.S. 789, 801 (2002). The lodestar method has two core “virtues.” Perdue , 559 U.S. at 551. First, it awards attorneys their 'prevailing market rates in the relevant community,' which satisfies “the aim of fee-shifting statutes.” Id. (citation modified). Second, “the lodestar method is readily administrable' and 'objective,' thereby limiting discretion, permitting judicial review, and producing predictable results. Id. at 551-52 (citation modified).

In some cases, however, courts concluded that the lodestar alone was not sufficient and sought to enhance it. The lodestar multiplier historically filled that role and was applied by the Supreme Court to account for such factors as:

(1) the time and labor required; (2) the novelty and difficulty of the questions; (3) the skill requisite to perform the legal service properly; (4) the preclusion of employment by the attorney due to acceptance of the case; (5) the customary fee; (6) whether the fee is fixed or contingent; (7) time limitations imposed by the client or the circumstances; (8) the amount involved and the results obtained; (9) the experience, reputation, and ability of the attorneys; (10) the “undesirability” of the case; (11) the nature and

Page 16length of the professional relationship with the client; and (12) awards in similar cases.

Hensley v. Eckerhart , 461 U.S. 424, 430 n.3 (1983); see also id. at 434 & n.9.

Yet even as the lodestar itself ascended in popularity, courts began to shy away from multipliers in statutory fee-shifting cases. In 1984, the Supreme Court held that one such enhancement factor-the 'novelty and complexity of the issues”—could not justify an enhancement because it is “fully reflected in the number of billable hours recorded by counsel” or, in some circumstances, in counsel’s hourly rates. Blum v. Stenson , 465 U.S. 886, 898 (1984). The Court also limited the applicability of representation-quality enhancements, cautioning against them unless supported by “specific evidence' of superior performance and, even then, only in 'exceptional' circumstances. Id. at 899 (quoting Hensley , 461 U.S. at 435).

Two years later, the Court expressed skepticism of multipliers more generally, noting a “strong presumption that the lodestar figure . . . represents a 'reasonable' fee' and instructing that, because the baseline lodestar already includes 'most, if not all, of the relevant factors' that courts may consider in setting reasonable attorneys' fees, using such factors to justify enhancements could amount to 'double counting.” Pennsylvania v. Del. Valley Citizens’ Council for Clean Air ( Delaware Valley I ), 478 U.S. 546, 565-66 (1986). The Court reiterated that view the following year, expressing Page 17doubt that contingency risk enhancements were permissible under federal fee-shifting statutes. Pennsylvania v. Del. Valley Citizens’ Council for Clean Air ( Delaware Valley II ), 483 U.S. 711, 720-23 (1987). And a few years later, in City of Burlington v. Dague , the Court clarified that contingency risk could never justify a multiplier. See 505 U.S. at 562, 567.

The Dague Court reasoned that contingency risk is a product of “(1) the legal and factual merits of the claim, and (2) the difficulty of establishing those merits.' Id. at 562. Because the difficulty of establishing the merits is reflected in the hours claimed by counsel or in the higher hourly-rate charged by a sufficiently skilled litigator, the Court viewed contingency risk as subsumed in the baseline lodestar. See id. It also observed that awarding multipliers for taking cases with weak merits would have the perverse effect of 'provid[ing] attorneys with the same incentive to bring relatively meritless claims as relatively meritorious ones.' Id. at 563.

The Supreme Court’s gradual distancing from the use of multipliers in cases through the 1990s became an unmistakable rejection in all but the most exceptional circumstances in 2010 with its opinion in Perdue . Perdue held that the lodestar typically subsumed not only factors like complexity and contingency risk, but also superior results and performance. While acknowledging that superior results “may be attributable to superior performance and commitment of resources by plaintiff's counsel,' the Court also recognized that superior results may be attributable to circumstances unrelated to plaintiff’s counsel’s skill, like inferior performance by Page 18opposing counsel, 'an unexpectedly sympathetic jury, or simple luck.” Perdue , 559 U.S. at 554. And even when there has been superior performance, the Court emphasized, an attorney’s skill is ordinarily reflected in her hourly rate. Id. at 553.

