Baerbel McKinney-drobnis v. Massage Envy Franchising, LLC

FOR PUBLICATION UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT BAERBEL MCKINNEY-DROBNIS; No. 20-15539 JOSEPH B. PICCOLA; CAMILLE BERLESE, individually and on behalf D.C. No. of all others similarly situated, 3:16-cv-06450- Plaintiffs-Appellees, MMC v. OPINION KURT ORESHACK, Objector-Appellant, v. MASSAGE ENVY FRANCHISING, LLC, Defendant-Appellee. Appeal from the United States District Court for the Northern District of California Maxine M. Chesney, District Judge, Presiding Argued and Submitted June 8, 2021 Seattle, Washington Filed October 20, 2021 2 MCKINNEY-DROBNIS V. ORESHACK Before: Ronald Lee Gilman, * Ronald M. Gould, and Eric D. Miller, Circuit Judges. Opinion by Judge Gould; Concurrence by Judge Miller SUMMARY ** Class Action Settlement The panel vacated the district court’s judgment overruling objections, certifying a class for settlement, approving the settlement, and granting most of class counsel’s requested fee award in a class action arising out of a dispute between Massage Envy Franchising, LLC (“MEF”), a membership-based spa-services company, and a putative nationwide class of current and former members. The class complaint alleged that MEF began periodically increasing membership fees in violation of the membership agreement. After extensive discovery and motions for class certification and summary judgment, the parties settled. In exchange for the release of all claims against MEF, class members could submit claims for “vouchers” for MEF products and services. The district court approved the settlement as “fair, reasonable, and adequate” under Fed. R. Civ. P. 23(e). Objector Kurt Oreshack challenged the * The Honorable Ronald Lee Gilman, United States Circuit Judge for the U.S. Court of Appeals for the Sixth Circuit, sitting by designation. ** This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader. MCKINNEY-DROBNIS V. ORESHACK 3 approval, contending that the vouchers provided to the class under the settlement were “coupons” under the Class Action Fairness Act (“CAFA”); and further contending that even if CAFA’s coupon restrictions did not apply, the district court abused its discretion by disregarding warning signs of class counsel’s self-interest that warranted additional scrutiny. The panel held that the district court erred in finding that the vouchers were not “coupons” under CAFA. If a form of class action settlement is considered a “coupon” under CAFA, then additional restrictions apply to the settlement- approval process. The panel, following the Fourth Circuit’s approach, held that de novo review applied to determine the applicability of CAFA’s coupon provisions. The panel, therefore, did not defer to the district court’s determination that the MEF vouchers were not coupons under CAFA. In In re Online DVD-Rental Antitrust Litigation, 779 F.3d 934 (9th Cir. 2015), this court outlined a three-factor test for determining whether an award constituted a coupon settlement. The panel held that the first factor – whether settlement benefits require class members “to hand over more of their own money before they can take advantage of” those benefits – was inconclusive as to whether the vouchers were coupons. The panel further held that under factor two – whether the credit was valid only for “select products or services” – the vouchers appeared to be coupons. Under factor three – how much flexibility the credit provided – the panel held that the vouchers were flexible, and this favored not viewing the vouchers as coupons. The panel concluded that no single Online DVD factor was dispositive, and held under de novo review that the vouchers were coupons, and subject to CAFA’s requirements for coupon settlements. The panel vacated the district court’s approval of the attorneys’ fee award and remanded for the district court to 4 MCKINNEY-DROBNIS V. ORESHACK use the value of the redeemed vouchers in awarding fees, as required by 28 U.S.C. § 1712(a). The panel next addressed Oreshack’s contention that, independent of CAFA’s applicability to the fee award, the district court erred by approving the settlement as “fair, reasonable, and adequate” under Rule 23(e). The panel noted, as a preliminary matter, that determining that vouchers were coupons under CAFA and vacating the fee award, did not necessarily require invalidating the entire settlement approval order. But given the objector’s challenge to the settlement agreement, the panel analyzed the entire agreement for fairness. The panel held that the district court abused its discretion by failing to adequately investigate and substantively grapple with some of the potentially problematic aspects of the relationship between attorneys’ fees and the benefits to the class. Because the errors made by the district court impacted the fairness of the entire settlement under Rule 23(e), and not just attorneys’ fees, the panel vacated the approval and remanded for the district court to analyze more deeply whether the settlement should be approved. Specifically, the panel held that the district court abused its discretion in failing to apply the requisite heightened scrutiny for pre-certification settlements. The district court did not apply the appropriate enhanced scrutiny because it failed to adequately investigate and address the three warning signs of implicit collusion articulated in In re Bluetooth Headset Products Liability Litigation, 654 F.3d 935 (9th Cir. 2011). Judge Miller joined the court’s opinion in full, but wrote separately to note his disagreement with this court’s approach to determining when vouchers are “coupons” MCKINNEY-DROBNIS V. ORESHACK 5 under CAFA. He wrote that in an appropriate case, the court should reconsider en banc In re Online DVD-Rental Antitrust Litigation, 779 F.3d 934 (9th Cir. 2015), and its three-factor test for determining whether an award constitutes a coupon settlement. COUNSEL Adam E. Schulman (argued) and Theodore H. Frank, Hamilton Lincoln Law Institute, Center for Class Action Fairness, Washington, D.C., for Objector-Appellant. Trenton R. Kashima (argued), Sommers Schwartz P.C., San Diego, California; Jeffery R. Krinsk and John J. Nelson, Finkelstein & Krinsk LLP, San Diego, California; for Plaintiffs-Appellees. Theodore J. Boutrous Jr. (argued), Kahn A. Scolnick, Martie P. Kutscher, and Daniel R. Adler, Gibson Dunn & Crutcher LLP, Los Angeles, California; Luanne Sacks, Cynthia A. Ricketts, Robert B. Bader, and Mike Scott, Sacks Ricketts & Case LLP, San Francisco, California; for Defendant- Appellee. 6 MCKINNEY-DROBNIS V. ORESHACK OPINION GOULD, Circuit Judge: This appeal concerns a district court’s approval of a class action settlement that the parties reached before class certification. When a federal court considers whether to approve a settlement, we require the court to closely scrutinize the agreement for any evidence that class counsel’s self-interest infected the negotiations at the expense of the class. When that approval comes before the class is certified and therefore before class counsel have expended substantial resources, there is an even greater risk that class counsel will breach the fiduciary duty owed to absent class members. The class action at issue in this appeal arose out of a dispute between Massage Envy Franchising, LLC (“MEF”), a membership-based spa-services company, and a putative nationwide class of current and former members. The class complaint alleged that MEF began periodically increasing membership fees in violation of the membership agreement. The district court approved the settlement as “fair, reasonable, and adequate” under Federal Rule of Civil Procedure 23(e). Objector Kurt Oreshack challenges that approval, contending that the vouchers provided to the class under the settlement are “coupons” under the Class Action Fairness Act, 28 U.S.C. § 1712 (“CAFA”). Oreshack also contends that even if CAFA’s coupon restrictions do not apply, the district court abused its discretion by disregarding warning signs of class