In Comcast v. Behrend, 569 U.S. 27 (2013), the Supreme Court held that a plaintiff cannot obtain class certification with an inadequate damages model. In the years since, courts have diverged over how much a plaintiff must do to satisfy this requirement. Often, plaintiffs seek class certification with nothing more than a skeletal proposal to develop and perform an analysis at some future point, using information they do not—and might never—possess. While some courts have found such adumbrative “models” sufficient at the class certification stage, the better decisions require more. As Comcast recognizes, Rule 23 “does not set forth a mere pleading standard.” Rather, a plaintiff “must affirmatively demonstrate” through “evidentiary proof” that damages are measurable on a class-wide basis through a common methodology. Faithful application of that principle obligates plaintiffs and their experts to offer a detailed methodology that is tailored to the facts of the case, and to show that any data that the model requires in fact exists and can be obtained.

The Comcast Decision

Comcast was an antitrust class action involving Comcast cable subscribers in the Philadelphia area. The plaintiffs moved to certify a damages class, asserting four different theories of antitrust injury—i.e., four different theories of how Comcast’s allegedly wrongful conduct affected the prices subscribers paid. Of those four theories, the district court accepted just one as appropriate for class treatment: that Comcast’s activities deterred other cable companies from building competing networks in areas where Comcast already operated. Id. at 30-32.

The plaintiffs’ expert offered a regression model for calculating damages that compared actual cable prices in the Philadelphia area with hypothetical prices that would have prevailed but for Comcast’s anticompetitive activity. But the model did not isolate those damages attributable to the single, relatively narrow type of anticompetitive activity that the district court had accepted for class treatment. The district court certified the class anyway, and a divided Court of Appeals affirmed, rejecting Comcast’s argument that the model improperly measured damages resulting from all four liability theories as an “attack on the merits of the methodology [that] had no place in the class certification inquiry.” Id. at 32 (cleaned up).

In a 5-4 decision, the Supreme Court reversed. Writing for the majority, Justice Scalia began by reiterating that Rule 23 is more than “a mere pleading standard.” It requires the plaintiff to demonstrate, “through evidentiary proof,” that questions common to class members “in fact” predominate over individualized questions. Id. at 33-34 (emphasis in original). Certification is proper only if this proof withstands a “rigorous analysis” that “frequently entail[s] overlap with the merits of the plaintiff’s underlying claim.” Id. at 33-34. The plaintiffs failed to satisfy the predominance requirement, the Court held, because their proffered damages model failed to “identif[y] [only] the damages that are the result of the wrong.” Id. at 37. As such, their model could not “possibly establish that damages [were] susceptible of measurement across the entire class for purposes of Rule 23(b)(3).” Id. at 35. And absent such a showing, the Court found, “[q]uestions of individual damage calculations [would] inevitably overwhelm questions common to the class.” Id. at 34.

Writing jointly for the dissent, Justices Ginsburg and Breyer sought to cabin the majority’s decision, which they deemed “good for this day and case only.” Id. at 42. The dissent asserted that the necessity of individualized damages calculations is not fatal to certification, and that the “predominance requirement is generally satisfied even if damages are not provable in the aggregate” so long as “adjudication of questions of liability common to the class will achieve economies of time and expense.” Id. The dissenters instructed lower courts to read Comcast narrowly, contending that the decision “breaks no new ground” and “should not be read to require, as a prerequisite to certification, that damages attributable to a class-wide injury be measurable ‘on a class-wide basis.’” Id. at 41.

The Lax Approach To Comcast

Despite the dissent’s spin, Comcast’s majority could hardly have been clearer that plaintiffs seeking certification must actually “prove” that individualized questions—including questions relating to damages—do not predominate over common questions. Nonetheless, some federal courts in consumer-product class actions have followed the dissent’s instruction to review plaintiffs’ proffered damages model with minimal rigor. These courts have found it enough for plaintiffs’ experts to promise that damages will be susceptible to class-wide proof using a model that they might construct using data they might obtain.

An oft-cited example is In re Scotts EZ Seed Litigation, 304 F.R.D. 397 (S.D.N.Y. 2015). There, a putative class of consumers alleged that the defendants had misrepresented that their “EZ Seed” product grows grass 50% thicker with half the water compared to ordinary seed. Id. at 403. The plaintiffs sought to recover the price premium they allegedly paid as a result of the claim, and offered an expert who testified that he intended to isolate that premium using one of three statistical methods (he had not yet decided which) and data that he expected would be available down the line. Id. at 413. The court held that these assurances were sufficient because “nothing in Comcast requires an expert to perform his analysis at the class certification stage”—or even to “describe his proposed methodologies in … detail.” Id. at 413-14 & n.11.