For these reasons, the Court deemed performance-enhancements appropriate only when 'superior attorney performance is not adequately taken into account in the lodestar calculation.' Id. at 554. Such circumstances include: (1) where the lodestar is calculated using an hourly rate that 'does not adequately measure the attorney's true market value,” e.g., if the hourly rate is based on “only a single factor (such as years since admission to the bar),' id. at 554-55, (2) where “the attorney’s performance includes an extraordinary outlay of expenses and the litigation is exceptionally protracted,” id. at 555, and (3) where the attorney has faced an 'exceptional delay in the payment of fees,” id. at 556. These circumstances, the Court cautioned, will be “‘rare’ and ‘exceptional,’ and require specific evidence that the lodestar fee would not have been ‘adequate to attract competent counsel.’” Id. at 554 (quoting Blum , 465 U.S. at 897); see also id. at 555 (requiring district courts to link the magnitude of the enhancement to objective evidence that justifies the enhancement).

So what is the current law of the land in the statutory fee-shifting context? District courts may still-at least in theory—use lodestar multipliers to account for a small subset of the factors outlined in Hensley . But they may do so only on Page 19the rare occasion when those factors are not already subsumed in the baseline lodestar, and only after overcoming the “‘strong presumption' that the lodestar figure is reasonable.' Id. at 554. Even then, the lodestar enhancement must be supported by “specific evidence” to ensure that the fee award is “capable of being reviewed on appeal,” id. at 553 (citation modified), and must be accompanied by “a reasonably specific explanation for all aspects” of the court’s decision to apply a multiplier, id. at 558.

We and our sister circuits have adhered to this guidance, reaffirming that the lodestar “carries a strong presumption of reasonableness and includes most, if not all, of the relevant factors constituting a reasonable attorney’s fee,” Souryavong v. Lackawanna County , 872 F.3d 122, 128 (3d Cir. 2017) (citation modified), and that enhancements are permissible only in “the rare circumstances in which the lodestar does not adequately take into account a factor that may properly be considered in determining a reasonable fee,” Augustyn v. Wall Twp. Bd. of Educ. , 139 F.4th 252, 260 (3d Cir. 2025) (citation modified); Parsons v. Ryan , 949 F.3d 443, 467 (9th Cir. 2020); Black v. SettlePou, P.C. , 732 F.3d 492, 502 (5th Cir. 2013); Millea v. Metro-North R.R. Co. , 658 F.3d 154, 168-69 (2d Cir. 2011).⁶

Although Perdue , as a statutory fee-shifting case, is not strictly binding in contractual fee-shifting cases, a lack of 'technically binding' authority does not give us license to “blithely disregard . . . precedent where its reasoning applies.” In re Home Depot Inc. , 931 F.3d at 1085. And we are persuaded here that the reasoning of Perdue and its predecessors applies with equal force to contractual fee-shifting cases. After all, the Court’s primary rationale for restricting multipliers in Perdue was simply that they double-count factors already subsumed in the lodestar, thereby Page 21increasing the award beyond a presumptively “reasonable” fee. 559 U.S. at 553-54; Dague , 505 U.S. at 562-63.

We see no reason that the Supreme Court’s definition of a “reasonable” fee—a fee sufficient to “enable private parties to obtain legal help' without awarding attorneys a windfallwould differ where the fee shifting happens to be contractual rather than statutory. Delaware Valley I , 478 U.S. at 565. In neither context is fee shifting “designed as a form of economic relief to improve the financial lot of attorneys.” Id. ; see also Blum , 465 U.S. at 893-95 (recounting legislative history of 42 U.S.C. § 1988 and noting that fees should 'not produce windfalls to attorneys”). And in both contexts, the lodestar is comprised of the same two factors—the number of hours that similarly skilled attorneys would ordinarily devote to the litigation and the “prevailing market rates” for “private counsel of comparable experience, skill, and reputation.” Blum , 465 U.S. at 892 n.5. So, in both types of cases, the lodestar method avoids a windful and 'produces an award that roughly approximates the fee that the prevailing attorney would have received if he or she had been representing a paying client.” Perdue , 559 U.S. at 551.