counsel’s self-interest that warrant additional scrutiny. We hold that (1) the district court erred in finding that the vouchers are not “coupons” under CAFA, and (2) the district court abused its discretion in failing to apply the MCKINNEY-DROBNIS V. ORESHACK 7 requisite heightened scrutiny for pre-certification settlements. Specifically, we conclude that the court did not apply the appropriate enhanced scrutiny because it failed to adequately investigate and address the three warning signs of implicit collusion that we articulated in In re Bluetooth Headset Products Liability Litigation, 654 F.3d 935 (9th Cir. 2011). I A MEF operates as a franchisor selling spa services to consumers through a system of more than 1,100 Massage Envy locations nationwide. The franchisee locations sell products and provide spa services under the Massage Envy brand name. MEF locations are membership-based. For a monthly fee, members receive one prepaid massage per month and lower prices than non-members pay for additional services. The prepaid services can accrue on the members’ accounts. Members enter into “Membership Agreements” with the franchisee location. MEF provides Membership Agreement templates to its franchisees for use with their members. Baerbel McKinney-Drobnis, Joseph Piccola, and Camille Berlese (collectively, “Plaintiffs”) represent a putative nationwide class of current and former MEF members who paid membership fee increases during the class period. The class members signed Membership Agreements with franchisees in different states (California, Arizona, and Texas). In their amended class complaint, Plaintiffs alleged that, beginning in 2013, MEF locations began unilaterally increasing membership dues without authorization. Many class members discovered an initial price increase of $0.99 per month, and then in some cases, a 8 MCKINNEY-DROBNIS V. ORESHACK second, bigger monthly increase of $10 or more. Based on the unauthorized membership fee increases, the amended complaint alleged breach of contract, intentional interference with contractual relations, and state consumer- protection-law violations. The parties vigorously dispute whether the fee increases violated the Membership Agreements that class members signed. B Six years before the settlement at issue in this case, class counsel sued MEF on behalf of different clients, asserting contract and tort claims that flowed from the loss of accrued services following membership termination. See Hahn v. Massage Envy Franchising, LLC, No. 12cv153-CAB, 2013 WL 12415927, at *1 (S.D. Cal. Feb. 25, 2013). After extensive discovery and motions for class certification and summary judgment, the parties settled. The class was divided up for settlement purposes: Hahn for former MEF members, and Zizian for current MEF members. The Hahn and Zizian settlements included a release provision that resolved the specific claims concerning accrued and unused massages and extinguished “any claim asserted or that could have been asserted in [Hahn/Zizian]” and any “claims that any Membership Agreement . . . constituted a fraudulent, unlawful, unfair, or deceptive business practice; was unconscionable; violated consumer protection statutes; and for breach of contract and breach of the covenant of good faith and fair dealing.” Both settlements were approved by the district court. After the objection and opt-out deadline expired in Zizian, Plaintiffs filed this case. MCKINNEY-DROBNIS V. ORESHACK 9 C Before the settlement in this case, the district court adjudicated several competing motions between Plaintiffs and MEF. In January 2017, Plaintiffs moved to strike MEF’s affirmative defenses. MEF responded by filing a motion for judgment on the pleadings, or in the alternative, to strike class allegations. MEF’s motion was based on the Zizian settlement release. The court granted the motion to strike as to 25 of MEF’s 29 affirmative defenses, and then denied MEF’s motion for judgment on the pleadings. The court also denied MEF’s request to certify the court’s order for interlocutory appeal. The parties began discovery. Plaintiffs propounded 55 document requests, 25 interrogatories, and several subpoenas. Plaintiffs “reviewed over 7,000 pages of documents.” Plaintiffs also benefitted from discovery regarding MEF’s business practices that came to light in the Hahn settlement. Plaintiffs began seeking electronic- discovery experts and scheduled depositions of MEF officers. These efforts continued until the date the parties reached the settlement at issue. During the discovery period, the parties periodically explored settlement. On October 27, 2017, the parties met to exchange their settlement positions. After one unsuccessful mediation and continued discovery, the parties met for a second mediation in November 2018. At the second mediation—and importantly, before any motion for class certification was filed—the parties agreed on the material terms of a settlement. 10 MCKINNEY-DROBNIS V. ORESHACK D The proposed settlement agreement was executed on March 11, 2019. The settlement class includes current and former members of MEF franchisees who paid membership- fee increases during the class period. In exchange for the release of all claims against MEF, class members can submit claims for “vouchers” for MEF products and services. The voucher that each class member receives corresponds to the fee increase the class member paid. The vouchers expire after eighteen months. The vouchers may be used at any MEF location to purchase retail products, massage sessions, enhancements, and/or facial sessions. MEF offers 251 different items for sale. The vouchers also have some flexibility because they are transferable, may be combined with other promotions and discounts, and can be used in multiple transactions until exhausted. On the other hand, the vouchers are not redeemable for cash and cannot be used to pay monthly membership fees or tips. The settlement provides for a $10 million “floor”; in other words, if class members do not claim enough vouchers to use up the full $10 million fund, the settlement will increase voucher amounts to claimants pro rata until the $10 million floor is reached. MEF also agreed to injunctive relief requiring the franchisees to adopt a template Membership Agreement that mandates a 45-day advance written notice before membership fees can be increased. Under the agreement, each named Plaintiff has the right to request a $10,000 incentive award without opposition. Two additional settlement terms are particularly relevant to this appeal. First, MEF agreed to a “clear-sailing” provision for attorneys’ fees, which means that MEF would MCKINNEY-DROBNIS V. ORESHACK 11 not object to class counsel’s fee request so long as counsel request no more than $3.3 million. Second, the settlement contains a “reverter” or “kicker” provision, which means that, if the court awards less than $3.3 million in fees, the excess funds revert to MEF rather than to the class. A direct-notice program reached an estimated 96.9% of the 1.7 million class members. After the claims period closed, a total of 105,693 class members (or about 6.2% of the class) submitted valid voucher requests. At the time of preliminary approval of the settlement, the estimated cost of notice and settlement administration was $450,000. The requested vouchers amounted to under $3 million in value, well below the $10 million floor provided in the settlement. As a result, each claimed voucher’s value was adjusted upwards on a pro rata basis in proportion to the fee increase that the class member paid. After the adjustment, the vouchers ranged in value from $36.28 to $180.68. The Attorneys General of Plaintiffs’ home states—Arizona, California, and Texas—scrutinized the settlement agreement and did not object. E The district court granted preliminary approval of the settlement in June 2019. Class counsel sought the maximum $3.3 million award that MEF had agreed not to oppose. In the fee request, counsel contended that CAFA, which governs attorneys’ fees