The United States District Court for the Northern District of California recently issued two critical decisions tackling Federal Rule of Evidence 702 challenges that underpin a district court’s Federal Rule of Civil Procedure 23 class certification analyses. The court found that when class certification requests “rise[] or fall[] on the validity and reliability of the opinions” of experts, an expert’s inability to establish class-wide methods for determining the elements of a cause of action can prove fatal to such requests. Still, once plaintiffs inevitably appeal trial court denials of class certification due to exclusion of testimony, the U.S. Court of Appeals for the Ninth Circuit has been considerably less open to such an application of Rule 702.  

Background

Traditionally, Rule 23 class certification analyses focused narrowly on early pleadings and shied away from analysis of merits of a case. Courts often interpreted Rule 23 to require class certification decisions at “an early practicable time after a person sues or is sued as a class representative,” often before discovery, and some district courts even imposed strict filing deadlines.^[1] The Supreme Court stated that “nothing in either the language or history of Rule 23 … gives a court any authority to conduct a preliminary inquiry into the merits” when deciding certification.^[2] The Ninth Circuit similarly held that plaintiffs need not prove the merits or “establish a probability that the action will be successful.”^[3] As a result, any discovery, let alone expert discovery and admissibility considerations, was generally excluded from class certification proceedings.

That stance slowly transitioned with the Supreme Court’s analysis in General Telephone Co. of the Southwest v. Falcon, expanding past a rigid reading of the rule.^[4] The Supreme Court further clarified its position in Wal-Mart Stores, Inc. v. Dukes, citing that “‘rigorous analysis’ will entail some overlap with the merits of the plaintiff’s underlying claim.”^[5] This ‘rigorous analysis’ standard began to open the door for further consideration of case merits in the determination of class certification. For a moment, it seemed that even the Ninth Circuit understood the need for a more merit-based undertaking to certification analyses.^[6]

Nevertheless, the Ninth Circuit’s treatment of expert evidence at class certification has been largely inconsistent. Although the Supreme Court in Wal-Mart suggested Daubert analyses likely apply at this stage, and various circuits have now recognized that inadmissibility of expert evidence can be fatal to certification, the Ninth Circuit’s approach has sent, at best, mixed signals.^[7] In Sali v. Corona Reg’l Med. Ctr., the appellate court held that “proof [in support of class certification] need not be admissible evidence,” and that “admissibility alone is not a proper basis to reject evidence submitted in support of class certification.”^[8] The court also held that while “a district court should evaluate admissibility under the standard set forth in Daubert,” “admissibility must not be dispositive.” In Olean Wholesale Grocery Coop., Inc. v. Bumble Bee Foods LLC, the court confirmed that “plaintiffs may use any admissible evidence” to meet Rule 23(b)(3)’s requirements.^[9] Again in Lytle v. Nutramax Labs, however, the court reaffirmed that “there is no requirement that the evidence relied upon by Plaintiffs to support class certification be presented in an admissible form.”^[10] These, and additional conflicting rulings that have followed, have left the Ninth Circuit’s approach increasingly at odds with Rule 702’s core purpose of excluding inadmissible expert testimony, they have deepened an entrenched circuit split, marking a departure from the Supreme Court’s guidance in Wal-Mart and its progeny.^[11]

The Current Stage

Within the past year, the Northern District issued two additional decisions that found the inadmissibility of expert evidence, in the absence of other evidence, to be fatal to the respective plaintiffs’ class certification efforts.

In Klein v. Meta Platforms, Inc., the district court denied class certification after concluding that plaintiffs relied solely on the inadmissible expert opinions of economist Dr. Nicholas Economides to establish class-wide antitrust injury. 766 F. Supp. 3d 956 (N.D. Cal. 2025). Plaintiffs allege that Meta’s violation of the Sherman Act and relevant California law created an illegal monopoly in the “personal social network services” (PSNS) market through misrepresentations about use and data collection practices that “deprive[d] its competitors of the ability to compete.” They contend that absent these misrepresentations, Meta would have faced a competitive PSNS market requiring payment to users for their data. The court emphasized that the admissibility of Dr. Economides’s opinions was the “threshold question” that would “effectively determine the fate of the user plaintiffs’ certification request.” The court was unpersuaded by Dr. Economides’s opinion that user costs take “the form of data collection and use,” and that competition would have forced Meta to reduce the effective cost of Facebook use to a “negative price” of a monthly $5.00 to users. Finding his analysis insufficient to show injury—namely, the alleged loss of a $5.00/month payment—the court excluded his opinions under Rule 702 and denied class certification.

