Challenging a TCPA Data Expert: Reverse Lookups and Rule 702

More than 75 percent. That’s the share of 666 supposed class members, flagged from a sample by a plaintiff’s data expert in a wrong-number prerecorded-call case, that a defense expert found were the bank’s own customers, according to the Fourth Circuit’s August 2025 opinion in Davis v. Capital One. In cases like Davis, the class list comes from an expert running the defendant’s call records through outside databases. A company that tests that work, number by number, often has its strongest certification defense.

How do plaintiffs’ experts turn call logs into a class list?

They start with the defendant’s own records and filter them through commercial data. The method can break at any one of those filters.

Cell phone or landline. The claims for prerecorded calls (calls that play an artificial or recorded voice) and autodialed calls (made with equipment that can generate numbers randomly or sequentially) to cell phones need proof that each number was assigned to wireless service. In Davis, the expert’s first step was to analyze the bank’s records “and records belonging to telecommunications carriers to determine which of Capital One’s calls had been made to cell phones,” the Fourth Circuit wrote. In a Northern District of California case, the plaintiff proposed to use “information from TransUnion to exclude any calls made to landline numbers,” according to the order in Revitch v. Citibank.

Do Not Call Registry matching. The National Do Not Call Registry is the federal list of numbers whose owners don’t want telemarketing calls. The FCC’s rule bars telephone solicitations to a “residential telephone subscriber who has registered his or her telephone number on the national do-not-call registry,” under 47 C.F.R. § 64.1200(c)(2). In Mantha v. QuoteWizard.com, the expert used a vendor’s Registry lookup, which also reported each number’s registration date, and then ran a business-number lookup and a keyword screen to drop numbers tied to businesses.

Wrong numbers and reassigned numbers. This is the hardest step. The expert has to show that the person who got the call wasn’t the customer who gave consent. In Davis, the tool was a reverse append. “A reverse historical append involves inputting a phone number into a database to find out who it belongs to,” the Davis opinion explains in a footnote. The Davis expert also planned to use the Reassigned Numbers Database, whose records, the court noted, only go back to January 2021. I’d treat any wrong-number method as suspect until someone checks it against the company’s own account history.

Every number that survives the filters can become a claim in the class total. For Do Not Call claims, the statute reaches a person who received “more than one telephone call within any 12-month period by or on behalf of the same entity” in violation of the rules, and it allows “up to $500 in damages for each such violation,” which a court may raise to as much as three times that amount if the violation was willful or knowing, under 47 U.S.C. § 227(c)(5). The same paragraph makes it an affirmative defense that the caller established and implemented, with due care, reasonable practices and procedures to prevent violations. The prerecorded-call claim in Davis and the autodialer claim in Revitch arise under a different subsection, § 227(b), with its own damages terms. Either way, each false positive the expert leaves in adds exposure.

What did the December 2023 change to Rule 702 do?

It made the burden explicit and reworded the last requirement. Federal Rule of Evidence 702 now allows expert testimony only “if the proponent demonstrates to the court that it is more likely than not” that the testimony meets each requirement, including that it rests on sufficient facts or data and that the expert applied the method reliably, under the rule’s current text. Subpart (d) now requires that “the expert’s opinion reflects a reliable application of the principles and methods to the facts of the case.” The Supreme Court’s order adopting the amendment provided that it “shall take effect on December 1, 2023.”

The Advisory Committee was blunt about why. Its 2023 note says many courts had treated the sufficiency of an expert’s basis and the application of the method as questions of weight, and then states: “These rulings are an incorrect application of Rules 702 and 104(a),” per the committee note on Cornell’s Rule 702 page.

I think that matters most for data experts, because their weak spots are usually application, not principle. Nobody disputes that you can look up a phone number in a commercial database. The fight is over whether this expert, with these vendors and these date ranges, sorted these numbers correctly.

Does the amendment change how Ninth Circuit courts treat experts at certification?

Not in any Ninth Circuit decision discussed here. Federal courts in California still work from a 2018 Ninth Circuit rule that evidence offered for certification doesn’t have to be admissible. In Sali v. Corona Regional Medical Center, the court said a district court should evaluate expert evidence under Daubert, and then: “But admissibility must not be dispositive.”

The en banc court later held in Olean Wholesale Grocery Cooperative v. Bumble Bee Foods that plaintiffs must prove the Rule 23 facts by a preponderance of the evidence, and it listed expert evidence showing “nonsensical results such as false positives” as one kind that can fail. Then in 2024, Lytle v. Nutramax Laboratories allowed a “limited” Daubert review at certification when an expert’s model isn’t fully executed, while warning: “Merely gesturing at a model or describing a general method will not suffice.”