in class-action settlements that provide for a recovery of “coupons” to class members, did not apply because the settlement vouchers were not “coupons” covered by CAFA. Class member and now Objector-Appellant Oreshack timely objected to the settlement, class certification, and attorneys’ fee request. Oreshack contended, among other 12 MCKINNEY-DROBNIS V. ORESHACK things, that: (1) the settlement was a “coupon” settlement under CAFA, but the settlement did not follow CAFA’s procedures (namely, CAFA requires class counsel’s fees to be calculated based on the value of vouchers that class members ultimately redeem, rather than the face value of the claimed 1 vouchers; here, $10 million); (2) the settlement unfairly benefits class counsel at the expense of the class because of the economic reality that many vouchers will expire unredeemed; and (3) the three Bluetooth 2 factors were present. Oreshack contended that, at a minimum, CAFA requires that the district court not award attorneys’ fees until the voucher redemption rate is known. Oreshack also requested that the court investigate the previous settlement negotiated between class counsel and MEF in Hahn; specifically, he asked the court to request that Plaintiffs’ and MEF’s respective counsel provide the redemption rate for the vouchers issued in that settlement. The class approved the proposed settlement. Of the 1.7 million class members, 523 (0.03%) opted out and 19, including Oreshack, objected. 1 We refer to vouchers for which class members submitted claims as “claimed” vouchers, and the rate at which class members submitted claims as the “claim rate.” By contrast, we refer to claimed vouchers that class members have used—submitted to the defendant business before the expiration date—as “redeemed” vouchers. We refer to the rate at which class members redeemed vouchers as the “redemption rate.” 2 In In re Bluetooth Headset Products Liability Litigation, 654 F.3d 935 (9th Cir. 2011), we described “subtle signs” that class counsel may have “allowed pursuit of their own self-interests . . . to infect the negotiations.” Id. at 947. Courts considering whether to approve a settlement exhibiting these signs must scrutinize the agreement closely for potential collusion. Id. at 946–47. MCKINNEY-DROBNIS V. ORESHACK 13 F The district court held a fairness hearing on February 28, 2020. The court overruled all objections, certified the class for settlement, approved the settlement, and granted most of class counsel’s requested fee award. The district court discussed whether the settlement was a coupon settlement with the parties. It concluded that the vouchers were not CAFA coupons because class members could purchase “quite a bit of variety” without spending their own money, the vouchers could be redeemed for a range of products that class members have shown interest in, and the vouchers provided “a fair amount of flexibility” because they can be transferred or stacked and do not expire for 18 months. In calculating attorneys’ fees, the district court stated that: “If it’s not a coupon settlement, at least [in] the Ninth Circuit, you look at what the fund is,” even though “everybody knows that that fund is not going to be used up.” The district court acknowledged that some of the voucher value would go back to MEF if the vouchers expired without being redeemed, but the court still decided to use the $10 million face value of the claimed vouchers to calculate fees instead of the expected or estimated voucher redemption rate as requested by Oreshack. The district court then calculated class counsel’s fees on a percentage-of-recovery basis by adding together the $10 million value of the voucher relief with the expected $450,000 in notice and administrative costs, and then awarding 25% of that total ($2,612,500) as the attorneys’ fee award. The district court determined that the benefits to the class were adequate relief given the significant obstacles that the class faced in litigating their claims, including the difficulty 14 MCKINNEY-DROBNIS V. ORESHACK of proving contract claims on behalf of a class that signed a variety of different membership agreements containing different language. The court also noted the low opt-out rate and that three states’ Attorneys General had scrutinized the settlement and did not object. The parties dispute the extent to which the district court considered the Bluetooth “red flag” factors, i.e., potential signs of class counsel’s self-interest. The court stated that it had not been “shown” that class counsel “receive[d] a disproportionate amount of the consideration,” and the court was not “prepared to find that there was collusion here.” The court noted that the settlement included a “clear-sailing” provision and a “reversion” to MEF of any difference between class counsel’s requested and awarded attorneys’ fees, but it determined that the settlement did not reflect collusion that weighed against settlement approval. In sum, the district court found the settlement to be “fair, reasonable, and adequate” under Federal Rule of Civil Procedure 23(e). The court issued an order granting final settlement approval on March 2, 2020, and its final judgment and dismissal order on March 20, 2020. By operation of the settlement’s reverter provision, and because the court awarded class counsel less than the requested $3.3 million, approximately $600,000 in unawarded fees reverted to MEF. Oreshack timely appealed. II We first consider whether the district court erred in finding that the vouchers are not “coupons” subject to CAFA’s restrictions. Upon de novo review, we hold that the vouchers are coupons. MCKINNEY-DROBNIS V. ORESHACK 15 A If a form of class action settlement relief is considered a “coupon” under CAFA, then additional restrictions apply to the settlement-approval process. CAFA requires courts (1) to apply “heightened scrutiny” to settlements that award “coupons” to class members, and (2) to base fee awards on the redemption value of the coupons, rather than on their face value. In re EasySaver Rewards Litig., 906 F.3d 747, 754–55 (9th Cir. 2018) (citing 28 U.S.C. § 1712). “Congress targeted such settlements for heightened scrutiny out of a concern that the full value of coupons was being used to support large awards of attorney’s fees regardless of whether class members had any interest in using the coupons.” Id. at 755 (citing S. Rep. No. 109-14, at 15–20 (2005), reprinted in 2005 U.S.C.C.A.N. 3, 15–20). Congress was concerned that if courts use the face value of the coupons—given that much of that value will go unused—the size of the settlement fund would be “inflated . . . without a concomitant increase in the actual value of relief for the class.” Id. By requiring courts to use the redemption-rate value of the coupons instead of the face value, CAFA “ensures that class counsel benefit[s] only from coupons that provide actual relief to the class.” Id. Because CAFA did not define the term “coupon,” “courts have been left to define that term on their own, informed by § 1712’s [CAFA’s] animating purpose of preventing settlements that award excessive fees while leaving class members with ‘nothing more than promotional coupons to purchase more products from the defendants.’” EasySaver, 906 F.3d at 755 (quoting In re Online DVD- Rental Antitrust Litig., 779 F.3d 934, 950 (9th Cir. 2015)). In Online DVD, we outlined three factors to guide the inquiry of whether settlement relief should be considered a 16 MCKINNEY-DROBNIS V. ORESHACK coupon under CAFA: “(1) whether class members have ‘to hand over more of their own money before they can take advantage of’ a credit, (2) whether the credit is valid only ‘for select products or services,’ and (3) how much flexibility the credit provides, including whether it expires or is freely transferrable.” EasySaver, 906 F.3d at 755 (quoting Online DVD, 779 F.3d at 951). Applying those factors in Online DVD, we concluded that a settlement providing $12 gift cards to Walmart was not a coupon settlement within the meaning of