The court agreed with Meta, concluding that “‘[t]he mere theory of ‘negative price markets’ … cannot bridge the analytical gap between how [Facebook] actually competes for users and how Economides speculates it would.’” Before undertaking a complete Rule 702 analysis, the court emphasized that “‘at all stages … Rule 702 tasks a district court judge with ensuring that an expert’s testimony both rests on a reliable foundation and is relevant to the task at hand.’”^[12] It then focused primarily on Dr. Economides’s theory of antitrust injury, explaining that to show cognizable class-wide harm, Dr. Economides must establish five separate links, the key one being that Meta would compete on price rather than quality. The court found this key link in Dr. Economides’s theory “too great an analytical” leap, finding it unreliably speculative and unsupported by the record, which showed that Meta and the broader PSNS market consistently competed on quality rather than price.

The court also identified broader flaws in plaintiffs’ evidence. It found no reliable economic literature supporting Dr. Economides’ conclusions and determined that his analyses of Meta’s brief, internal pricing proposals or compensated‑data ideas—none of which matured into real proposals—did not address the issue at hand. The court emphasized that Dr. Economides consistently failed to explain why Meta would choose user payments as its competitive strategy, and it rejected the notion that this gap was merely impeachable, noting, rather, that it was central to his entire theory of a cognizable antitrust injury.

Lacking a sufficient showing that “‘questions of law and fact common to class members predominate over any questions affecting only individual members’” under Rule 23(b)(3), the district court found that plaintiffs failed to establish commonality or predominance for their antitrust claims.^[13] The court indicated that plaintiffs has not offered a class-wide method for proving antitrust injury supported by admissible expert testimony and, without reaching the remaining Rule 23 factors, denied class certification. The court subsequently granted Meta’s motion to exclude Dr. Economides in the individual plaintiffs’ remaining claims and, thereafter, granted summary judgment in favor of Meta—plaintiffs’ antitrust claims could no more withstand the exclusion of Dr. Economides’s opinions on summary judgment than at certification. Although the Ninth Circuit had earlier denied Plaintiff’s immediate appeal of the class certification denial, Plaintiffs’ appeal of both rulings is now pending before the Ninth Circuit.

The Northern District reaffirmed its general approach to Rule 702 in In re Apple iPhone Antitrust Litigation, adopting a more cautious pathway to class certification and, yet, still ending with a decertified class. No. 4:11-cv-6714-YGR, 2025 LX 555485 (N.D. Cal. Oct. 27, 2025). In its initial denial of certification, the court reminded the parties that certification requires “rigorous analysis … ‘of the persuasiveness of the evidence presented,’” and, citing Ellis v. Costco Wholesale Corp., noted that a “district court must consider the merits if they overlap with the Rule 23(a) requirements.”^[14] It later reiterated that “Rule 23(b)(3)’s predominance requirement is demanding,” and that plaintiffs “must actually prove – not simply plead” compliance with all Rule 23 requirements, including predominance.^[15] If plaintiffs intend to rely on expert opinions, the methodology must be “capable of showing class-wide… impact.”^[16] Here, the district court concluded that plaintiffs’ Rule 702 shortcomings ultimately necessitated decertification because they failed to heed the court’s concerns about proving their claims class-wide, and those unresolved issues predominated over any common ones. Despite multiple attempts to satisfy the court’s reliability requirements for their experts’ methodologies, plaintiffs ultimately failed to show common proof for a central element of their antitrust claims—the injury.

Plaintiffs allege that Apple Inc. unlawfully monopolized the aftermarket for iPhone applications in violation of Section 2 of the Sherman Act, causing consumers to purportedly pay higher prices for iOS applications or in-app purchases. Plaintiffs claim Apple imposed supracompetitive commissions on iOS app developers, enabling it to control the application aftermarket, and that developers passed on those costs to consumers, resulting in the asserted injury.

With an eye towards certification, the district court conducted a detailed Rule 702 admissibility analysis of plaintiffs’ expert testimony and methodologies before denying their initial certification bid, finding that they failed to satisfy Rule 23(b)(3)’s predominance requirement or show a class-wide injury. The court explained that “at the class certification stage, ‘the relevant inquiry is a tailored Daubert analysis which scrutinizes the reliability of the expert testimony in light of the criteria for class certification and the current state of the evidence.’”^[17]

Plaintiffs relied on a proprietary econometric model and accompanying report from Professor Robert McFadden to estimate how competitive commission rates would affect app prices. The court declared his damages model unreliable, criticizing its “cherry-picked” values and confounding errors. The court granted Apple’s Rule 702 motion in part as to aspects of Professor McFadden’s opinions. Because plaintiffs had not met “their predominance burden because they rel[ied] on an unsound methodology” that failed to reliably illustrate which consumers were injured or in what numbers, the court denied class certification. Moreover, the court also noted that the model identified a substantial number of potentially uninjured class members.