Most, if not all, of the other concerns articulated by the Court in the statutory fee-shifting context also carry over. 7Page 22 Compensating for contingency risk in contract cases likewise could motivate attorneys to litigate unmeritorious cases, Dague , 505 U.S. at 563, and produce 'burdensome satellite litigation” to determine counsel’s eligibility for enhancements, id. at 566. In both contexts, compensating for superior results risks enhancing fees based merely on “inferior performance by defense counsel” or “simple luck.” Perdue , 559 U.S. at 554. Also in both cases, attorneys 'understand[] that payment of fees will generally not come until the end of the case, if at all,” so lodestar enhancements for the outlay of expenses and delays in payment are reasonable only when such expenses or delays are truly “exceptional.” Id. at 555-56. In short, where federal law applies and fees are awarded under the lodestar method, fees awarded by contract, no less than fees awarded by statute, are subject to the strictures of Perdue .

In coming to this conclusion, we are in good company. No Court of Appeals has held to the contrary, and the Sixth, Ninth, and Eleventh Circuits all agree that Perdue and its predecessors apply when a contractual fee-shifting provision is evaluated under federal law. See Chambers v. Whirlpool Corp. , 980 F.3d 645, 656, 665 (9th Cir. 2020); Linneman v. Vita-Mix Corp. , 970 F.3d 621, 629 n.2, 632 (6th Cir. 2020); In re Home Depot Inc. , 931 F.3d at 1079-80, 1086.

Page 23We pause, however, to emphasize the limits of our decision. Our holding applies (1) to fees awarded under contractual fee-shifting provisions, (2) where the contract calls for the award of reasonable attorneys' fees as defined by federal law, and (3) where the district court employs the lodestar method. We express no opinion on whether statutory fee-shifting case law applies to fees awarded under state law, which federal courts may be called upon to apply, e.g. , Chieftain Royalty Co. v. Enervest Energy Institutional Fund XIII-A, L.P. , 888 F.3d 455, 461 (10th Cir. 2017); In re Volkswagen & Audi Warranty Extension Litig. , 692 F.3d 4, 14-17 (1st Cir. 2012), and which may authorize enhancements more (or less) readily than federal law, see In re Apex Oil Co. , 297 F.3d 712, 719-20 (8th Cir. 2002) (affirming enhancement under Texas law); Mangold v. Cal. Pub. Utils. Comm’n , 67 F.3d 1470, 1479 (9th Cir. 1995) (affirming enhancement under California law).

Nor do we opine on whether Perdue applies to contractual fee-shifting awards treated as constructive common funds, or to equitable common-fund awards—an issue that neither this Court nor the Supreme Court has directly addressed.⁸ See Brytus v. Spang & Co. , 203 F.3d 238, 243-44 (3d Cir. 2000).

Page 24Finally, our opinion does nothing to disrupt our existing case law permitting district courts to employ a multiplier in the context of a so-called 'lodestar cross-check,' where the lodestar method is used “to cross-check the reasonableness of a percentage-of-recovery fee award.' In re Cendant Corp. PRIDES Litig. , 243 F.3d 722, 742 n.26 (3d Cir. 2001); see also In re Rite Aid Corp. Sec. Litig. , 396 F.3d at 306. As we have explained, when the lodestar method is used for that purpose, we require “neither mathematical precision nor bean-counting,” and permit district courts to “rely on summaries submitted by the attorneys,” In re Rite Aid Corp. Sec. Litig. , 396 F.3d at 306-07, without the need to abide by Dague ’s limits on lodestar enhancements, see In re Cendant Corp. PRIDES Litig. , 243 F.3d at 742 n.26. Of course, even where “multipliers for risk or counsel’s expertise are Page 25appropriate in the lodestar cross-check” they still 'require . . . scrutiny and justification.' Brytus , 203 F.3d at 243 (citation modified).

The Settlement Agreement calls for the application of federal law, awards fees under a contractual fee-shifting arrangement, and assesses fees according to the lodestar method. But without the benefit of our decision today, the District Court applied a 1.75 lodestar multiplier that did not account for Perdue ’s limitations, including its “strong presumption” that the unenhanced lodestar figure was reasonable. 559 U.S. at 554.