CAFA, “given Walmart’s extensive inventory of low-cost products.” Id. at 756 (citing Online DVD, 779 F.3d at 951). Importantly, the gift cards did not expire, were freely transferable, and gave class members the option to receive $12 in cash instead. Id. In EasySaver, we applied the Online DVD factors to a class action suit alleging that the defendant company enrolled the class members in a membership-rewards program without their consent and then mishandled their billing information. Id. at 752. As part of the settlement, defendants agreed to email every class member a $20 credit to purchase items on the company’s website. Id. at 753. The credits were fully transferable, but they included restrictions such as a one-year expiration date, blackout periods, an inability to use the credit for same-day orders, and an inability to combine with other promotions. Id. We held that the $20 credits were CAFA coupons because the credits were “categorically different from Walmart gift cards” due to the defendant company’s small universe of products and the numerous restrictions on class members’ use of the credits. Id. at 756–57. Most recently, we applied the Online DVD framework in a class action brought by dishwasher purchasers against the manufacturer concerning a design defect. Chambers v. MCKINNEY-DROBNIS V. ORESHACK 17 Whirlpool Corp., 980 F.3d 645 (9th Cir. 2020). We determined that the rebates offered in this settlement constituted coupon relief under CAFA, reasoning that even with the rebate, class members had to spend hundreds of dollars to purchase a dishwasher, the rebate applied only to the brands that contained the defect at issue, and the expiration period constituted a small fraction of the average life of the product. Id. at 660. The Online DVD framework thus provides us with a helpful guide in determining whether the MEF vouchers are coupons. B The parties disagree about the standard of review that we should use in analyzing whether the vouchers in question are coupons. Oreshack contends that we should review the district court’s determination that the vouchers are not coupons de novo because that determination involves statutory interpretation. See, e.g., United States v. Havelock, 664 F.3d 1284, 1289 (9th Cir. 2012) (en banc). MEF responds that interpreting the word “coupon” in Section 1712 might constitute a question of law, but it is one we have already answered by articulating the Online DVD factors; we should thus ask instead whether the district court abused its discretion in applying that established framework. We have not previously designated the correct standard of review for deciding the applicability of CAFA’s coupon provisions. See EasySaver, 906 F.3d at 755 n.5; Online DVD, 779 F.3d at 950 n.8. The Fourth Circuit recently chose de novo review, and it appears to be the only court of appeals to have expressly selected a standard of review to use in this context. In re Lumber Liquidators Chinese-Manufactured Flooring Prods. Mktg., Sales Pracs. & Prods. Liab. Litig., 952 F.3d 471, 488 (4th Cir. 2020). We now follow the Fourth Circuit’s approach and hold that de novo review 18 MCKINNEY-DROBNIS V. ORESHACK applies to determine the applicability of CAFA’s coupon provisions for the following reasons. First, determining whether CAFA applies to a particular settlement is necessarily a question of statutory interpretation. And “[l]ike all other questions of statutory interpretation,” we should review this question of law de novo. United States v. Paulk, 569 F.3d 1094, 1094 (9th Cir. 2009). In the CAFA context specifically, we tend to review the “construction, interpretation, or applicability” of CAFA’s requirements de novo. Bush v. Cheaptickets, Inc., 425 F.3d 683, 686 (9th Cir. 2005). And when, as here, “elements of legal analysis and statutory interpretation” factor into the attorneys’ fee award, those elements are properly reviewed de novo. K.C. v. Torlakson, 762 F.3d 963, 966 (9th Cir. 2014) (citation omitted). MEF is correct that we already construed the statute when we developed the Online DVD framework. But establishing a framework for interpreting a statute does not alter the fact that statutory interpretation is a legal question that we review de novo. The Online DVD factors help answer the legal question of whether settlement relief constitutes a “coupon” as Congress intended that term under CAFA. This is still “primarily a legal question” to which we should apply de novo review in seeking our answer. United States v. Marbella, 73 F.3d 1508, 1515 (9th Cir. 1996). The Supreme Court has provided guidance in this area. “[W]hen applying the law involves developing auxiliary legal principles of use in other cases,” appellate courts should use de novo review, but when questions “immerse courts in case-specific factual issues” such as weighing evidence and making credibility judgments, appellate courts should typically review with deference. U.S. Bank Nat’l MCKINNEY-DROBNIS V. ORESHACK 19 Ass’n ex rel. CWCapital Asset Mgmt. LLC v. Vill. at Lakeridge, LLC, 138 S. Ct. 960, 967 (2018). Further, section 1712(a) uses mandatory language, stating that if a proposed settlement provides for coupon relief, the attorneys’ fee award “shall” be based on the redemption value of the coupons. 28 U.S.C. § 1712(a). Given this mandatory command, it follows that abuse of discretion cannot be the appropriate standard of review for this question. See United States v. Ped, 943 F.3d 427, 433 (9th Cir. 2019) (“The word ‘shall’ generally imposes a nondiscretionary duty.” (quoting SAS Inst., Inc. v. Iancu, 138 S. Ct. 1348, 1354 (2018))). We therefore do not defer to the district court’s determination that the Massage Envy vouchers are not coupons under CAFA. Instead, we review this question de novo using the Online DVD three-factor framework as a guide. 3 The first factor—whether settlement benefits require class members “to hand over more of their own money before they can take advantage of” those benefits—is somewhat inconclusive as to whether the vouchers are coupons. On the one hand, even class members receiving the smallest voucher, in the amount of $36.28, would be able to purchase entire products without spending their own money. Cf. EasySaver, 906 F.3d at 757 (noting that, with 3 Oreshack construes Online DVD as a “limited exception” to CAFA’s coupon definition when the credits or vouchers do not expire and can be used to purchase many different types of products. He also suggests that EasySaver clarified the Online DVD holding by concluding that this exception “applies only to coupons that all class members view as ‘equivalently useful’” to cash. See EasySaver, 906 F.3d at 758. We do not construe EasySaver as establishing a mandatory checklist or a “cash equivalent” test. We construe Online DVD and EasySaver as establishing a three-factor balancing test. 20 MCKINNEY-DROBNIS V. ORESHACK shipping charges, class members would inevitably have to spend more of their own money to use the $20 credit). On the other hand, a $36.28 voucher is not enough to purchase most of Massage Envy’s services. Class members receiving the $36.28 voucher could not even purchase a single massage—the service that is the basis for the membership fee that class members were allegedly injured by—without spending their own money. Because the ability to get a massage (rather than ancillary products) is central to the membership program of Massage Envy, on balance we view this factor as favoring the conclusion that the vouchers are coupons. Under factor two—whether the credit is valid only for “select products or services”—the vouchers appear to be coupons. We recognize that MEF offers much more than massages. It sells 251 different products within the sphere of health and wellness: spa services and add-ons to spa services; skincare products such as lotions, oils, exfoliants, sun protection, anti-aging products, and skin care kits; and fitness equipment, including products like foam rollers, foot exercisers, and stretchers. Even so, the 251 products that MEF