Irma Allen v. Ollies Bargain Outlet Inc

37 F.4th 890

• Court: Court of Appeals for the Third Circuit
• Decided: June 24, 2022
• Docket: 21-2121
• Status: Published

Opinion

PRECEDENTIAL

UNITED STATES COURT OF APPEALS FOR THE THIRD CIRCUIT ____________

No. 21-2121 ____________

IRMA ALLEN; BARTLEY MICHAEL MULLEN, Jr., Individually and on behalf of all others similarly situated

v.

OLLIE’S BARGAIN OUTLET, INC., Appellant ____________

On Appeal from the United States District Court for the Western District of Pennsylvania (D.C. Civil No. 2-19-cv-00281) District Judge: Honorable William S. Stickman, IV ____________

Argued: March 15, 2022

Before: JORDAN, KRAUSE, and PORTER, Circuit Judges.

(Filed: June 24, 2022) ____________ Richard L. Etter [ARGUED] Ogletree Deakins One PPG Place Suite 1900 Pittsburgh, PA 15222

David L. Schenberg Ogletree Deakins 7700 Bonhomme Avenue Suite 650 St. Louis, MO 63105

Counsel for Appellant

R. Bruce Carlson Carlson Brown 222 Broad Street Sewickley, PA 15143

Gary F. Lynch Elizabeth Pollock-Avery Kelly K. Iverson Jamisen A. Etzel [ARGUED] Nicholas Colella Lynch Carpenter 1133 Penn Avenue 5th Floor Pittsburgh, PA 15222

Counsel for Appellees

2 ____________

OPINION OF THE COURT ____________

PORTER, Circuit Judge.

Irma Allen and Bartley Mullen are disabled and need wheelchairs to move about. Hoping to find “Good Stuff Cheap,” they went shopping at two different bargain stores owned by Ollie’s Bargain Outlet, Inc. (“Ollie’s”). But once inside Ollie’s, they encountered an obstacle course: pillars, clothing racks, and boxes blocked their way. Dissatisfied with their shopping experiences, they filed a putative class action against Ollie’s under Title III of the Americans with Disabili- ties Act (“ADA”). They seek permission to sue on behalf of every similarly disabled individual who shops at any Ollie’s store in the United States and has or will encounter interior access barriers. The District Court certified the proposed class. We will vacate and remand. The District Court abused its dis- cretion by certifying an overly broad class based on inadequate evidence of numerosity and commonality.

I

A

Ollie’s owns and operates over four hundred retail stores across twenty-nine states.1 Allen and Mullen visited two different Ollie’s stores in Monaca and New Castle,

1 Ollie’s Bargain Outlet Holdings, Inc., 2021 Annual Report (Form 10-K) at 1 (“We have grown to 431 stores in 29 states as of January 29, 2022.”).

3 Pennsylvania. There, they encountered obstacles blocking their path of travel, including inventory on the floor, clothing racks placed too close together, boxes, pallets, and structural pillars. Pictures taken later at these stores show aisles similarly nar- rowed by inventory carts, pallets, columns, boxes, or goods on the floor. Suspecting a pattern, Allen and Mullen’s lawyers hired investigators to take photographs and measure aisle width at several Ollie’s stores in Pennsylvania. After this pre- liminary investigation, Allen and Mullen sued Ollie’s under Title III of the ADA.

B

Title III of the ADA prohibits retailers like Ollie’s from discriminating “on the basis of disability in the full and equal enjoyment of the goods, services, facilities, privileges, advantages, or accommodations” they offer to the public. 42 U.S.C. § 12182(a). This general prohibition has several spe- cific definitions that extend disability discrimination beyond disparate treatment or invidious discrimination. Plaintiffs fo- cus their complaint and argument on three specific definitions of Title III discrimination. We discuss these for background.

First, Title III discrimination includes “a failure to make reasonable modifications in policies, practices, or procedures, when such modifications are necessary to afford” goods, ser- vices, and the like to “individuals with disabilities.” Id. § 12182(b)(2)(A)(ii). “To comply with this command, an indi- vidualized inquiry must be made to determine whether a spe- cific modification for a particular person’s disability would be reasonable under the circumstances as well as necessary for that person . . . .” PGA Tour, Inc. v. Martin, 532 U.S. 661, 688 (2001).

4 Second, Title III discrimination includes “a failure to remove architectural barriers . . . in existing facilities, . . . where such removal is readily achievable.” 42 U.S.C. § 12182(b)(2)(A)(iv). The Department of Justice gives the term “architectural barriers” a broad scope. For example, shelves, tables, chairs, vending machines, display racks, and furniture are treated as “architectural.” 28 C.F.R. § 36.304(b)(3), (4). Architectural barriers must be removed only when “readily achievable,” a standard that “means easily accomplishable and able to be carried out without much diffi- culty or expense.” 42 U.S.C. § 12181(9).

Third, facilities built or altered after the ADA’s effec- tive dates must be “readily accessible to and usable by” the dis- abled. 42 U.S.C. § 12183(a); 28 C.F.R. §§ 36.401(a)(1), 402(a)(1). To be readily accessible, a facility must comply with the standards for accessible design. 28 C.F.R. § 36.406. Under section 403.5.1 of the most recent 2010 standards, aisles must generally be at least thirty-six inches wide, but can measure as little as thirty-two inches wide for short distances. 36 C.F.R. pt. 1191, app. D. Department of Justice rules require facilities to maintain accessible aisles “in operable working condition.” 28 C.F.R. § 36.211(a).

Plaintiffs’ “core contention” is that “Ollie’s deliberately directs the placement of merchandise within aisles,” causing a corporate-wide failure to maintain accessible aisles. Appellees’ Br. 28. Under plaintiffs’ theory, retail stores fail to maintain accessible aisles “in operable working condition” if they inten- tionally and recurringly block them with movable objects, a position supported by Ninth Circuit precedent. See Chapman v. Pier 1 Imports (U.S.) Inc., 779 F.3d 1001, 1009 (9th Cir. 2015) (retail store violated ADA when it had a pattern of obstructing aisles with objects like “step ladders”). Plaintiffs

5 claim that Ollie’s failure to modify its corporate policies to pre- vent this alleged merchandising practice is discriminatory, and they also suggest that some or all merchandising goods count as “architectural” barriers that must be removed.

C

After completing targeted discovery, plaintiffs moved to certify the following class under Federal Rule of Civil Pro- cedure 23(b)(2):

All persons with qualified mobility disabilities who have attempted, or will attempt, to access the interior of any store owned or operated by [Ollie’s] within the United States and have, or will have, experienced access barriers in interior paths of travel.