Plaintiffs took a second, briefly successful, bite at the apple by addressing the court’s primary critiques of Professor McFadden’s methodology and narrowing the class to payors rather than individual Apple ID account holders. Plaintiffs represented that deduplicating Apple’s payor data would reduce the number of uninjured members, matching payors with multiple account records to consumers. The court, in granting certification, agreed that under the promised revised model, the predominating issue would be the “common question of whether Apple’s conduct caused class members to suffer an antitrust injury.” But it warned that Professor McFadden’s model, using Apple payor data, should reliably calculate damages and illustrate class-wide anticompetitive impact before trial, or the class would face modification or decertification.

The district court’s earlier warning proved prescient of the Rule 702 issues to come. After expert discovery, Apple moved to exclude the testimony of Darryl Thompson of JND Legal Administration and sought class decertification. Apple argued—and the court ultimately agreed—that plaintiffs still had not met the requirements of Rule 23(a) and (b), as they failed to establish predominance for both injury or damages and could not reliably identify class members or their injury status through admissible expert, or other, evidence. Plaintiffs had engaged Thompson to deduplicate data and “match” payor records to consumers to identify class members and calculate class damages, but the court found this support unsuccessful and insufficient.

The court, again, found Thompson’s opinions neither relevant nor reliable, warranting exclusion on multiple grounds. The court gave considerable weight to the rebuttal report of Apple’s witness, Dr. Victoria Stodden, in identifying the substantial errors in Thompson’s methodology and conclusions. First, the district court determined that Thompson’s opinions lacked “reliable foundation,” and he failed to offer “knowledge and experience of [the relevant] discipline.”^[18] Although experienced with basic data deduplication, Thompson was not qualified to perform the matching analysis needed to identify injured payors, lacked experience with the required techniques, failed to utilize best practices, and was unfamiliar with the appropriate statistical tools. Second, the court was unconvinced of the reliability of Thompson’s methodology. Consistent with Ninth Circuit guidance and the factored approach in Hankey, the court found that Thompson’s methodology 1) could not be tested, 2) had not been peer-reviewed, 3) did not have a determinable error rate, and 4) did not “enjoy general acceptance within the relevant scientific community.” The court further criticized the “black box” nature of Thompson’s methods, especially given significant errors in the resulting dataset. Finally, the court agreed with Dr. Stodden and Apple that his methods could not be reliably applied to Apple’s data, as his methods appeared to neither clean nor deduplicate the dataset with sufficient certainty or accuracy.

The court concluded that decertification was the only appropriate action, as the plaintiffs had failed to meet the predominance requirement of Rule 23(b)(3). Stressing that the burden to demonstrate the Rule 23 requirements still lies with “the party seeking certification,” the court specifically noted that given the plaintiffs’ proffered evidence in the form of expert opinion alone, the predominance inquiry was reliant on and overlapped greatly with its Daubert analysis. With no further evidence to identify plaintiffs’ injury status and damages, the court found that plaintiffs had failed to “prove … an essential element of the cause of action.”^[19] The court viewed plaintiffs’ position as “back to square one,” as there was no path to establishing injury “through a common body of evidence, applicable to the whole class.” Plaintiffs’ appeal of this matter is likewise pending before the Ninth Circuit.

The Next Act

Defendants in the Ninth Circuit can take some satisfaction from the Northern District’s recent fulsome application of Rule 702 analysis at the class-certification stage. But those decisions have limited effect in resolving the escalating circuit split absent clearer Ninth Circuit guidance on admissibility at this stage. In its recent decision in Noohi v. Johnson & Johnson Consumer, Inc., the Ninth Circuit, while affirming yet another class certification marred by a “not yet fully developed or executed” damages model, outlined potential guardrails for plaintiffs’ future requests. 146 F.4th 854, 861 (9th Cir. 2025). The court found a “full-blown Daubert assessment” unnecessary at certification because evidence need not be “presented in an admissible form at the class certification stage.” However, the Ninth Circuit did add that persuasive rebuttal evidence may undermine commonality and typicality-type inquiries. It further warned that without the eventual materialization of a properly executed methodological approach, plaintiffs risk renewed Daubert challenges or summary judgment against the entire class. Or, as seen in the Apple saga, perhaps swift decertification. The Noohi matter was denied certiorari by the Supreme Court, leaving the Ninth Circuit’s ruling in place and underscoring defendants’ ongoing efforts to put to rest this Ninth Circuit‑driven split.