sells pale in comparison to the millions of low-cost products that Walmart sells, as in Online DVD. MEF is a highly specialized retailer and, although it offers more than 200 products, all of the products fall under the same umbrella category of health and wellness. And although the vouchers do not expressly limit which MEF products or services can be obtained using the vouchers, they are practically limited by the fact that MEF does not sell products online and not all 251 Massage Envy products and services are available at every Massage Envy location. The limited range of products and services available at Massage Envy, even considering the breadth offered within the MCKINNEY-DROBNIS V. ORESHACK 21 product category, favors viewing the vouchers as coupons under CAFA. Under factor three—how much flexibility the credit provides—the vouchers are flexible, as noted by the district court. The vouchers may be transferred, sold, and aggregated. Unlike the credits rejected in EasySaver, the vouchers here have no blackout dates and remain valid for more than one year. The vouchers do not have a “use it or lose it” restriction, meaning that class members could keep going back and buying products over time until the value of their voucher is fully extinguished. Because of these terms, MEF’s vouchers are more flexible than settlement benefits that we have held are coupons. See, e.g., In re HP Inkjet Printer Litig., 716 F.3d 1173, 1176 (9th Cir. 2013) (coupons expired in six months, were non-transferrable, and could not be used with other discounts or coupons); see also Hadley v. Kellogg Sales Co., No. 16-CV-04955-LHK, 2020 WL 836673, at *1 (N.D. Cal. Feb. 20, 2020) (coupon benefits “expire in a mere four months, must be used in a single transaction, and are only stackable to the extent permitted by retailers” (internal quotation marks omitted)). This factor favors not viewing the vouchers as coupons. In sum, factors one and two support finding that the Massage Envy vouchers are coupons. And although the third factor supports the opposite conclusion, no single Online DVD factor is dispositive. See EasySaver, 906 F.3d at 756 n.7 (describing that while one factor weighed in favor of the district court’s determination about CAFA’s applicability to the settlement, the remaining factors did not). Here, the relatively narrow range of products offered (Factor Two), combined with the vouchers failing to allow most class members to buy massage services—MEF’s flagship offering—without spending their own money (Factor One), 22 MCKINNEY-DROBNIS V. ORESHACK suggests that these vouchers should be viewed in law as coupons. Although flexible, the vouchers do ultimately expire, and there is no evidence that a secondary market for Massage Envy vouchers exists. See Redman v. RadioShack Corp., 768 F.3d 622, 628 (7th Cir. 2014) (“[T]he secondary market in coupons is bound to be thin.”). Upon de novo review of the vouchers under the Online DVD framework, we hold that they are coupons and, consequently, are subject to CAFA’s requirements for coupon settlements. We therefore vacate the district court’s approval of the attorneys’ fee award and remand for the district court to use the value of the redeemed vouchers in awarding fees, as required by 28 U.S.C. § 1712(a). III We next address Oreshack’s contention that, independent of CAFA’s applicability to the fee award, the district court erred by approving the settlement as “fair, reasonable, and adequate” under Rule 23(e). As in Roes, 1– 2 v. SFBSC Management, LLC, 944 F.3d 1035 (9th Cir. 2019), “the main thrust of Objector[’s] argument on appeal is that the district court abused its discretion in approving a class action settlement that does not provide enough benefit to class members and contains indicia of collusion.” Id. at 1044. As a preliminary matter, determining that the vouchers are coupons under CAFA and vacating the fee award does not necessarily require invalidating the entire settlement approval order. See Bluetooth, 654 F.3d at 945. But when an objector brings a challenge to the settlement agreement under Rule 23(e), we must analyze the entire agreement for fairness, looking in particular at the Bluetooth warning signs when, as here, we are reviewing a pre-certification settlement agreement. See Chambers, 980 F.3d at 669; see MCKINNEY-DROBNIS V. ORESHACK 23 also In re Sw. Airlines Voucher Litig., 799 F.3d 701, 706 (7th Cir. 2015) (finding that CAFA applies to the fee award, and then examining the entire settlement agreement for fairness using the Ninth Circuit’s Bluetooth signs). Oreshack makes two independent arguments to support his request to vacate the settlement approval. First, he contends that the district court erred by valuing the vouchers for purposes of attorneys’ fees at $10 million. Second, he contends that the court erred in approving a settlement that exhibits preferential treatment to class counsel under Bluetooth. We need not consider Oreshack’s first argument, given our holding in Part II and instructions on remand for the district court to recalculate the fee award using the value of the redeemed vouchers. As for Oreshack’s second argument, we hold that the court abused its discretion by failing to adequately investigate and “substantively grapple with some of the potentially problematic aspects of the relationship between attorneys’ fees and the benefit to the class.” Roes, 944 F.3d at 1051 (cleaned up). Because the errors made by the court impact the fairness of the entire settlement under Rule 23(e), and not just attorneys’ fees, we vacate the approval and remand for the court to analyze more deeply whether the settlement should be approved. To that end, the court may employ whatever procedures it considers helpful to its more rigorous analysis. A Before approving a class settlement under Rule 23(e), district courts must scrutinize the settlement and ensure that it is “fair, reasonable, and adequate.” Fed. R. Civ. P. 23(e). These requirements are in place “[b]ecause of the unique due process concerns relating to absent class members and the inherent risk of collusion between class counsel and defense counsel.” Roes, 944 F.3d at 1048. 24 MCKINNEY-DROBNIS V. ORESHACK “We review a district court’s approval of a class action settlement for clear abuse of discretion.” Bluetooth, 654 F.3d at 940. “When the issue presented is the substantive fairness of the settlement, we must refrain from ‘substitut[ing] our notions of fairness for those of the district judge.’” Campbell v. Facebook, Inc., 951 F.3d 1106, 1121 (9th Cir. 2020) (alteration in original) (quoting Bluetooth, 654 F.3d at 950). Notwithstanding our limited appellate review of substantive fairness, “we hold district courts to a higher procedural standard when making that determination of substantive fairness: ‘To survive appellate review, the district court must show it has explored comprehensively all factors, and must give a reasoned response to all non- frivolous objections.’” Allen v. Bedolla, 787 F.3d 1218, 1223–24 (9th Cir. 2015) (emphasis in original) (citation omitted). This is because, “[p]rior to formal class certification, there is an even greater potential for a breach of fiduciary duty owed the class during settlement.” Bluetooth, 654 F.3d at 946. District courts must ensure that class counsel do not “collude with the defendant to strike a quick settlement without devoting substantial resources to the case.” Briseño v. Henderson, 998 F.3d 1014, 1024 (9th Cir. 2021). Accordingly, when a settlement precedes class certification, as it did here, the district court must apply “an even higher level of scrutiny.” Roes, 944 F.3d at 1049 (quoting Bluetooth, 654 F.3d at 946). In 2018, Congress amended Rule 23(e)(2) to provide specific factors for a district court to consider in determining whether a settlement is “fair, reasonable, and adequate.” Fed. R. Civ. P. 23(e)(2). Those factors include whether: MCKINNEY-DROBNIS V. ORESHACK 25 (A) the class representatives and class counsel have adequately represented the