App. 171. Before proceeding as a class under Rule 23(b)(2), plaintiffs had to satisfy Federal Rule of Civil Procedure 23(a). Under Rule 23(a), they had to “demonstrate, first, that ‘(1) the class is so numerous that joinder of all members is impractica- ble; ‘(2) there are questions of law or fact common to the class; ‘(3) the claims or defenses of the representative parties are typ- ical of the claims or defenses of the class; and ‘(4) the repre- sentative parties will fairly and adequately protect the interests of the class.’” Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 345 (2011) (quoting Fed. R. Civ. P. 23(a)). To satisfy Rule 23(b)(2), plaintiffs then had to show that Ollie’s “has acted or refused to act on grounds that apply generally to the class, so that final injunctive relief or corresponding declaratory relief is appropriate respecting the class as a whole.” Id. at 345–46 (quoting Fed. R. Civ. P. 23(b)(2)).

6 Before the District Court, Plaintiffs argued that joinder of class members was impracticable given the size of the class. They introduced three strands of evidence to support this assertion. First, data from the U.S Census Bureau’s 2018 American Community Survey, estimating the number of peo- ple with ambulatory disabilities—meaning serious difficulty walking or climbing stairs—for each zip code with an Ollie’s store. Second, twelve emails received by Ollie’s customer ser- vice over three years from or on behalf of patrons that use wheelchairs or have a mobility disability. Third, a declaration stating that over seven days, sixteen persons using wheelchairs or scooters were recorded by video at the two Ollie’s locations where Allen and Mullen shopped.

Plaintiffs at first argued there were common questions based on Ollie’s alleged failure to adopt ADA-specific stand- ard operating procedures and employee training practices. In their reply brief, plaintiffs urged a narrower commonality argument, one they now press on appeal. They asserted that Ollie’s “employees” have a common “practice” of “placing merchandise displays and stock in locations that block or limit accessibility,” and they attributed this alleged practice to Ollie’s corporate “merchandise stocking and display prac- tices.” App. 901–02. To support this commonality argument, plaintiffs cited allegations in their complaint, Allen’s and Mullen’s depositions, and photographs of Pennsylvania stores showing a “pattern and practice of path of travel obstructions.” App. 901 n.8, 902 n.9.

D

The District Court certified the proposed class. The District Court agreed with plaintiffs that joinder of all class members would be impracticable. Allen v. Ollie’s Bargain

7 Outlet, Inc., No. 2:19-CV-281, 2021 WL 1152981, at *6 (W.D. Pa. Mar. 26, 2021). Adding Allen and Mullen, the twelve cus- tomer emails, and the sixteen individuals observed in two stores over seven days, the District Court concluded that plain- tiffs “have concretely shown that thirty people with potential mobility disabilities are customers of Ollie’s stores.” Id. In the District Court’s judgment, the circumstantial evidence of thirty potentially disabled patrons, together with the community sur- vey estimates, was enough. Id. As the District Court put it, “[t]he statistical evidence presented already indicates that there is a good chance that the proposed class is numerous, and any speculation accompanying the statistical data alone is over- come by the addition of the concrete, case-specific evidence of written complaints and video footage.” Id. at *6 (citation omit- ted). Ollie’s objected to the use of the customer complaints as inadmissible hearsay, but the District Court overruled the objection, holding that non-expert evidence like the customer complaints need not be admissible to certify a class. Id. at *5 n.5.

The District Court also held the proposed class pre- sented common questions. It relied on a syllogism. First, “Ollie’s policies are uniform and company-wide.” Id. at *7. Second, “[i]f Ollie’s policies and procedures do, in fact, cause access barriers to unlawfully restrict individuals with disabili- ties from obtaining their desired goods, then proposed mem- bers who endured violations have suffered the same injury, the resolution of which will resolve a central issue in one fell stroke.” Id. “As a result,” the District Court held, “Plaintiffs have satisfied their burden of showing by a preponderance of the evidence that there are questions of law or fact common to the proposed class.” Id. After finding the remaining require- ments of Rule 23(a) were met, the District Court held that the

8 proposed class satisfied Rule 23(b)(2) because “[a]n injunction requiring the removal of the existing access barriers, and the modification of Ollie’s policies to prevent the use of access barriers restricting disabled individuals’ use and enjoyment of Ollie’s goods would provide appropriate relief to the proposed class.” Id. at *8.

This appeal followed.

II

The District Court had subject-matter jurisdiction under 28 U.S.C. § 1331. We have appellate jurisdiction under Rule 23(f) and 28 U.S.C. § 1292(e). Mielo v. Steak ’n Shake Operations, Inc., 897 F.3d 467, 473–74 (3d Cir. 2018). If the case proceeds to summary judgment or trial, the result may be different, but Allen and Mullen have adequately alleged Article III standing at this stage. Id. at 478–82.

We review a class certification order for abuse of dis- cretion, which occurs if the trial court’s decision rests on a clearly erroneous finding of fact, an errant conclusion of law, or an improper application of law to fact. Marcus v. BMW of N. Am., LLC, 687 F.3d 583, 590 (3d Cir. 2012). We review questions of law, including whether the trial court applied the correct legal standard, de novo. Steak ’n Shake, 897 F.3d at 474.

III

Under Rule 23, the proposed class must be “so numer- ous that joinder of all members is impracticable.” Fed. R. Civ. P. 23(a)(1). This “rule prevents putative class representatives

9 and their counsel, when joinder can be easily accomplished, from unnecessarily depriving members of a small class of their right to a day in court to adjudicate their own claims.” Marcus, 687 F.3d at 594–95. As with every Rule 23 requirement, plain- tiffs must show the class is numerous enough by a preponder- ance of the evidence. Steak ’n Shake, 897 F.3d at 483–84. We presume joinder is impracticable when the potential number of class members exceeds forty. Id. at 486. This is a guidepost: showing the number of class members exceeds forty is neither necessary nor always sufficient. Marcus, 687 F.3d at 595. “The text” of Rule 23(a)(1) is “conspicuously devoid of any numer- ical minimum required for class certification.” In re Modafinil Antitrust Litig., 837 F.3d 238, 249 (3d Cir. 2016). But while a class of forty-one does not automatically satisfy Rule 23(a)(1), a putative class that size faces a relaxed burden under our prec- edent. By contrast, the “inquiry into impracticability should be particularly rigorous when the putative class consists of fewer than forty members.” Id. at 250.

In recent opinions, we have given the numerosity re- quirement “real teeth.” Steak ’n Shake, 897 F.3d at 484. When plaintiffs cannot directly identify class members, they “must show sufficient circumstantial evidence specific to the prod- ucts, problems, parties, and geographic areas actually covered by the class definition to allow a district court to make a factual finding. Only then may the court rely on ‘common sense’ to forgo precise calculations and exact numbers.” Marcus, 687 F.3d at 596. And “where a putative class is some subset of a larger pool, the trial court may not infer numerosity from the number in the larger pool alone.” Hayes v. Wal-Mart Stores, Inc., 725 F.3d 349, 358 (3d Cir. 2013). In Steak ’n Shake, for example, census data showing “there are between 14.9 million to 20.9 million persons with mobility disabilities who live in

10 the United States” was not enough to show numerosity under Rule 23(a)(1). 897 F.3d at 486. Applying these principles, we conclude the District Court abused its discretion when it found that plaintiffs had met their numerosity burden.