The Ninth Circuit applied Lytle again in July 2025. In Noohi v. Johnson & Johnson Consumer Inc., it affirmed certification based on an unexecuted damages model but added that “JJCI must be given the opportunity to test the admissibility and reliability of Dr. Roberts’ model once it has been fully executed.” Neither opinion turned on the 2023 amendment.

In the two Northern District of California orders discussed here, judges didn’t treat the amendment as a reason to change course. In a pet-food case, the defendant argued that the 2023 note requires an admissibility ruling at certification, and the Northern District of California answered, “But the Court does not read the advisory committee notes so broadly,” in its October 2024 order in Flodin v. Central Garden & Pet. The same court still denied certification because the damages model didn’t fit the theory of liability. Three weeks earlier, another judge in that district noted that the plaintiff called the amendment’s application “debatable,” found the plaintiff’s experts’ opinions, as relevant to certification, “satisfy amended Rule 702,” denied the defendant’s motions to exclude, and certified the class in Johnson v. Glock.

Here’s my view. In the Ninth Circuit, a motion to exclude the plaintiff’s data expert is worth filing, but it shouldn’t carry the whole defense. Build the record so the error rate defeats certification under Rule 23 even if the judge reads Sali to mean the testimony comes in. Olean’s warning about false positives is a close analogy to a reverse-lookup list that sweeps in the company’s own customers.

Which TCPA rulings excluded or credited the plaintiff’s expert?

The outcomes track the difficulty of the question the expert was asked to answer.

Excluded: Davis v. Capital One (E.D. Va. 2023, aff’d 4th Cir. 2025). The proposed class covered prerecorded calls to cell numbers not assigned to a current Capital One account holder. The district court excluded the plaintiff’s expert under Rule 702 in an October 2023 order, and the Fourth Circuit affirmed in an unpublished opinion. According to the appellate opinion, the expert never ran the final reverse-append step, the defense expert found that over 75 percent of the 666 numbers she flagged in a sample were customers, and the named plaintiff himself wouldn’t have been found because he’d registered his phone under a pseudonym. The court agreed that the error rate “weighs heavily against the admissibility of Verkhovskaya’s expert opinion.”

Certification denied: Revitch v. Citibank (N.D. Cal. 2019). The plaintiff’s method matched names from a TransUnion reverse lookup against the bank’s account records. The court found that “his expert did not examine or evaluate” the “last seen” date that TransUnion reports for each name, according to the order. Counting that date, the bank’s expert found that 120 of the 176 numbers the plaintiff’s method first flagged were documented in TransUnion’s data as associated with the customer on the account at the time of the call. The court denied certification because individual consent questions would predominate, which is the only point about consent I’ll make here: a consent defense needs evidence behind it before it can defeat certification.

Admitted: Mantha v. QuoteWizard.com (D. Mass. 2024). This was a Do Not Call text class, and the court applied the amended Rule 702 text. It denied the motion to exclude and certified the class, writing that “Using a widely available, long-standing, commercial service to run a database check is a perfectly normal and reasonable methodology,” in its August 2024 order. The court found the district court’s 2023 Davis ruling unpersuasive because it arose under a different section of the TCPA and involved a different analysis by the same expert, and it noted that “Numbers on the NDNCR are presumed residential.”

I read these cases this way. In Mantha, Registry matching and a business-number screen were treated as routine database work. In Davis and Revitch, the method had to say who held a phone on a given day, and that’s where it failed. Not every wrong-number case comes out that way: the Revitch order noted that district courts have split on certifying wrong-number classes, and it described an earlier Northern District of California decision that certified one after finding that a reverse number lookup could resolve consent class-wide.

What should a defense expert test?

The defense expert re-runs the plaintiff’s method on real data and reports what it gets wrong. The table maps the usual steps to the tests worth running.

In my view, two tests do most of the work. First, take a random sample of the expert’s output and check it against the company’s own account records. Olean treated results showing false positives as one way expert evidence can fail to produce common answers, though it set no error-rate cutoff, so the size and source of the errors matter. Second, run the method on the named plaintiff. When the method can’t find the person who filed the case, it’s hard to argue it will find anyone else.

Watch the dates as well. The FCC’s safe harbor for Registry compliance requires use of a version of the Registry obtained “no more than 31 days prior to the date any call is made,” under § 64.1200(c)(2)(i)(D), so a scrub against today’s Registry doesn’t show what the caller faced on the call date. For more on proving whether a line was a landline or a cell phone, see our post on prerecorded calls to landlines and their exemptions. For the call records the FTC’s rule already requires sellers to keep, see our post on Telemarketing Sales Rule recordkeeping.