class; (B) the proposal was negotiated at arm’s length; (C) the relief provided for the class is adequate, taking into account: (i) the costs, risks, and delay of trial and appeal; (ii) the effectiveness of any proposed method of distributing relief to the class, including the method of processing class- member claims; (iii) the terms of any proposed award of attorney’s fees, including timing of payment; and (iv) any agreement required to be identified under Rule 23(e)(3); and (D) the proposal treats class members equitably relative to each other. Id. (emphasis added). Under the plain language of the Rule, courts must scrutinize “the terms of any proposed award of attorney’s fees.” Id. at 23(e)(2)(C)(iii). We have interpreted the amended Rule 23(e)(2) as imposing an obligation on district courts to “examine whether the attorneys’ fees arrangement shortchanges the class. In other words, the new Rule 23(e) makes clear that courts must balance the ‘proposed award of attorney’s fees’ vis-à-vis the ‘relief 26 MCKINNEY-DROBNIS V. ORESHACK provided for the class’ in determining whether the settlement is ‘adequate’ for class members.” Henderson, 998 F.3d at 1024. B It is well established that class settlements present the unavoidable risk that class counsel might not have adequately represented the interests of absent class members, and it is equally well established that this concern is salient in the pre-certification settlement context. In Bluetooth, we recognized that class counsel’s self-interest could lead counsel to negotiate a disproportionate share of settlement relief for itself compared to the relief obtained by absent class members. 654 F.3d at 945–46. Given these due process concerns, we must review “a pre-certification settlement approval not only for whether the district court has explored comprehensively all factors, given a reasoned response to all non-frivolous objections, and adequately developed the record to support its final approval decision, but also for whether the district court has looked for and scrutinized any subtle signs that class counsel have allowed pursuit of their own self-interests to infect the negotiations.” Roes, 944 F.3d at 1043 (cleaned up). We have identified three such “subtle signs,” which we call the Bluetooth factors: (1) “when counsel receive[s] a disproportionate distribution of the settlement; (2) when the parties negotiate a clear-sailing arrangement, under which the defendant agrees not to challenge a request for an agreed- upon attorney’s fee; and (3) when the agreement contains a kicker or reverter clause that returns unawarded fees to the defendant, rather than the class.” Henderson, 998 F.3d at 1023 (cleaned up). District courts must apply the Bluetooth factors in examining pre-certification settlements “to smoke out potential collusion.” Id. MCKINNEY-DROBNIS V. ORESHACK 27 If we conclude that the district court did not adequately consider the Bluetooth factors, and therefore did not adequately consider signs that the parties had negotiated an unreasonable amount of attorneys’ fees in assessing settlement fairness in the first instance, then “we must vacate and remand the Approval Order [in addition to the attorneys’ fee award], so that the court may appropriately factor this into its Rule 23(e) analysis.” Bluetooth, 654 F.3d at 946. In Roes, we applied Bluetooth and considered the settlement relief provided to a putative class of exotic dancers who worked at adult-entertainment clubs managed by the defendant. 944 F.3d at 1039. The settlement provided $2 million in cash, distributed as follows: class counsel received $950,000 in fees, the class received $864,115, and the rest went to administrative costs and incentive payments. Id. at 1043. The settlement also provided non-cash relief in the form of injunctive relief and “dance fee payment vouchers.” Id. For purposes of attorneys’ fees, the district court valued both forms of relief at $1 million each, with the objectors challenging both valuations on appeal. Id. at 1051–52. As relevant here, the district court valued the dance fee payment vouchers at their full face value of $1 million instead of $370,000—the value of the dance fee payments that the class claimed but had not yet redeemed. Id. at 1042. We held that the district court in Roes abused its discretion by failing to apply “‘an even higher level of scrutiny’ for evidence of collusion or other conflicts of interest than is ordinarily required under Rule 23(e).” Id. at 1043, 1060 (quoting Bluetooth, 654 F.3d at 936). The Roes settlement agreement included “subtle signs of implicit collusion,” including a clear-sailing agreement, a disproportionate cash distribution for attorneys’ fees, 28 MCKINNEY-DROBNIS V. ORESHACK disproportionate incentive payments to the named plaintiffs, and reversionary clauses that would return unclaimed funds to the defendants. Id. at 1049–50. We noted that “the district court did not substantively investigate or address” some of the objectors’ concerns, and it did not “explain why the [vouchers] should [] be valued at its $1 million maximum” even though only a portion of that maximum had been claimed. Id. at 1052. Importantly, we stated that “[r]egardless of whether the dance fee payment vouchers are officially ‘coupons’ within the meaning of [CAFA], the district court should have recognized that some of the same concerns applicable to coupon settlements also apply here and warranted closer scrutiny of the [vouchers as settlement relief].” Id. Instead, the court dismissed the voucher objection by asserting that the dance fee payment vouchers provide “a tangible benefit” that was “not the ordinary illusory coupon payment.” Id. Because the court failed to “substantively grapple” with whether the Bluetooth warning signs created an unfair settlement, id. at 1051, we vacated the settlement approval and remanded for “a more searching inquiry.” Id. at 1050 (quoting Allen, 787 F.3d at 1224). C MEF contends that Oreshack asks us to ignore all of the Rule 23(e) factors except for the part of Rule 23(e)(2)(C) that requires courts to balance the adequacy of class relief in light of “the terms of any proposed award of attorney’s fees.” It is true that Rule 23(e)(2)(C) directs courts to evaluate settlement fairness in light of, among other things, the “costs, risks, and delay of trial and appeal.” Fed. R. Civ. P. 23(e)(2)(C)(i). And here, as the district court acknowledged, the class would have faced obstacles if the case were litigated further. So, according to MEF, as in Campbell v. Facebook, Inc., Oreshack is challenging “only a subset of MCKINNEY-DROBNIS V. ORESHACK 29 the considerations that were relevant to the district court’s holistic assessment of the settlement’s fairness.” See 951 F.3d 1106, 1122 (9th Cir. 2020) (affirming the settlement approval where the objector challenged the district court’s award of “worthless injunctive relief” (citation omitted)). We disagree. First, Campbell is readily distinguishable. For one thing, although we noted that the objector in Campbell was challenging only a subset of relevant considerations, it was in the context of the objector’s failure to weigh the value of the injunctive relief against the strength of the claims that the class members would have given up. Id. at 1122–24. We then separately concluded that, unlike here, the district court had adequately considered the Bluetooth factors. Id. at 1125. Finally, Campbell is factually distinguishable because it involved a post-certification Rule 23(b)(2) settlement for solely injunctive relief. Id. We concluded that the district court did not abuse its discretion in finding that Bluetooth’s second and third warning signs— clear-sailing agreements and reverter provisions—were inapplicable in that context. Id. Second, MEF’s focus on the strength of Plaintiffs’ case does not persuade us because even a recognition that the substantive claims present a weak case cannot cure a district court’s failure to apply the requisite heightened scrutiny to a pre-certification settlement agreement. Before Rule 23(e)’s 2018 amendment provided factors for courts to consider in assessing settlement approval, we “filled in the gaps,” Henderson, 998 F.3d at 1023, by instructing courts to weigh the following factors: (1) the strength of the plaintiffs’ case; (2) the risk, expense, complexity, and likely duration of further litigation; (3) the risk of 30 MCKINNEY-DROBNIS V. ORESHACK maintaining class action status throughout the trial; (4) the amount offered in settlement; (5) the extent of discovery completed and the stage of the proceedings; (6) the experience and views of counsel; (7) the presence of a governmental participant; and (8) the reaction of the class members to the proposed settlement. Churchill Vill., LLC v. Gen. Elec., 361 F.3d 566, 575 (9th Cir. 2004) (the “Churchill factors”) (emphasis added). 