Plaintiffs argue that the 2018 American Community Survey estimates of persons with mobility disabilities would alone allow us to affirm the District Court’s numerosity find- ing. But these survey estimates prove little. The survey measures anyone who reports serious difficulty walking or climbing steps. Plaintiffs acknowledge that the more relevant number of disabled persons—individuals needing wheel- chairs—is about an order of magnitude lower, and they ask us to extrapolate more accurate regional numbers from different national census estimates. The national census study they cite estimates that persons needing wheelchairs are a small fraction of the population that has severe difficulty walking or climbing stairs. See Mathew W. Brault, U.S. Census Bureau, Americans with Disabilities: 2010, P70-131, Table A-1, 17 (July 2012) (8.3% of the U.S. population fifteen and older has a severe mobility disability, but only 1.5% uses a wheelchair), https://perma.cc/5V96-H5DS. But extrapolating the relevant number across every region would be hazardous speculation. “Trained experts commonly extrapolate from existing data.” Gen. Elec. Co. v. Joiner, 522 U.S. 136, 146 (1997). Generalist Article III judges typically do not.

Regional population statistics like the survey are in any event insufficient. The District Court was right “that a district court’s finding premised almost exclusively on statistical data is not enough to satisfy numerosity—something more is required.” Ollie’s, 2021 WL 1152981, at *6. Steak ’n Shake

11 rejected reliance on statistical data documenting the number of disabled people not because it was national in scope, but because it did not allow us to “determine—rather than specu- late about—the portion of those disabled individuals who have actually patronized a relevant Steak ’n Shake restaurant, let alone the portion who have experienced or will experience an ADA violation at one of those restaurants.” 897 F.3d at 486. The same remains true for regional population statistics. Regional estimates of a disabled population showing proximity to a store may be more probative of disabled customers than national ones, depending on the quality and reliability of the study’s statistical methods and practices, but they alone do not support a finding that a class is numerous. Consider Ollie’s store in Monaca, Pennsylvania. The survey suggests there are about one thousand persons with mobility disabilities living in the same zip code. Plaintiffs would extrapolate that about a tenth of these residents, or one hundred Monaca residents, need wheelchairs to move about. Even if that extrapolation is accurate, however, we would still be left with no basis to determine what portion of those one hundred wheelchair- bound residents of Monaca are customers of Ollie’s, let alone what portion have suffered a common ADA injury. We cannot infer numerosity from this large pool of residents.

The “something more” required by Steak ’n Shake is concrete evidence of class members who have patronized a public accommodation and have suffered or will likely suffer common ADA injuries. We reject plaintiffs’ argument that the community survey estimates alone are enough to carry their burden of proof.

12 2

We next turn to the other two strands of non-statistical evidence the District Court thought set this case apart from Steak ’n Shake. While this evidence is probative, after examin- ing all the evidence, we are still left with head-scratching spec- ulation, insufficient to support a factual finding.

a.

The first strand of non-statistical evidence is plaintiffs’ declaration stating that over seven days, sixteen persons using wheelchairs or scooters were recorded by video at the two Ollie’s locations where Allen and Mullen shopped.2 We agree that this declaration is “probative of the number of potentially disabled individuals visiting Ollie’s stores.” Ollie’s, 2021 WL 1152981, at *5 n.6. But it is not enough to satisfy plaintiffs’ burden of proof on numerosity, even considered alongside the community survey of disabled residents.

For one, the declaration does not allow us to determine what portion of disabled residents shop at Ollie’s. Plaintiffs ask us to extrapolate customer numbers from a limited video sam- ple of two stores over seven days, arguing the video footage suggests more than three hundred wheelchair-using customers

2 Ollie’s does not challenge the admissibility of this declaration, and for good reason. When a class certification “motion relies on facts outside the record,” Federal Rule of Civil Procedure 43(c) allows trial courts to “hear the matter on affidavits.” And the Judicial Code permits declarations instead of affidavits. 28 U.S.C. § 1746. Regardless of whether the Federal Rules of Evidence govern, the declaration was properly before the District Court.

13 shop at Ollie’s every day. But that extrapolation rests on spec- ulation, not a reasonable inference. The video, for starters, does not allow us to determine what portion of those wheelchair- using customers are disabled. To be disabled, the customers would need to have “a physical or mental impairment that sub- stantially limits one or more major life activities,” including “walking.” 42 U.S.C. § 12102(1)(A), (2)(A). At least some wheelchair- or scooter-using customers may not qualify. Cf. Richardson v. Chicago Transit Auth., 926 F.3d 881, 884 (7th Cir. 2019) (holding extreme obesity does not qualify as a phys- ical impairment if it is not the result of a physiological disorder or condition). To be sure, we agree the District Court does not need to “determine as a matter of law that each of the sixteen individuals seen using a wheelchair are mobility disabled un- der the ADA before considering the photographs.” Ollie’s, 2021 WL 1152981, at *5 n.6. In this case, the District Court used common sense to infer that at least some of the customers using wheelchairs are likely disabled. Id. We cannot say that common-sense inference was an abuse of discretion. It is fair to infer that at least some of the wheelchair-using customers are likely disabled under the ADA. Even if we accept the District Court’s conclusion, however, the number of disabled customers observed in the video could range from zero to six- teen. Some evidence buttressing a correlation between the wheelchair-using and ADA-disabled populations would signif- icantly strengthen this evidence.

Still, even assuming all sixteen customers were likely disabled and that none of them were repeat visitors, we have no basis to assume that the rate of wheelchair-using customers observed in the video footage sample is representative of Ollie’s stores. Before we can extrapolate the limited sample across four hundred stores, our precedent requires at least some

14 evidence supporting a factual finding that disabled customers visit Ollie’s “in roughly equal proportions” to the rate observed in the video. Marcus, 687 F.3d at 596. Otherwise, we remain in the realm of speculation, not common-sense inferences. And even if the declaration allowed us to determine the pool of wheelchair-bound Ollie’s customers, the declaration still does not allow us to “determine—rather than speculate about—the portion of those disabled individuals who . . . have experienced or will experience an ADA violation at one of those” stores. Steak ’n Shake, 897 F.3d at 486. The declaration does not sug- gest that the wheelchair-using customers observed in the video suffered an ADA violation in common with the class. At best, the declaration is evidence of the general pool of wheelchair- using Ollie’s customers, not the more relevant subset of wheelchair-bound customers who have suffered common ADA injuries. The District Court appears to have assumed that evidence of injured customers was unnecessary to support numerosity. See Ollie’s, 2021 WL 1152981, at *6 n.7. If so, that was error. Evidence establishing the subset of injured cus- tomers, not just the general pool of wheelchair-using customers who shop at Ollie’s, is necessary to support a finding that a class is likely numerous enough. See Marcus, 687 F.3d at 595. The putative class consists of persons with mobility disabilities who encountered or will encounter inaccessible aisles at an Ollie’s store. There may well be millions of wheelchair-bound Ollie’s customers across all twenty-nine states, but if none of them suffered or will likely suffer similar class injuries, they are not class members and do not support a finding of numer- osity.

b.