The court conducted a similarly relaxed analysis in Goldemberg v. Johnson & Johnson Consumer Cos., 317 F.R.D. 374 (S.D.N.Y. 2016). There, a putative class of consumers alleged that Johnson & Johnson’s Aveeno “Active Naturals” products misleadingly implied that all of their ingredients were natural, as opposed to at least one. Id. at 382-83. The court held that the plaintiffs had satisfied their Comcast burden based on their expert’s assurances that the techniques he had “proposed” would “eventually [be able to] compute the damages associated with” the alleged misrepresentation. Id. at 394 (cleaned up). The defendants’ criticisms of “what the proposal does not yet do”—e.g., that the expert had not even clarified what variables or data he would use—were “unavailing at this juncture.” Id. at 395 n.19.

These decisions are, to say the least, in considerable tension with Comcast. It is true, as these courts are fond of saying, that Comcast does not require plaintiffs to definitively prove the amount of their damages at the class-certification stage. But Comcast certainly does require plaintiffs to submit “evidentiary proof,” sufficient to withstand “rigorous analysis,” that their damages are “in fact” capable of calculation in a manner conducive to class treatment. An expert’s high-level proposal to use a particular modeling technique, coupled with vague assurances that the method will be adaptable to the facts of the particular case and that the necessary data will be available, does not meet this standard.

The Rigorous Approach To Comcast

The better reasoned cases require plaintiffs to come forward with more than boilerplate expert testimony about generally accepted models, and to provide case-specific evidence of workability and effectiveness. See Werdebaugh v. Blue Diamond Growers, 2014 U.S. Dist. LEXIS 173789, at *47 (N.D. Cal. Dec. 15, 2014) (“[T]he Court is obligated to do more than rubberstamp a proposed damages class merely because the plaintiff’s expert [proposes] to … use[] a peer reviewed methodology….”). Cases applying this approach “grant[] that absolute precision is not required [of a damages model] at the class certification stage”—but they nevertheless hold that “real explanation is necessary” and that “vague and abstract” proposals do not suffice. Bruton v. Gerber Prods. Co., 2018 U.S. Dist. LEXIS 30814, at *34 (N.D. Cal. Feb. 13, 2018). At a minimum, these cases require that the expert’s model be fleshed out, with “a clearly defined list of variables”; “a meaningful explanation as to how the variables will be addressed”; and proof that “the data related to … [the] proposed … variables exists.” Id. (citations omitted).

They may also require preliminary results from the proposed model to prove that the technique can actually work on the facts of the particular case. See In re Graphics Processing Units Antitrust Litig., 253 F.R.D. 478, 492 (N.D. Cal. 2008) (noting that courts are “increasingly skeptical of … experts who offer only generalized and theoretical opinions that a particular methodology may serve [their] purpose without also submitting a functioning model that is tailored to market facts in the case at hand”); cf. In re ConAgra Foods, 90 F. Supp. 3d 919, 1031-32 (C.D. Cal. 2015) (granting certification where plaintiff’s expert “ha[d] conducted a preliminary hedonic regression that indicates there is a price premium associated with the [challenged claim]”).

A recent example of this rigorous approach is Ang v. Bimbo Bakeries USA, Inc., where a putative class of consumers claimed that the defendant misleadingly represented that its products were a “good” or “excellent” source of grain and “100% Whole Wheat.” 2018 U.S. Dist. LEXIS 149395, at *2 (N.D. Cal. Aug. 31, 2018). The plaintiffs’ expert proposed to calculate damages using a regression analysis that would isolate the effect of the alleged misrepresentations on price. Id. at *38-39. However, the court observed, the expert had not “plausibly explain[ed] how he will be able to do so,” merely promising that he would perform “subsequent analysis” “once he has the appropriate data (if the class is certified).” Id. at *39-40. This was inadequate to satisfy Comcast, which “does not permit the Court to wait until after certification to assess whether [the proposed model] is capable of measuring economic injury attributable to the alleged misbranding on a classwide basis.” Id. at *40-41.

Opperman v. Kong Technologies offers another example of the “rigorous analysis” that courts should employ post-Comcast. 2017 U.S. Dist. LEXIS 116333 (N.D. Cal. July 25, 2017). There, a putative class of consumers claimed that Apple misled them about the security of personal information on its devices. Id. at *8-9. The plaintiffs proposed conducting a conjoint analysis to determine the market value of the relevant security features. Id. at *38. The court denied class certification because the plaintiffs’ proposal failed to specify the variables to be used in the analysis and, further, failed to isolate the value that consumers attribute to the particular privacy issue in the case as opposed to privacy generally. Id. at *38-44. The expert’s mere “assurance that he c[ould] build a model to calculate damages” tied to plaintiffs’ theory of liability was inadequate. Id. at *39.