4 But we have also held that adequately considering the Churchill factors is insufficient if the district court failed to adequately investigate or address the Bluetooth factors. Indeed, even where several Churchill factors militate towards settlement approval, if “a settlement agreement is negotiated prior to formal class certification, consideration of these eight Churchill factors alone is not enough to survive appellate review.” Bluetooth, 654 F.3d at 946 (emphasis in original). Accordingly, we conclude that Oreshack does not impermissibly challenge only a subset of the district court’s fairness assessment. D Oreshack contends that the district court erred by failing to apply enhanced scrutiny to a pre-certification settlement. We agree that the court did not apply the requisite scrutiny and thereby abused its discretion in failing to “investigate or 4 The amended Rule 23(e) did not “displace” this court’s previous articulation of the relevant factors, and it is still appropriate for district courts to consider these factors in their holistic assessment of settlement fairness. See Fed. R. Civ. P. 23 advisory committee’s note to the 2018 amendment. MCKINNEY-DROBNIS V. ORESHACK 31 adequately address” the economic reality of the settlement relief and the Bluetooth warning signs. See Roes, 944 F.3d at 1049; Bluetooth, 654 F.3d at 947–48. Oreshack contends that all three Bluetooth warning signs are present in the settlement. We address each Bluetooth factor in turn. The first factor is whether class counsel has received a “disproportionate distribution of the settlement.” Bluetooth, 654 F.3d at 947 (quoting Hanlon v. Chrysler Corp., 150 F.3d 1011, 1021 (9th Cir. 1998)). Because we have determined that the vouchers are coupons and have directed the district court to reassess the fee award, on remand it will be important for the district court to reconsider whether class counsel received a “disproportionate distribution of the settlement” in light of the adjusted fee award. The parties do not dispute that the second and third Bluetooth factors—the clear-sailing and reverter provisions—are present. We agree with Oreshack that the district court did not adequately investigate or address the implications of those provisions. In Bluetooth and subsequent cases, we have considered how the Bluetooth factors can operate on their own and in tandem to provide warning signs of collusion. When a clear-sailing provision is paired with a reverter, the terms together increase the risk that class counsel will unreasonably raise the amount of requested fees, and the class members will have less incentive to push back because the recovery of any unawarded fees will inure to the benefit of the defendants, not the class members. In Roes, we noted that clear-sailing arrangements are not prohibited, but that they are “disfavored” because they are “important warning signs of collusion.” Id. at 1050–51 (citations omitted). This is because “[a] clear sailing provision signals the potential that 32 MCKINNEY-DROBNIS V. ORESHACK a defendant agreed to pay class counsel excessive fees in exchange for counsel accepting a lower amount for the class members.” Henderson, 998 F.3d at 1027 (citing Bluetooth, 654 F.3d at 949). Although the presence of a clear-sailing provision is not a “death knell,” the district court has a duty to scrutinize the agreement for signs that the fees requested by counsel are unreasonably high. Kim v. Allison, 8 F.4th 1170, 1180 (9th Cir. 2021). Similarly, we have identified “reverter” or “kicker” provisions as red flags because if the defendant “is content to pay [millions of dollars] to class counsel but the court finds the full amount unreasonable, there is no plausible reason why the class should not benefit from the spillover of excessive fees.” Id. So, unless the court makes a finding in this case that the two provisions together promote the best interest of the class, and not just class counsel, it is “less likely that the settlement can be approved.” Bluetooth, 654 F.3d at 949. The district court assessed these warning signs by first stating that “we do have what might be called a clear sailing provision here.” Without further analysis of the clear-sailing provision—and no party disputes that the term in question is a clear-sailing provision—the court ultimately found that “since there aren’t some of the other red flags,” it wasn’t “prepared to find that there was collusion here.” Thus, the court’s only reference to the clear-sailing provision was to say that because other red flags are not present, the clear- sailing agreement is not dispositive. This is a questionable conclusion given that the agreement also contains a reverter provision. Moreover, although it is true that such clear- sailing arrangements are not per se prohibited, it is also true that “[t]he very existence of a clear sailing provision increases the likelihood that class counsel will have MCKINNEY-DROBNIS V. ORESHACK 33 bargained away something of value to the class.” Bluetooth, 654 F.3d at 948 (emphasis added) (citation omitted). The district court next considered the reverter provision and stated: “The settlement provides that the fees sought but not awarded would revert to the defendant rather than remain in the settlement. Well, they won’t be paid if they’re not paid.” The parties dispute the meaning of this statement— “[w]ell they won’t be paid if they’re not paid”—but we conclude that whether the “they” in question refers to class counsel or to counsel’s fees, the court did not adequately scrutinize the provisions for evidence of whether class counsel’s self-interest had “infect[ed] the negotiations.” Id. at 947. Here, the reverter operated to return almost $602,000 to MEF rather than to the class, even though MEF had shown itself willing to pay that amount in connection with the settlement. The court acknowledged only that class counsel would not be paid an unreasonable fee if the court chose to reduce that fee, not that the fee reduction would itself benefit MEF rather than the class. When a settlement provides non-cash relief and a reverter provision, a district court must be on the alert for an attorneys’ fee award that is artificially inflated in relation to the relief provided to the class. See Roes, 944 F.3d at 1053– 54. The more undesirable or inflexible a voucher is in comparison to cash or to a gift card, the greater the risk that the settlement value may be overinflated. This is because the risk that such settlement relief will be artificially inflated “is even more grave when the value of unused coupons will revert back to defendants.” Id. In other words, if the likely redemption rate is low—which is to be expected in a consumer class action providing non-cash relief—then MEF and class counsel can inflate the perceived settlement value while knowing that MEF is unlikely to pay more than a 34 MCKINNEY-DROBNIS V. ORESHACK fraction of that amount. Id. at 1054 (“Unchecked, such reversions would allow defendants to create a larger coupon pool than they know will be claimed or used, just to inflate the value of the settlement and the resulting attorneys’ fees, because they know that they will not be on the hook for the