The District Court also relied on what it characterized as “the written complaints of twelve individuals complaining,

15 in one way or another, of various barriers adversely affecting the navigation of individuals who are wheelchair-bound.” Ollie’s, 2021 WL 1152981, at *5. Unlike the community sur- vey or the video, at least some of these customer complaints support the existence of putative class members with common ADA injuries. But there are far too few complaints, and not all of them support the District Court’s finding.

At the outset, we note the parties dispute whether the customer complaints are admissible as evidence. Ollie’s argues that the Federal Rules of Evidence apply during the class cer- tification stage, and that the customer complaints are inadmis- sible hearsay. Plaintiffs respond that “fact” testimony—as opposed to expert opinion—need not be admissible to support class certification. The District Court agreed with plaintiffs, holding the Federal Rules of Evidence are inapplicable to non- expert evidence used to certify a class. Ollie’s, 2021 WL 1152981, at *5 n.5. We decline to decide this question. For even assuming—and it is only an assumption—that the Federal Rules of Evidence do not govern the admissibility of the cus- tomer complaints, the record still would not establish numer- osity.3

To begin, at least one of the twelve customer complaints does not support membership in the putative class. The rele- vant email says:

Please pass this on to the management at the Columbus, GA store. My husband and I recently visited this store for the first time, and we were very impressed. My husband is a paraplegic, and

3 Ollie’s has preserved this argument and may raise it again on remand.

16 uses a wheelchair while shopping. There were very few places he could not get into, and every employee he encountered asked if he needed help. The aisles were clear, and the merchandise was—for the most part—easy for him to reach. When we asked for help, it was given cheerfully and quickly. We enjoyed the experience, and plan to become regular customers. The employ- ees of this store went above and beyond, and I just wanted you to know.

App. 711. The class definition is limited to disabled individuals who have experienced access barriers in interior access aisles. This disabled customer reportedly experienced clear aisles at his local Ollie’s store in Georgia, so he is not a potential class member. The District Court clearly erred by relying on this email as evidence of a potential class member.

There may be others. Ollie’s argues that other customer complaints, closely read, similarly do not support the existence of class members. For example, Ollie’s argues that the District Court should have excluded two customer complaints made outside of Pennsylvania’s two-year statute of limitations for personal injury claims. The District Court never considered these arguments. On remand, the District Court should deter- mine whether the remaining complaints support the existence of putative class members. To do so, the District Court must be able to infer from the complaints that an Ollie’s customer with a mobility disability suffered or will suffer a common ADA injury that falls within the putative class definition. Otherwise, the District Court cannot rely on the customer complaints to determine the existence of putative class members.

17 In any event, even assuming all eleven remaining cus- tomer complaints support a finding that there are at least eleven putative class members, and considering the declaration and the statistical evidence together, as the District Court did, we still find the evidence far too speculative. To recap, the com- munity survey tells us nothing concrete about the portion of disabled residents who shop at Ollie’s stores. The declaration tells us nothing about what portion of disabled customers suf- fered common ADA injuries, and little about what number of disabled residents shop at Ollie’s. And the customer com- plaints are few. Eleven complaints over almost four years of company operations are hardly evidence of a sizable class. The customer complaints overall give us little reason to conclude that judicial economy supports depriving the apparently small number of complainants of their day in court by aggregating their individual claims in a classwide suit.

In short, after considering the record evidence, we have proof of a class that consists of Allen, Mullen, and at most eleven others. To establish numerosity, plaintiffs must do more to prove the existence of actual class members. See In re Modafinil Antitrust Litig., 837 F.3d at 250 (suggesting a class of twenty or less would be too small to justify a class action). If plaintiffs cannot carry the burden on numerosity, Allen and Mullen may always seek relief individually.

While “the number of class members is the starting point,” trial courts should weigh other factors relevant to the practicability of joinder under Rule 23(a)(1), including “judi- cial economy, the claimants’ ability and motivation to litigate as joined plaintiffs, the financial resources of class members, the geographic dispersion of class members, the ability to iden-

18 tify future claimants, and whether the claims are for injunctive relief or for damages.” In re Modafinil Antitrust Litig., 837 F.3d at 250, 253. Plaintiffs argue that factors other than the numerosity of the class also support the District Court’s finding that joinder of class members would be impracticable. That may well be. But the District Court never exercised its broad discretion to consider these other Rule 23(a)(1) factors, and we are a court of review, so we decline to weigh these factors for the first time on appeal. On remand, however, the District Court remains free to consider plaintiffs’ arguments and decide whether joinder would be impracticable based on all the rele- vant factors. We do not decide whether plaintiffs may show that joinder would be impracticable on this record. We hold only that the numerosity evidence considered alone is not enough to satisfy Rule 23(a)(1).

A class may be certified only if “there are questions of law or fact common to the class.” Fed. R. Civ. P. 23(a)(2). “Commonality requires the plaintiff to demonstrate that the class members have suffered the same injury. This does not mean merely that they have all suffered a violation of the same provision of law.” Dukes, 564 U.S. at 349–50 (citation and quotation marks omitted). Instead, the claims “must depend upon a common contention.” Id. at 350. “That common con- tention, moreover, must be of such a nature that it is capable of classwide resolution—which means that determination of its truth or falsity will resolve an issue that is central to the validity of each one of the claims in one stroke.” Id. This test ensures that the “claims can productively be litigated at once.” Id. When deciding whether the class raises a common question, “the court cannot be bashful. It must resolve all factual or legal disputes relevant to class certification, even if they overlap

19 with the merits—including disputes touching on elements of the cause of action.” Marcus, 687 F.3d at 591 (quotation marks omitted).

The District Court abused its discretion when finding commonality for two reasons. First, it misapplied the relevant standards and certified a geographically overbroad class. Sec- ond, as we explained in Steak ’n Shake, a broad term like “access barriers” does not give rise to a common injury under the ADA.

The District Court found commonality satisfied for a class consisting of all Ollie’s stores in the United States. The District Court reasoned that “[i]f Ollie’s policies and proce- dures do, in fact, cause access barriers to unlawfully restrict individuals with disabilities from obtaining their desired goods, then proposed members who endured violations have suffered the same injury, the resolution of which will resolve a central issue in one fell stroke.” Ollie’s, 2021 WL 1152981, at *7. “As a result,” the District Court concluded, “Plaintiffs have satisfied their burden of showing by a preponderance of the evidence that there are questions of law or fact common to the proposed class.” Id. The conclusion does not follow from the premise.

