Class Actions Brief Blog

Class certification can be a key battleground in class actions. Successful certification of a class often gives plaintiffs leverage in a case, while defendants that successfully oppose class certification may significantly limit or eliminate their potential liability.

Parties often rely on experts to support their class certification arguments, particularly where large data sets must be analyzed to identify members of the proposed class or to analyze alleged damages. The expert testimony can assist the fact finder in evaluating the relevant evidence.

To ensure that an expert’s opinion is useful to the fact finder, Federal Rule of Evidence 702 and Daubert require that the expert is qualified, based on their background and knowledge, to offer an opinion on the relevant issue. Rule 702 and Daubert also require that the expert’s opinion is the product of reliable principles and methods, and that the expert reliably applied those principles and methods.

The application of the Daubert standard in class certification proceedings has been a heated topic, especially in the Ninth Circuit, as we have noted in posts regarding standards used to analyze expert models, the Ninth Circuit’s en banc decision regarding uninjured class members and statistical evidence, and the Supreme Court’s denial of certification in a case allowing certification based on an incomplete damages model. 

A recent case in the Northern District of California highlights again the importance of expert selection in a class certification fight. In In re Apple iPhone Antitrust Litigation, No. 4:11-cv-06714, 2025 WL 3124160 (N.D. Cal. Oct. 27, 2025), the court granted Apple’s Daubert motion to exclude the testimony of the plaintiffs’ expert, finding that he was unqualified and his methodology was unreliable. Because class certification depended on the expert’s analysis, the trial court also decertified the class.

According to the plaintiffs, Apple charges iOS app developers a commission that is higher than it would be in a competitive market, and the developers then pass that cost on to consumers by charging higher prices to buy the apps or to make in-app purchases. The plaintiffs allege that this practice allows Apple to unlawfully monopolize the aftermarket for iOS apps, which violates the Sherman Act.

After denying the plaintiffs’ first motion for class certification in 2021, the court granted their second motion in early 2024. The court cautioned, however, that the plaintiffs must deduplicate Apple’s payor data to maintain the class. Because a new payor record is created each time the payor creates a new Apple ID or changes their name, address, or payment method, a single consumer may have multiple accounts or payor records. Deduplication would ensure that the actual class of injured consumers could be ascertained, as required for class certification.

After the court initially certified the class, the plaintiffs began attempting to match over a billion payor records to individual consumers. The plaintiffs engaged Darryl Thompson, the Chief Information Officer of a legal administration company, for this work. He first attempted to “clean” the data to improve its quality. For example, this could include ensuring that “Apt 2” and “Apartment 2” in an address are read as the same data point. He then attempted to “match” the payor records to individual payors and validate those results.

Apple also engaged an expert, Victoria Stodden, Ph.D, an associate professor of industrial and systems engineering at the University of Southern California. Dr. Stodden identified numerous errors in Mr. Thompson’s work. For example, Mr. Thompson treated named plaintiff “Rob Pepper” and “Robert Pepper” as two different payors, even though the home address and credit card information associated with each record was the same, and he aggregated all payors with the first name “Kim.” He also identified 1.9 million “unique payors” with addresses in King Salmon, Alaska, a town with only 375 residents. In addition to identifying these errors, Dr. Stodden also noted that she could not replicate Mr. Thompson’s methods.

In considering Apple’s Daubert motion, the court relied heavily on Dr. Stodden’s report to identify errors in Mr. Thompson’s methodology and analysis. While the opinion does not explicitly say so, the opinion suggests that Dr. Stodden’s work was trustworthy and accurate.

The court identified several reasons why Mr. Thompson was not qualified to offer an expert opinion on data cleaning and matching for this case. The court noted that Mr. Thompson is not a statistician and does not have a degree or training in statistics. Mr. Thompson supervises class notice and distribution during claims administration for class action lawsuits. He had never previously performed a cleaning and matching exercise like he did in this lawsuit. Nor was his methodology grounded in literature or studies relating to data matching.

The court also found his methodology unreliable because: (1) it could not be tested or replicated due to Mr. Thompson’s failure to document many of his steps; (2) it was not based on a method that was peer-reviewed or generally accepted in the scientific community; and (3) Mr. Thompson could not provide an error rate or confidence interval, or otherwise meaningfully validate his work.

Finally, the court found that Mr. Thompson’s analysis was incorrect because he sought to match the names associated with payor records rather than matching actual payors to the records. As an example, the court noted that a child may use her parent’s credit card for in-app purchases, making the parent the ultimate payor despite the child’s name being listed in the record. If the parent also makes purchases, the payor identification for the parent should include both the parent’s and child’s in-app purchases.

For all of these reasons, the court excluded Mr. Thompson’s testimony and decertified the class.

This case shows that choosing the right expert can make or break class certification.^[1] Litigants should examine their expert’s background and experience carefully to ensure that they align with the expert’s role in the case. Showing a court that an expert possesses the relevant background specific to the analysis at hand can build the expert’s credibility with the court. Experts should also choose a methodology accepted by the scientific community, document that methodology, closely analyze any errors, and provide a clear explanation of their work for the court to evaluate.

[1] Earlier this year in an unpublished opinion, the Fourth Circuit affirmed a district court decision similarly finding an expert’s testimony inadmissible under Federal Rule of Evidence 702 and denying class certification. Davis v. Capital One N.A., No. 24-1507, 2025 WL 2445880 (Aug. 26, 2025).