We cannot say on this record that this error was harmless. To the contrary, the District Court explicitly based the enhancement on a number of factors we now recognize were subsumed under the lodestar. Among other factors, it rested on Class Counsel's 'risk of nonpayment' (because Class Counsel took “the case on a contingency basis”), “the complex and technical” issues in the case, “the size of the fund,” the settlement’s “innovative terms,” and its observation that Class Counsel was 'highly regarded' and 'highly skilled.' App. 28-32 (citation modified). But the first two factors are improper considerations under Perdue and its predecessors because they are always subsumed in the baseline lodestar, see Perdue , 559 U.S. at 553, 558; Dague , 505 U.S. at 562 (contingency risk); Blum , 465 U.S. at 898-99 (complexity), and the remainder boil down to attorney performance which, Page 26following Perdue , can serve as the basis of lodestar enhancements only in “rare” and “exceptional” circumstances, 559 U.S. at 554 (citation modified). Yet the District Court did not explain how, if at all, this case qualifies as exceptional or why the lodestar alone would not have attracted similarly skilled counsel. The District Court also double-counted the case’s technical complexity when setting fees, relying on that complexity both to approve Class Counsel's unusually high number of hours and to approve the enhancement.

Nor was the enhancement in this case supported by any specific record evidence that it was reasonable, or an explanation of the “amount of the enhancement . . . attributable to [each] factor.' Id. at 558. Yet Purdue teaches that “when a trial judge awards an enhancement on an impressionistic basis, a major purpose of the lodestar method-providing an objective and reviewable basis for fees—is undermined.” Id. (citation omitted). Given these defects in the enhancements, the District Court’s fee award cannot stand.

We turn next to BMW’s argument that the baseline lodestar itself was unreasonable because (1) over 80% of the 2,877 hours requested by Class Counsel were billed by partners at partner rates, and (2) some of those hours were inefficient or duplicative, including 262 hours for drafting complaints, 374 hours for discovery, and 172 hours for mediation, among others.

Page 27When it comes to setting the baseline lodestar— determining the reasonable number of hours and the reasonable hourly rate—district courts have “substantial discretion . . . because they are ‘better informed than an appellate court about the underlying litigation and an award of attorney fees is fact specific.’” United States ex rel. Palmer v. C&D Techs., Inc. , 897 F.3d 128, 137 (3d Cir. 2018) (quoting Pub. Interest Rsch. Grp. of N.J., Inc. v. Windall , 51 F.3d 1179, 1184 (3d Cir. 1995)). But that discretion is not unlimited. See Gelis I , 49 F.4th at 377. And here, given the number of hours claimed by Class Counsel for the tasks in question, and the fact that the vast majority were incurred by partners at partner billing rates, we cannot conclude that the District Court acted within those limits.

We begin by addressing the overall proportion of work performed by partners-a startling 80% (more than 2,300 hours). We have no doubt that this proportion is unusual when compared with other class actions and that, generally, over-allocation of work to partners may justify reducing attorneys’ fees. See Ursic v. Bethlehem Mines , 719 F.2d 670, 677 (3d Cir. 1983). After all, a 'Michelangelo should not charge Sistine Chapel rates for painting a farmer’s barn.” Id. Here, the District Court gave Class Counsel “some leeway” in the allocation of work to partners because of what it viewed as the 'uniquely complex nature of this case' such that only partners, with specialized knowledge, could “adequately perform [certain] tasks.” App. 24 (citation modified). Even so, that specialized knowledge-as well as the years of Page 28experience reflected in partners' significantly higher billing rates-is reasonably expected to produce commensurately greater billing efficiencies in performing tasks. And where, as here, hours are billed at partner rates because of the alleged need for expert and specialized knowledge, but the number of hours billed does not reflect any greater efficiency, there is reason for concern about impermissible '[d]ouble-dipping.' Ursic , 719 F.2d at 677.

That concern prompted us, in BMW’s first appeal, to vacate and remand for the District Court to develop the record and to address “whether certain hours [we]re duplicative . . . or whether the total hours billed were reasonable for the work performed.” Gelis I , 49 F.4th at 380. We appreciate the fuller explanation the District Court has now provided. But we do not view that explanation as justifying approval of the full 2,877 hours claimed by Class Counsel.