Before certifying the proposed class, the District Court must answer the very question it asked: whether plaintiffs have significant proof that Ollie’s corporate policies, procedures, or practices in fact cause discrimination by stores nationwide. Posing a hypothetical common question is not enough to sat- isfy plaintiffs’ burden of proof. There must be evidence the class proceeding will likely “produce a common answer.”

20 Dukes, 564 U.S. at 352. By failing to answer the commonality question, the District Court deferred plaintiffs’ need to show commonality.

The District Court’s legal error is not harmless. Our review of the record shows that commonality is not met by a preponderance of the evidence for this nationwide class. It is not enough that Ollie’s has corporate policies and that some or all stores in Pennsylvania pay inadequate attention to aisle accessibility. Stitching together a corporate-wide class requires more.

“Rule 23 requires more than allegations, initial evi- dence, or a threshold showing. It requires a showing that each of the Rule 23 requirements has been met by a preponderance of the evidence at the time of class certification.” Ferreras v. Am. Airlines, Inc., 946 F.3d 178, 184 (3d Cir. 2019). When proceeding on a corporate-wide basis, the Supreme Court has required proof of a policy or practice of discrimination before certifying a corporate-wide class. Dukes, 564 U.S. at 353. Without a corporate-wide policy that causes discrimination (including disparate impacts when relevant, as under Title VII), a plaintiff must have significant proof of a common mode of exercising discretion that “pervades the entire company,” not just stores in some states or regions of the country. Rodriguez v. Nat’l City Bank, 726 F.3d 372, 383–85 (3d Cir. 2013) (quoting Dukes, 564 U.S. at 356). There is no significant proof of either here.

Like any large retailer, Ollie’s has several corporate pol- icies governing its stores. These include visual store standards governing the placement and marketing of goods, general safety, loss prevention, and maintenance policies, and a “Yes, I Can” program, requiring stores to retrieve goods for patrons

21 that have trouble accessing them. Ollie’s specifically requires stores to ensure wheelchairs can pass easily through aisles. So as in Dukes, Ollie’s “announced policy forbids” the discrimi- natory conduct alleged by the class—inaccessible interior aisles. Dukes, 564 U.S. at 353. Ollie’s, to be sure, allows local stores discretion when maintaining adequate paths of travel for wheelchairs. But that kind of discretionary decision-making “is just the opposite of a uniform . . . practice that would provide the commonality needed for a class action.” Id. at 355. As the Supreme Court said in Dukes, allowing stores discretion is “a very common and presumptively reasonable way of doing business.” Id. It is not evidence of a common corporate-wide injury.

On appeal, plaintiffs focus on Ollie’s visual store stand- ards. They argue that the standards emphasize placing as much stock as possible on the sales floor. For example, they point out that photographs in the visual standards illustrate items stacked to the side of aisles as well as tight placement of clothing racks. Plaintiffs stress in their briefing that Ollie’s visual store stand- ards are “a plausible explanation,” a “plausible causal connec- tion,” or “a plausible, direct cause of the proliferation of alleg- edly discriminatory barriers.” Appellees’ Br. 15, 27. Perhaps. But plaintiffs must do more than assert a plausible causal explanation at this stage. They must show that the visual store standards are more likely than not a common cause of a failure to maintain accessible aisles across Ollie’s stores in the United States.

They have not met that burden. There is no proof that the visual standards cause inaccessible aisles across all Ollie’s stores nationwide. The investigative record is limited to stores in Pennsylvania. On this record, we do not know whether the visual standards “may have resulted in” discrimination “in

22 some regions . . . but not at all in others.” Rodriguez, 726 F.3d at 385. Proceeding on a corporate-wide basis against a corpo- ration with over four hundred stores in twenty-nine states requires more than plausible allegations backed by Pennsylvania-only evidence.

The only evidence from outside Pennsylvania is less than a dozen customer emails reporting inaccessible aisles.4 Setting aside the potential inadmissibility of the emails, Dukes rejected “anecdotal evidence” as “too weak” to support a com- mon practice. Dukes, 564 U.S. at 358. Dukes involved 120 employee affidavits, signed under penalty of perjury, sharing stories of employment discrimination by Wal-Mart supervi- sors. Id. The Supreme Court said these affidavits proved noth- ing: “More than half of these reports are concentrated in only 6 States (Alabama, California, Florida, Missouri, Texas, and Wisconsin); half of all States have only one or two anecdotes; and 14 States have no anecdotes about Wal–Mart’s operations at all.” Id. The anecdotal evidence here is far weaker than in Dukes. Less than a dozen email anecdotes over four years, from a corporation with over four hundred stores in twenty-nine states and thousands of employees exercising discretion, “prove nothing at all.” Id. at 358 n.9.

The District Court abused its discretion by certifying a corporate-wide class on this record. We leave it to the District Court to decide whether a geographically narrower class lim-

4 As we noted earlier, at least one customer reportedly experienced clear aisles. Another customer complained about the same store in Monaca, Pennsylvania, that plaintiff Mullen visited.

23 ited to some or all Ollie’s stores in Pennsylvania would satisfy the commonality requirement.

The District Court also abused its discretion when find- ing commonality for a separate reason. The District Court cer- tified a class embracing all persons with qualified mobility dis- abilities who have “experienced access barriers in interior paths of travel.” Ollie’s, 2021 WL 1152981, at *8–9. That class definition conflicts with our decision in Steak ’n Shake. A class that includes any and all access barriers is overbroad.

In Steak ’n Shake, the trial court certified a class con- sisting of persons with qualified disabilities who “encountered accessibility barriers at any Steak ’n Shake restaurant.” 897 F.3d at 487–88. We reversed the trial court’s commonality finding for two independent reasons.

First, the class representatives’ alleged injuries were based on excessively steep parking slopes, but the class was not limited to restaurant patrons who suffered an injury in a parking lot. Id. at 489–90. Second, and more relevant here, even if the class definition were limited to parking facilities, we observed, “the wide variety of [ADA] regulations . . . reveal that there are still various types of ADA violations that could occur specifically in a parking facility.” Id. at 490. Access bar- riers could include excessively steep parking lots but could also include inadequate signs. Id. “The wide variety of potential ADA violations captured in the broad class definition,” we held, meant that the claims could not be litigated together all at once. Id.