full coupon pool.”). In the pre-certification context, the district court must do more than acknowledge that warning-sign provisions exist and then conclude that they are not dispositive without further apparent scrutiny. This is especially true when the court does not “substantively grapple,” Roes, 944 F.3d at 1051, with the ways in which the red-flag provisions and specific voucher characteristics work together to present “multiple indicia of possible implicit collusion,” Bluetooth, 654 F.3d at 947. Here, the court asserted that it is “not unusual for there to be a fund in which not all the funds are used,” and “one can always argue” that unawarded fees should have gone to the class. Even if that is so, the court is not free to discount settlement terms that we have held are evidence of potential collusion without adequate investigation and analysis. See id. Accordingly, because the district court did not conduct the required heightened inquiry, we hold that the court abused its discretion in granting approval of the settlement. See Roes, 944 F.3d at 1060. On remand, we do not restrict the scope of the court’s inquiry regarding whether the settlement should be approved. It might be that in the end, after adjusting the attorneys’ fees using the voucher redemption rate and applying the heightened scrutiny that Bluetooth requires, the court will conclude that the settlement agreement is fair. But because we hold the court to a higher procedural standard, the court must “provide the necessary explanations” in making that finding. Bluetooth, MCKINNEY-DROBNIS V. ORESHACK 35 654 F.3d at 945 (citation omitted). We remand for that purpose. IV Under Rule 23(e), a federal court may approve a class action settlement only if it finds the agreement is “fair, reasonable, and adequate.” Fed. R. Civ. P. 23(e). For the foregoing reasons, we vacate and remand the district court’s approval of the settlement and its fee award. We instruct the court to use the value of the redeemed vouchers as required by CAFA and to analyze the pre-certification settlement agreement with heightened scrutiny. In so holding, we express no opinion on the ultimate fairness of the settlement that the parties have negotiated—a conclusion properly within the purview of the district court. Id. at 949–50. VACATED AND REMANDED. Each party shall bear its own costs on appeal. MILLER, Circuit Judge, concurring: I join the court’s opinion in full. I write separately to note my disagreement with our circuit’s approach to determining when vouchers are “coupons” under the Class Action Fairness Act of 2005 (CAFA), Pub. L. No. 109-2, 119 Stat. 4. District courts must review proposed class-action settlements to determine whether they are “fair, reasonable, and adequate.” Fed. R. Civ. P. 23(e)(2). An inherent danger in class-action settlements is that the defendant (who cares only about the total amount of the settlement, not how it is distributed) will agree to a settlement in which most of the 36 MCKINNEY-DROBNIS V. ORESHACK recovery flows to class counsel, with only modest benefits to the class members (none of whom individually has enough at stake to have an incentive to object). That danger is particularly acute when the benefits to the class come in the form of coupons for the defendant’s products. If the court were to assess the reasonableness of the settlement based on the nominal value of the coupons—many of which the class members might never use—then the apparent value of the settlement fund would be artificially inflated and would exceed the actual benefit to the class. See Roes, 1–2 v. SFBSC Mgmt., LLC, 944 F.3d 1035, 1053–54 (9th Cir. 2019). To avoid that result, CAFA provides that “[i]f a proposed settlement in a class action provides for a recovery of coupons to a class member, the portion of any attorney’s fee award to class counsel that is attributable to the award of the coupons shall be based on the value to class members of the coupons that are redeemed.” 28 U.S.C. § 1712(a). By directing the court to count only “the value to class members of the coupons that are redeemed,” id., the statute “ensures that class counsel benefit only from coupons that provide actual relief to the class, lessening the incentive to seek an award of coupons that class members have little interest in using—either because they might not want to conduct more business with defendants, or because the coupons are too small to make it worth their while,” In re EasySaver Rewards Litig., 906 F.3d 747, 755 (9th Cir. 2018). But although determining whether a settlement involves coupons is central to calculating attorney’s fees correctly, CAFA does not define the term “coupon.” Normally, when a statute does not define a term, we look to its ordinary meaning. Mississippi Band of Choctaw Indians v. Holyfield, 490 U.S. 30, 47 (1989). As Judge Friedland has observed, MCKINNEY-DROBNIS V. ORESHACK 37 the ordinary meaning of “coupon” encompasses “any type of award that is not cash or a product itself, but that class members can redeem to obtain products or services or to help make future purchases.” Hendricks v. Ference, 754 F. App’x 510, 514 (9th Cir. 2018) (Friedland, J., concurring in part and dissenting in part); see 3 Oxford English Dictionary 1050–51 (2d ed. 1989) (defining “coupon” as “a form, ticket, part of a printed advertisement, etc., entitling the holder to a gift or discount”); Webster’s Third New International Dictionary 522 (2002) (defining “coupon” as a “form, slip, or section of a paper resembling a bond coupon in that it may be surrendered in order to obtain some article, service, or accommodation” or a “form or check indicating a credit against future purchases or expenditures”). Under that definition, the vouchers in this case, which have no cash value but simply grant class members an amount ranging from $36.28 to $180.68 off Massage Envy products or services, are plainly coupons—so plainly that class representatives’ counsel repeatedly (albeit unintentionally) referred to them as “coupons” during oral argument. It would be best if we could resolve this case by stating the obvious: A voucher is a coupon, so class counsel’s attorney’s fees must be calculated based on the value of any Massage Envy vouchers that are redeemed. See Redman v. RadioShack Corp., 768 F.3d 622, 636 (7th Cir. 2014) (concluding that the term “coupon” is “interchangeable with ‘voucher’”). Unfortunately, our precedent commands otherwise. In In re Online DVD-Rental Antitrust Litigation, 779 F.3d 934 (9th Cir. 2015), we prescribed a three-factor test for determining whether an award constitutes a coupon settlement: “(1) whether class members have ‘to hand over more of their own money before they can take advantage of’ 38 MCKINNEY-DROBNIS V. ORESHACK a credit, (2) whether the credit is valid only ‘for select products or services,’ and (3) how much flexibility the credit provides, including whether it expires or is freely transferrable.” EasySaver, 906 F.3d at 755 (quoting Online DVD, 779 F.3d at 951). That test has no basis in the statutory text. And as Judge Friedland has observed, it introduces “needless complication and confusion” to the evaluation of class-action settlements. Hendricks, 754 F. App’x at 516 (Friedland, J., concurring in part and dissenting in part). This case is a good example. We hold that one of the three factors is “somewhat inconclusive” but “on balance” points one way; another “appear[s]” to point the same direction; and a third points to the opposite conclusion. Just how to balance the factors against each other is unclear because they are not readily commensurable. Here, we conclude that the vouchers are coupons. If one of the three factors were slightly different, would the conclusion be different? Further litigation will be required before anyone can know for sure. None of this is a criticism of today’s decision; the court does as well as anyone could in applying the Online DVD test. The problem is with the test itself. In an appropriate case, we should reconsider Online DVD en banc.