One more tip from the cases: data vendors don’t always vouch for their own products once litigation starts. In Davis, a data vendor’s employee declared that determining definitively whether and when a person used a number “cannot reliably be achieved using the . . . service PacificEast provides,” as the Fourth Circuit quoted. I’d subpoena the vendor early. A vendor’s disclaimer carries the most weight against a method that depends on the database to say who held a number on a given date. In Mantha, by contrast, the court called a vendor’s refusal to warrant its Registry lookup insignificant.

What to secure once a class complaint arrives

1. Put a litigation hold on the call and text data from your dialer and platforms, in native format, along with the consent records for every number at issue.
2. Export opt-out requests and internal do-not-call entries with their dates, and keep the account-level notes that explain them.
3. Save each National Do Not Call Registry scrub file along with the date of the Registry version it used.
4. Gather vendor and lead-source contracts, scripts, recorded or AI-voice message files, and any wireless-scrub or append reports a vendor ran for you.
5. Bring in counsel before answering a demand letter, contacting the plaintiff, or making any settlement offer.

The firm’s California TCPA defense lawyers page explains how it approaches claims involving wrong-number calls and Do Not Call Registry texts.

Questions companies ask about TCPA data experts

Should we hire our own data expert before class certification?

Usually, yes, once the plaintiff discloses an expert. A rebuttal expert can re-run the method and measure its error rate against your records. In Davis, the defense expert’s testing of the 666 numbers the plaintiff’s expert flagged in a sample was among the grounds the district court relied on, and the Fourth Circuit cited the resulting error rate in affirming. Without your own analysis, the court hears only the plaintiff’s expert describe the data.

Is a reverse lookup ever reliable enough to certify a TCPA class?

Sometimes. The Revitch order noted that district courts have split on wrong-number classes. In Mantha, the court accepted a vendor’s business-number screen partly because the expert used it only to narrow the list. Wrong-number classes are harder, because the method must prove the person called wasn’t the customer who consented.

Does Rule 702 even apply at class certification in California federal courts?

It applies, but Ninth Circuit precedent says admissibility alone shouldn’t decide certification. Under Sali and Lytle, courts review the method’s reliability, sometimes in a limited form, and weigh it under Rule 23. In the Northern District of California orders discussed here, one judge found the experts met the amended rule and another declined to read the 2023 note to require more. Plan to win on both rules.

What if the plaintiff’s expert hasn’t finished the analysis by the certification motion?

Make that the centerpiece. Lytle, a damages-model case, allows an unexecuted model only if the expert charts a path to all the necessary data and shows the method will work on the facts of the case. In Davis, the expert’s admission that she never ran most of her method weighed against her.

If a data expert is building the class against your company

If your company is defending a TCPA class action built on a reverse lookup or a Registry match, the expert’s error rate on your own records will often decide whether a class is certified. Rosenberger + Kawabata represents businesses defending TCPA lawsuits and class actions through its business litigation practice, including claims involving wrong-number calls and Do Not Call Registry texts. Contact Rosenberger + Kawabata online for a free and confidential consultation, or call (310) 894-6921.

Sources

The statute is the Telephone Consumer Protection Act at 47 U.S.C. § 227. The FCC’s rules are at 47 C.F.R. § 64.1200, as published in the eCFR through October 1, 2025. Federal Rule of Evidence 702 and its 2023 committee note are at Cornell’s Legal Information Institute, and the effective date is in the Supreme Court’s April 2023 order.

The appellate decisions are Davis v. Capital One, N.A., No. 24-1507 (4th Cir. Aug. 25, 2025) (unpublished); Sali v. Corona Regional Medical Center, 909 F.3d 996 (9th Cir. 2018); Olean Wholesale Grocery Cooperative, Inc. v. Bumble Bee Foods LLC, 31 F.4th 651 (9th Cir. 2022) (en banc); Lytle v. Nutramax Laboratories, Inc., 114 F.4th 1011 (9th Cir. 2024); and Noohi v. Johnson & Johnson Consumer Inc., No. 23-55190 (9th Cir. July 25, 2025).

The district court orders are Davis v. Capital One, N.A., No. 1:22-cv-00903 (E.D. Va. Oct. 20, 2023); Revitch v. Citibank, N.A., No. C 17-06907 WHA (N.D. Cal. Apr. 28, 2019); Mantha v. QuoteWizard.com, LLC, No. 19-12235-LTS (D. Mass. Aug. 16, 2024); Flodin v. Central Garden & Pet Co., No. 4:21-cv-01631-JST (N.D. Cal. Oct. 23, 2024); and Johnson v. Glock, Inc., No. 3:20-cv-08807-WHO (N.D. Cal. Sept. 30, 2024).

This post is general information about the Telephone Consumer Protection Act and related laws as of October 1, 2025. It is not legal advice, and reading it does not create an attorney-client relationship. Courts and the FCC change the rules in this area often, and the outcome of any claim depends on its facts. Past results do not guarantee future outcomes.