24 The class definition here similarly applies to any “access barriers.” Ollie’s, 2021 WL 1152981, at *9. Plaintiffs argue that Steak ’n Shake is distinguishable because the class is limited to access barriers in interior paths of travel, not every part of a store, but that entirely ignores our second reason for finding no commonality. In Steak ’n Shake, we warned against the broad term “accessibility barriers,” as it sweeps in a broad array of potential claims with little in common. The same is true here. Some “access barriers” are fixtures, like pillars, fixed tables, or aisle shelves. There is no evidence those types of fixed barriers result from any common policy or employee practice. Plaintiffs have not shown that Ollie’s has any central- ized blueprint or policy that requires stores to build narrow aisles or place pillars, tables, and shelving in the middle of the way. Cf. Colorado Cross Disability Coal. v. Abercrombie & Fitch Co., 765 F.3d 1205, 1216–17 (10th Cir. 2014) (porch design common to all 231 stores involved raised common question). Without evidence of a centralized store blueprint, we cannot say claims against those types of access barriers in interior paths of travel can be productively litigated together.

On appeal, plaintiffs focus their argument on movable barriers like merchandising, clothing racks, inventory carts, and the like. Plaintiffs mainly argue that Ollie’s stores violate their obligation to maintain 36-inch-wide accessible aisles by recurringly placing merchandising in the way. But that is not the class the District Court certified. At plaintiffs’ request, the District Court certified a class that applies to any kind of access barrier in interior paths of travel, not just merchandising wares blocking accessible aisles. We cannot cure the overbreadth of the class definition on appeal. We leave it to the District Court to decide whether a narrower class limited to particular mer- chandising wares or particular merchandising display practices

25 blocking interior aisles could satisfy the commonality require- ment.

* * *

Plaintiffs have failed to clear the first two hurdles of Rule 23(a). We will vacate and remand. We need not decide whether the remaining requirements of Rule 23 were satisfied. On remand, however, the District Court should clarify what classwide legal theory or theories of liability plaintiffs are pur- suing and determine whether each is suitable for classwide proof and common relief. As we have explained, trial courts must include in the certification order or opinion “(1) a readily discernible, clear, and precise statement of the parameters defining the class or classes to be certified, and (2) a readily discernible, clear, and complete list of the claims, issues or defenses to be treated on a class basis.” Steak ’n Shake, 897 F.3d at 488 n.21. Trial courts must then “determine what ele- ments plaintiffs would have to prove under that theory to reach a finding of liability and relief, and then assess whether this proof can be made within the parameters of Rule 23.” Hohider v. United Parcel Serv., Inc., 574 F.3d 169, 197 (3d Cir. 2009). And under Rule 23(b)(2), the provision under which plaintiffs here seek class certification, trial courts must explain how classwide relief would be appropriate for each legal injury. “Rule 23(b)(2) applies only when a single injunction or declar- atory judgment would provide relief to each member of the class. It does not authorize class certification when each indi- vidual class member would be entitled to a different injunction or declaratory judgment against the defendant.” Dukes, 564 U.S. at 360.

There are significant cohesion concerns with some of the theories of classwide relief advocated by plaintiffs. For

26 example, to the extent plaintiffs seek removal of “architectural barriers” in Ollie’s existing facilities, liability turns on a variety of individualized factors, including “the nature and cost of” the steps needed to remove each barrier. 42 U.S.C. § 12181(9)(A). Common proof and common relief relevant to that theory may prove elusive. On remand, the District Court must address these differing ADA standards and rules to determine whether common proof and common relief would be available for each distinct claim raised by the putative class.

27 PORTER, Circuit Judge, concurring.

Today, we sidestep one of the principal legal issues raised by this appeal: whether the Federal Rules of Evidence apply to fact evidence introduced in support of class certifica- tion. Respectfully, I see no reason to duck the question. The issue is properly presented and adequately briefed. The District Court may still need to answer this question on remand. And courts are divided. By our indecision, we prolong needless uncertainty in an important area of the law, and we undermine the uniformity required by the Federal Rules of Evidence. I would prefer to end any lingering uncertainty now, by holding that statutory text and precedent require applying the Federal Rules of Evidence before certifying a class under Rule 23.

As the Supreme Court has emphasized, “Rule 23 does not set forth a mere pleading standard. A party seeking class certification must affirmatively demonstrate his compliance with the Rule—that is, he must be prepared to prove that there are in fact sufficiently numerous parties, common questions of law or fact, etc.” Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350 (2011). “This calls for a rigorous analysis that usually requires courts to make factual findings and legal conclusions that overlap the underlying merits of the suit.” Mielo v. Steak ‘n Shake Operations, Inc., 897 F.3d 467, 482 (3d Cir. 2018).

In Blood Reagents, we held rigorous analysis means “that a plaintiff cannot rely on challenged expert testimony . . . to demonstrate conformity with Rule 23 unless the plaintiff also demonstrates, and the trial court finds, that the expert tes- timony satisfies the standard set out in Daubert.” In re Blood Reagents Antitrust Litig., 783 F.3d 183, 187 (3d Cir. 2015). We rejected the trial court’s acceptance of evidence that “could evolve” into admissible form later. Id. at 186. Daubert, of course, is based on Federal Rule of Evidence 702. See Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 589–92 (1993). So, under Blood Reagents, expert evidence used to certify a class action must be admissible under Federal Rule of Evidence 702.

We have never addressed whether fact evidence, rather than expert opinion, must likewise be admissible. The District Court rejected Ollie’s hearsay objection to the customer ser- vice emails by holding that non-expert evidence used to certify a class need not be admissible. See Allen v. Ollie’s Bargain Outlet, Inc., No. 2:19-CV-281, 2021 WL 1152981, at *5 n.5 (W.D. Pa. Mar. 26, 2021). That conclusion conflicts with the Federal Rules of Evidence.

The Federal Rules of Evidence are an exercise of legis- lative authority, so we read the rules “as we would any statute.” Daubert, 509 U.S. at 587. “The specific courts and proceedings to which the rules apply, along with exceptions, are set out in Rule 1101.” Fed. R. Evid. 101(a). Rule 1101 says that the rules apply to “United States district courts” and in “civil cases and proceedings.” Fed. R. Evid. 1101(a), (b). That means all civil cases and proceedings unless an exception applies.

Rule 1101 makes three exceptions, but as the Seventh Circuit has noted, Rule 23 proceedings are “not among the pro- ceedings excepted.” Mars Steel Corp. v. Cont’l Bank N.A., 880 F.2d 928, 938 (7th Cir. 1989) (en banc) (holding that Rule 23(e) fairness hearings are not exempt). The first two excep- tions apply to determinations “on a preliminary question of fact governing admissibility” and “grand-jury proceedings,” so they are irrelevant. Fed. R. Evid. 1101(d)(1), (2). A third and

2 broader exception applies to “miscellaneous proceedings such as: extradition or rendition; issuing an arrest warrant, criminal summons, or search warrant; a preliminary examination in a criminal case; sentencing; granting or revoking probation or supervised release; and considering whether to release on bail or otherwise.” Fed. R. Evid. 1101(d)(3). No civil proceedings are listed in the miscellaneous-proceedings exception.