Finally, an important appellate-level (but pre-Comcast) example is In re Hydrogen Peroxide Antitrust Litigation, 552 F.3d 305 (3d Cir. 2008). There, the plaintiff’s expert pointed to “two ‘potential approaches’ to estimating damages on a class-wide basis: benchmark analysis, which would compare actual prices during the alleged [antitrust] conspiracy with prices that existed before the class period; and (2) regression analysis, through which it ‘may be possible … to estimate the relationship between price of hydrogen peroxide … and the various market forces that influence prices….’” Id. at 313 (emphasis added). The district court granted certification, concluding that it “did not matter that [the plaintiffs’] expert had not [actually] completed any benchmark or regression analyses” and that the plaintiffs need “not … show at the certification stage that either method would [in fact] work.” Id. at 315. The Third Circuit reversed, holding that the district court had “applied too lenient a standard of proof.” Id. In particular, an expert’s “assurance to the court that it intends or plans to meet [Rule 23’s] requirements is insufficient.” Id. at 319 (emphases added).[1]

Conclusion

The latter approach to class-certification damages models is more faithful to Comcast’s demand for “evidentiary proof” and “rigorous analysis.” Moreover, it ensures fairness to litigants and avoids the waste of time and resources. Granting certification based on generic assurances that a damages model will work runs the risk that damages issues will turn out to predominate once the rubber actually hits the road. In fact, when courts have followed the lax approach described above, they have often been forced to decertify the class once the expert’s model inevitably turns out to fall short of her initial promises. See Herron v. Best Buy Stores LP, 2018 U.S. Dist. LEXIS 70556, at *11-13 (E.D. Cal. Apr. 26, 2018) (citing examples). By the time this comes to pass, the parties will have been forced to incur additional months or years of costly discovery and motion practice—assuming the court’s improvident grant of certification does not force the defendant to settle a case that should never have been certified at all. This is precisely what Comcast’s “rigorous analysis” requirement is intended to prevent. As courts often note, “a district court’s certification order often bestows upon plaintiffs extraordinary leverage, and its bite should dictate the process that precedes it.” Hydrogen Peroxide, 552 F.3d at 320 (quoting Oscar Private Equity Invs. v. Allegiance Telecom, Inc., 487 F.3d 261, 267 (5th Cir. 2007)). Granting class certification based on an expert’s generic promises, on the assumption that any problems will work themselves out, disserves Rule 23 and basic principles of fairness.

[1] See also Weisfeld v. Sun Chem. Corp., 84 F. App’x 257, 261-64 (3d Cir. 2004) (denying certification where expert “state[d] he planned to use … multiple regression” but provided little “independent analysis” or “discussion of the evidence on which [the] analysis was based”); Piggly Wiggly Clarksville, Inc. v. Interstate Brands Corp., 100 F. App’x 296, 299-300 (5th Cir. 2004) (same, where expert “did not offer a formula based on regression analysis, but merely opined that one could be found”); In re Dial Complete Mktg. & Sales Practices Litig., 2015 U.S. Dist. LEXIS 164346, at *100, 106-07 (D.N.H. Dec. 8, 2015) (same, where experts “discuss[ed] how … conjoint analysis works generally,” but “d[id] not provide any specifics”); Miller v. Fuhu Inc., 2015 U.S. Dist. LEXIS 162564, at *64-68 (C.D. Cal. Dec. 1, 2015) (same, where expert proposed to use “Contingent Valuation” or “Conjoint Analysis,” but “the few concrete details [he] provided … [were] too vague for the Court to determine … whether it ‘properly [could] be applied to the facts in issue’ in this case”); Saavedra v. Eli Lilly & Co., 2014 U.S. Dist. LEXIS 179088, at *18-20 (C.D. Cal. Dec. 18, 2014) (same, where expert “ha[d] yet to design the survey … he [would] use in his conjoint analysis” and “ha[d] not decided which attributes will be included”); Randolph v. J.M. Smucker Co., 303 F.R.D. 679, 697-98 (S.D. Fla. 2014) (same, where expert proposed to use “hedonic regression and/or conjoint analysis,” but offered “no hard-and-fast evidence that the [price] premium is capable of measurement,” only “bald, unsupported assertion” that his model would work); Jones v. ConAgra Foods, Inc., 2014 U.S. Dist. LEXIS 81292, at *77-78 (N.D. Cal. June 13, 2014) (same, where expert proposed to use “regression analysis,” but his proposal was “vague and abstract,” did not “provide a clearly defined list of variables,” did not show that “the [relevant] data … exist[ed],” and did not show “how he would determine … which competing and complementary products he would use”); In re ConAgra Foods, Inc., 302 F.R.D. 537, 549-53, 577-78 (C.D. Cal. 2014) (same, where expert “opine[d] that it is possible to determine damages” using hedonic regression and conjoint analysis, but “d[id] not actually perform either analysis or describe in any detail their specific application to this case”); Kottaras v. Whole Foods Mkt., Inc., 281 F.R.D. 16, 25-26 (D.D.C. 2012) (same, where expert stated he “plan[ned] to run a regression analysis … but his proposal [was] tentative” and “too vague for the Court to even evaluate”; rejecting expert’s unsupported “assurance” that “merits discovery [would] further refine [his] assessment”); Weiner v. Snapple Beverage Corp., 2010 U.S. Dist. LEXIS 79647 (S.D.N.Y. Aug. 5, 2010) (same; recognizing that a plaintiff’s expert need not fully “implement … his methodology at the class certification stage,” but that merely “identify[ing] … possible approaches and assert[ing] that they will work” is insufficient).