Even with a likelihood for Ninth Circuit reversal of the above decisions, the Northern District’s consistent use of rigorous Rule 702 evaluations underscores the growing need for defendants to leverage meaningful opportunities in fast‑moving class actions. Certification can pressure defendants into settling cases built on inadmissible expert testimony and flawed methodologies, producing inefficient and inaccurate class‑wide outcomes. Where expert evidence is central, defendants should raise Rule 702 challenges early to secure prompt rulings that may defeat certification and halt the progression of inadmissible expert testimony, renew those challenges if certification is conditional, and, where certification is denied for lack of admissible expert support, consider pursuing early summary judgment.

[1] Fed. R. Civ. P. 23(c) (emphasis added); C.D. Ca. Local Rule 23-3; E.D. Pa. Local Rule 23.1(c).

[2] Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 177 (1974).

[3] Blackie v. Barrack, 524 F.2d 891, 901 (9th Cir. 1975).

[4] 457 U.S. 147, 160 (1982) (“Sometimes the issues are plain enough from the pleadings …, and sometimes it may be necessary for the court to probe behind the pleadings before coming to rest on the certification question. Even after a certification order is entered, the judge remains free to modify it in the light of subsequent developments in the litigation.”); Comcast Corp. v. Behrend, 569 U.S. 27, 33, 133 S. Ct. 1426, 1432 (2013).

[5] 564 U.S. 338, 351, 131 S. Ct. 2541, 2551 (2011).

[6] Abs Entm’t v. CBS Corp., 908 F.3d 405, 427 (9th Cir. 2018) (“To undertake that [rigorous analysis] may require discovery.”); see id. (citing Kamm v. California City Development Company, 509 F.2d 205 (9th Cir. 1975) (“deny[ing] discovery in [such cases] would be an abuse of discretion”).

[7] 564 U.S. 338 (2011); 569 U.S. 27 (2013) (where an expert’s methodology cannot narrow damages to the injury in question, it does not provide the required assurance that it is capable of damages calculations class-wide); Messner v. Northshore Univ. Health Sys., 669 F.3d 802, 812 (7th Cir. 2012); In re Blood Reagents Antitrust Litig., 783 F.3d 183, 187 (3d Cir. 2015).

[8] 909 F.3d 996, 1004 (9th Cir. 2018).

[9] 31 F.4th 651, 665 (9th Cir. 2022) (emphasis added).

[10] 114 F.4th 1011, 1024 (9th Cir. 2024).

[11] In re Telescopes Antitrust Litig., 348 F.R.D. 455, 467 (N.D. Cal. 2025); Bess v. Ocwen Loan Servicing LLC, 334 F.R.D. 432, 437, 400 (W.D. Wash. 2020); Flodin v. Cent. Garden & Pet Co., No. 21-CV-01631-JST, 2024 U.S. Dist. LEXIS 192904, 2024 WL 4565340, at *2-3 (N.D. Cal. Oct. 23, 2024); Aberin v. Am. Honda Motor Co. Inc., No. 16-CV-04384-JST, 2021 U.S. Dist. LEXIS 71680, 2021 WL 1320773, at *4 (N.D. Cal. Mar. 23, 2021).

[12]  766 F. Supp. 3d at 961 (emphasis added) (citing In re Google Play Store Antitrust Litig., No. 21-md-02981-JD, 2023 WL 5532128, at *5 (N.D. Cal. Aug. 28, 2023) (cleaned up) (quotation omitted)).

[13] Fed. R. Civ. P. 23(a); Fed. R. Civ. P. 23(b)(3) (emphasis added).

[14] No. 11-cv-6714-YGR, 2022 U.S. Dist. LEXIS 79593, at *30 (N.D. Cal. Mar. 29, 2022) (citing Ellis v. Costco Wholesale Corp., 657 F.3d 970, 982-83 (9th Cir. 2011)).

[15] In re Apple, 2025 LX 555485 at *23. (citing Halliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258, 275 (2014)).

[16] Id.

[17] Id. at *9 (citing Rai v. Santa Clara).

[18] Id. at *8 (citing Kumho Tire Co., Ltd. v. Carmichael, 526 U.S. 137, 141, 149 (1999)).

[19] Id. (citing Olean).