While the list is not exclusive, in context, the miscella- neous-proceedings exception is best read as limited to closely analogous collateral proceedings, like hearings to transfer a juvenile delinquent for prosecution as an adult. See Gov’t of Virgin Islands in Interest of A.M., 34 F.3d 153, 161–62 (3d Cir. 1994) (allowing hearsay in a juvenile transfer hearing because the hearing was analogous to a preliminary examination in a criminal case). Otherwise, the exception would swallow the rule. Even if the exception may be extended to some ordinary civil proceedings, class certification proceedings are not closely analogous to any of the listed “miscellaneous proceed- ings,” so context suggests they do not fall under this exception, much like Rule 23(e) hearings. The conclusion is clear: Class certification proceedings are not exempt from the rules of evi- dence. See Anderson Living Tr. v. WPX Energy Prod., LLC, 306 F.R.D. 312, 378 n.39 (D.N.M. 2015) (“The similarity of a class certification hearing to a trial suggests that a class certifi- cation hearing is not a ‘miscellaneous proceeding such as’ a hearing on sentencing, extradition, preliminary examination, probation violation, or setting bail.”).

Our decision in Blood Reagents, moreover, prevents us from dispensing with the Federal Rules of Evidence. The District Court distinguished Blood Reagents as involving expert evidence. Ollie’s, 2021 WL 1152981, at *5 n.5. But for purposes of this question there is no principled basis for distin-

3 guishing between fact and expert evidence. Nothing in the rules of evidence allows us to selectively apply them. On the contrary, Federal Rule of Evidence 1101 says that “[t]hese rules”—meaning all rules, including hearsay rules—apply to civil proceedings generally. Fed. R. Evid. 1101(a), (b).

No hearsay exception applies to Rule 23 proceedings either. Hearsay is generally inadmissible unless allowed by rules adopted by the Supreme Court or statute. Fed. R. Evid. 802. Several rules of civil procedure permit proof by affidavit instead of live testimony, allowing modest exceptions to the hearsay rule. See, e.g., Fed. R. Civ. P. 4(l), 32(a)(1)(B), 65(b). One general exception is Federal Rule of Civil Procedure 43(c). That rule allows affidavits “[w]hen a motion relies on facts outside the record.” Fed. R. Civ. P. 43(c); see also 28 U.S.C. § 1746 (permitting the use of declarations instead). But while Federal Rule of Civil Procedure 43(c) allows considera- ble flexibility in avoiding the live testimony required by the hearsay rule, it does not allow simply attaching hearsay—like the customer complaints—to a motion. That is what happened here.

In short, “simple logic indicates,” and statutory text confirms, that Rule 23 is not satisfied when the “evidence prof- fered would not be admissible as proof of anything.” Behrend v. Comcast Corp., 655 F.3d 182, 215 n.18 (3d Cir. 2011) (Jordan, J., concurring in the judgment and dissenting in part), rev’d, 569 U.S. 27 (2013). Rigorous analysis and statutory text demand nothing less than admissible evidence at the time of certification.

4 II

Overlooking Federal Rule of Evidence 1101, several circuits have held that the rules of evidence do not apply to class certification proceedings. The first circuit to openly adopt this approach was the Eighth Circuit. See In re Zurn Pex Plumbing Prod. Liab. Litig., 644 F.3d 604, 611–14 (8th Cir. 2011). In Zurn, the Eighth Circuit affirmed a trial court’s watered-down Daubert analysis, rejecting the defendants’ argument that evidence under Rule 23 must “ultimately be admissible at trial.” Id. at 611. Judge Gruender dissented. Id. at 626–30. The Ninth Circuit followed some years later. In the Ninth Circuit, “[i]nadmissibility alone is not a proper basis to reject evidence submitted in support of class certification.” Sali v. Corona Reg’l Med. Ctr., 909 F.3d 996, 1004 (9th Cir. 2018).1 A trial court “may consider whether the plaintiff’s proof is, or will likely lead to, admissible evidence.” Id. at 1006. “But admissibility must not be dispositive. Instead, an inquiry into the evidence’s ultimate admissibility should go to the weight that evidence is given at the class certification stage.” Id. The Sixth Circuit has recently joined in part. It has held, “as have the Eighth and Ninth Circuits, that . . . ‘eviden- tiary proof’ need not amount to admissible evidence, at least with respect to nonexpert evidence.” Lyngaas v. Curaden AG, 992 F.3d 412, 428–29 (6th Cir. 2021).

Several circuits disagree with some or all of this. The First Circuit has rejected inadmissible hearsay evidence to sup- port standing for class members, reasoning that “class certifi-

1 But see Dukes v. Wal-Mart Stores, Inc., 603 F.3d 571, 639– 40 (9th Cir. 2010) (Ikuta, J., dissenting) (arguing that inadmissible expert testimony cannot be used to meet Rule 23), rev’d, 564 U.S. 338 (2011).

5 cation provides no occasion for jettisoning the rules of evi- dence and procedure, the Seventh Amendment, or the dictate of the Rules Enabling Act.” In re Asacol Antitrust Litig., 907 F.3d 42, 53 (1st Cir. 2018). That same logic would require applying the Federal Rules of Evidence to support class certi- fication. The Fifth Circuit has held that “findings must be made based on adequate admissible evidence to justify class certifi- cation.” Unger v. Amedisys Inc., 401 F.3d 316, 319 (5th Cir. 2005).2 At least two circuits have held, as we did in Blood Reagents, that expert testimony must be admissible under Daubert at the class certification stage. See Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 812–13 (7th Cir. 2012); Prantil v. Arkema Inc., 986 F.3d 570, 575–76 (5th Cir. 2021) (“[I]f an expert’s opinion would not be admissible at trial, it should not pave the way for certifying a proposed class.”).

I agree with the First and Fifth Circuits: Evidence used to certify a class must be admissible. The Sixth, Eighth, and Ninth Circuits overlook Federal Rule of Evidence 1101 and the rigorous analysis required by precedent. The various argu- ments they marshal in support of dispensing with the rules of evidence are unpersuasive. I will address each argument in turn.

2 The Sixth Circuit dismissed Unger’s admissibility requirement as dictum. Lyngaas v. Curaden AG, 992 F.3d 412, 430 (6th Cir. 2021). But the Fifth Circuit has relied on Unger to require admissible evidence. Prantil v. Arkema Inc., 986 F.3d 570, 575–76 & n.18 (5th Cir. 2021).