As for the 262 hours spent drafting complaints, 222 of which were claimed by partners, the District Court acknowledged that the hours were 'excessive . . . at first blush,” but nonetheless approved them on the ground that the work was “deceptively substantial.” App. 16. As recounted by the District Court, the complaints raised 'over a dozen claims” involving multiple different state consumer-protection laws, required Class Counsel to 'expend[] . . . time to vet possible class representatives,” involved three separate iterations, and dealt with issues that were “highly technical in nature.” App. 16-17. Yet neither individually nor collectively do these circumstances render 262 hours reasonable. The fact Page 29that an inordinate number of hours spent drafting were performed by partners-a fact not addressed by the District Court—makes that number more, not less, problematic. The number of claims raised in the complaints and the need to vet class representatives are hardly unusual. Though Class Counsel did draft three complaints, the first and second were “largely . . . identical,” App. 16, and the third differed in only minor respects, see Gelis v. BMW of N. Am., LLC , No. 2:17-cv-07386, Dist. Ct. Dkt. No. 45-2. And though technical in nature, the alleged timing chain defect was not so complex as to justify over six forty-hour workweeks of full-time drafting. The requested hours are particularly unreasonable when considering that the drafters were experienced partners and that Class Counsel separately claimed nearly 100 hours for “pre-litigation investigation” of the timing-chain defect—the “bulk” of which were billed by a partner who was “trained as a mechanic and therefore was uniquely positioned” to perform an efficient investigation. App. 16.

The hours approved for discovery also appear excessive. Class Counsel claimed approximately 100 hours (nearly all by partners) for “reviewing BMW’s documents,” which included “over 12,000 pages of largely technical materials produced by BMW . . . without the benefit of an index or other organization.” App. 20-21. The District Court found those hours reasonable because Class Counsel reviewed documents at a rate of 'approximately thirty seconds per page' which other courts have found 'to be an 'objectively reasonable pace.'' App. 20-21 (quoting Scott Hutchison Enters. v.

Page 30Cranberry Pipeline Corp. , 318 F.R.D. 44, 57 (S.D. W. Va. 2016)). But as with the hours spent drafting the complaints, it is not clear that the District Court accounted for the expectation that experienced partners with high billing rates will review documents with greater efficiency and thus require fewer hours.

The District Court also approved an additional 279 hours, again almost all by partners, attributed to 'production by Plaintiffs, meet and confers, communication between Class Counsel and their clients, and correspondence and conferences with the Court.” App. 21. According to the District Court, those hours were “reasonable given the circumstances of the case.” Id. This explanation does not indicate that the District Court factored in Class Counsel’s top-heavy staffing model, does not identify any circumstances surrounding discovery— such as the specifics of Plaintiffs’ productions or the frequency of the parties’ meet and confers—that would render these hours reasonable, and is not “a sufficient basis to review for abuse of discretion,” Gelis I , 49 F.4th at 377 (citation modified).

Finally, the District Court concluded it was not unreasonable for Class Counsel to bill 172 hours for “negotiation and settlement,” including 97 hours for sending three partners to mediation, so long as those partners “[we]re not unreasonably doing the same work and [we]re being compensated for the[ir] distinct contribution[s].” App. 19-20 (quoting Norman v. Hous. Auth. of Montgomery , 836 F.2d 1292, 1302 (11th Cir. 1988)). Fair enough. But the Court made no findings, nor does the record reflect, that each partner Page 31was indeed doing different work or making distinct contributions.

In short, in addition to the problematic lodestar enhancements, the errors here in calculating the baseline lodestar also require us to vacate and remand the fee award. We do not opine on what portion of Class Counsel’s time spent drafting complaints, performing discovery, or mediating was reasonable and will leave that to the District Court on remand. With our additional guidance today, we have every confidence that the District Court’s calculation of attorneys’ fees on this remand will be reasonable in view of both Perdue and the record.

For the foregoing reasons, we will vacate the District Court's order awarding attorneys' fees and will remand for further proceedings consistent with this opinion.

Page 32Counsel for Appellant Melissa Bayly Christopher J. Dalton [Argued] Argia J. DiMarco BUCHANAN INGERSOLL & ROONEY

Counsel for Appellees Gary S. Graifman KANTROWITZ GOLDHAMER & GRAIFMAN

Bruce H. Nagel [Argued]

Robert H. Solomon

NAGEL RICE