6 A

First, these circuits point to Rule 23(c)’s requirement that class actions be certified at “an early practicable time after a person sues or is sued as a class representative.” Sali, 909 F.3d at 1004 (quoting Fed. R. Civ. P. 23(c)(1)(A)). The need for speed, these courts reason, weighs against applying the Federal Rules of Evidence. That is unpersuasive. This rule was amended in 2003 to abrogate the certify-first-ask-questions- later practice followed in some circuits. In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 318–19 (3d Cir. 2008). Before 2003, Rule 23 said the class must be certified “as soon as practicable after commencement of an action.” Id. at 318. The change to “an early practicable time” was meant to encour- age rigorous compliance with the requirements of Rule 23 before certifying a class. Id. Under the current rule, “class cer- tifications are no longer conditional,” so a trial court “should delay certifying a class until it is satisfied that all Rule 23 requirements have been met.” Zurn, 644 F.3d at 629 (Gruender, J., dissenting). Requiring that evidence be admissi- ble does not conflict with this open-ended rule. If anything, the 2003 amendment suggests trial courts should not defer admis- sibility rulings relevant to certification until trial. See, e.g., Sali, 909 F.3d at 1006 (deferring admissibility in tension with 2003 amendment).

Second, these circuits assert that an order certifying a class is merely “tentative” and “preliminary,” as “[a]n order that grants or denies class certification may be altered or amended before final judgment.” Id. at 1004 (quoting Fed. R. Civ. P. 23(c)(1)(C)). Because class certification orders are not technically final, “common sense,” they say, suggests “the for-

7 mal strictures of trial” should not apply at the certification stage, including the Federal Rules of Evidence. Id. Otherwise, they argue, class certification proceedings would turn into evidentiary shooting matches. Id. For legal support, these cir- cuits lean on the Supreme Court’s statement that in class pro- ceedings, “a preliminary determination of the merits may result in substantial prejudice to a defendant, since of necessity [a class proceeding] is not accompanied by the traditional rules and procedures applicable to civil trials.” Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 178 (1974), quoted in Zurn, 644 F.3d at 613–14. There are many problems with this line of argument.

For one, that snippet of Eisen preceded the 2003 amend- ments to Rule 23 and has since been repudiated as dictum. The relevant part of Eisen held only that class representatives could not shift the cost of providing class-member notice to defend- ants by showing the class was likely to prevail on the merits. Eisen, 417 U.S. at 177–78. In Dukes, the Supreme Court said that Eisen’s general warning against preliminary determina- tions of the merits is “the purest dictum and is contradicted by our other cases.” 564 U.S. at 351 n.6. Eisen’s related statement about “traditional rules and procedures” is dictum too. For good measure, Dukes also expressed “doubt” at the trial court’s conclusion that Daubert did not apply when certifying a class. Id. at 354. One could even say that “the Supreme Court has expressed disapproval of the position taken by” the Sixth, Eighth, and Ninth Circuits. Zurn, 644 F.3d at 627 (Gruender, J., dissenting); see also Behrend, 655 F.3d at 215 n.18 (Jordan, J., concurring in the judgment and dissenting in part) (noting that “it is implicit” in Dukes that Daubert applies). But even if it has not, after the 2003 amendments and Dukes, Eisen’s cryp-

8 tic dictum about “traditional rules and procedures” does not support dispensing with the rules of evidence.

For another, Rule 23 certification orders are not “tenta- tive” in any practical sense. Trial courts cannot make “tenta- tive” Rule 23 findings. “When courts harbor doubt as to whether a plaintiff has carried her burden under Rule 23, the class should not be certified.” Steak ’n Shake, 897 F.3d at 483. Under our precedent, “it is no longer accurate—however true it might have been in the past—that class certification hearings are preliminary or conditional in the sense that a judge is going to go back and reconsider his or her class certification order.” Linda S. Mullenix, Putting Proponents to Their Proof: Evidentiary Rules at Class Certification, 82 Geo. Wash. L. Rev. 606, 636 (2014). “Although a judge subsequently may revise a class certification order, this practice has become extremely rare.” Id. at 637. In all but exceptional cases, an or- der certifying a class will be the trial court’s final word on the matter.

For similar reasons, the rhetoric about evidentiary shooting matches is also behind the times. Class certification proceedings are already evidentiary shooting matches, some- times requiring extensive evidentiary hearings. Id. at 639–41. The question is whether the shooting match will be played according to the uniform rules enacted by Congress, no rules at all, or only the rules judges really like. The correct answer is the rules enacted by Congress.

Characterizing Rule 23 orders certifying a class action as “tentative” and “preliminary,” moreover, trivializes the con- sequences of certifying a class. “As a practical matter, the cer- tification decision is typically a game-changer, often the whole ballgame, for plaintiffs and plaintiffs’ counsel.” Marcus v.

9 BMW of N. Am., LLC, 687 F.3d 583, 591 n.2. Rule 23(b)(2) class certifications compel unnamed persons to join a lawsuit they do not control, litigated by lawyers they did not choose, where a judgment binds them, win or lose. Cooper v. Fed. Reserve Bank of Richmond, 467 U.S. 867, 874 (1984); Dukes, 564 U.S. at 362. Defendants also face significant practical con- sequences. Once a class is certified, the risk of “devastating loss” often leads to “in terrorem” class settlements even for “questionable claims.” AT&T Mobility LLC v. Concepcion, 563 U.S. 333, 350 (2011); see also Shady Grove Orthopedic Assocs., P.A. v. Allstate Ins. Co., 559 U.S. 393, 445 n.3 (2010) (Ginsburg, J., dissenting) (“A court’s decision to certify a class . . . places pressure on the defendant to settle even unmeritori- ous claims.”); Newton v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 259 F.3d 154, 165 (3d Cir. 2001) (“[C]ertifying the class may place unwarranted or hydraulic pressure to settle on defendants.”). And plaintiffs, too, may be denied meaningful redress if defendants are allowed to defeat class actions by lard- ing the record with inadmissible hearsay, unauthenticated rec- ords, or unreliable opinion evidence. The search for truth encouraged by the Federal Rules of Evidence cuts both ways.

The Eighth Circuit has also suggested that because class certification findings are made by a judge, not a jury, there is less reason to apply Daubert rigorously, and presumably other rules of evidence too. Zurn, 644 F.3d at 613. But this distinc- tion finds no support in the Federal Rules of Evidence or our caselaw. The Federal Rules of Evidence require applying Daubert faithfully in bench trials too. UGI Sunbury LLC v. A Permanent Easement for 1.7575 Acres, 949 F.3d 825, 832–33 (3d Cir. 2020). Some rules of evidence, to be sure, expressly reference jury trials and do not apply to bench trials. Rule 403,

10 for example, allows a trial court to “exclude relevant evidence if its probative value is substantially outweighed by a danger of . . . misleading the jury.” Fed. R. Evid. 403. That rule is irrelevant in a bench trial. But hearsay is generally inadmissi- ble no matter who the trier of fact happens to be. Fed. R. Evid. 802. So, as with Daubert, there is no “bench trial” exception to hearsay.

Reasonable minds may disagree over the wisdom or practicality of applying the Federal Rules of Evidence, or hear- say rules specifically, in Rule 23 certification proceedings. But those policy judgments are for the Supreme Court and Congress to make. See 28 U.S.C. § 2072(a). We must apply the rules of evidence faithfully within their proper scope. That scope includes Rule 23 proceedings.