Proving Retail Sales Figures In Consumer Class Actions: Different Approaches Lead To Very Different Results
To prove damages in a consumer class action, the named plaintiff must show—among other things—how many units of the defendant’s product were purchased by consumers in the relevant state (or states). This is easier said than done. Manufacturers generally keep records of their own wholesale transactions—i.e., how much product they shipped to distributors or large retail chains. But they generally don’t have direct visibility into sales at the retail level, since they aren’t a party to those transactions. If not all of the product sold at wholesale ends up being purchased by consumers, manufacturers’ records may not reflect this. Likewise, if the product that a manufacturer ships to an address in State A (e.g., a regional distribution center) ends up being moved to State B before reaching store shelves, manufacturers’ records will not reflect this either. What, then, is a class-action plaintiff to do?
Two recent bench rulings in the Southern District of New York illustrate this problem and the different approaches that plaintiffs have taken to prove state-specific retail sales. In the first case, the court deemed the approach of the plaintiff’s expert fundamentally flawed, resulting in decertification on the eve of trial. The second case involved the very same expert; this time, however, the court concluded that any methodological errors were merely fodder for cross-examination and allowed the trial to proceed. In this post, we explore these contrasting rulings and attempt to discern what lessons prospective class-action defendants should learn from them.
In Re Amla Litigation, 16-cv-6593 (S.D.N.Y. Jan 31, 2019)
In Re Amla Litigation (“Amla”) concerned allegations that the defendant’s hair relaxer caused harm to users’ hair and scalp. The plaintiffs—represented by celebrity lawyer Mark Geragos—obtained certification of a class of New York purchasers and survived motions for summary judgment and decertification. The case was heading to trial when the defendant moved to exclude the plaintiffs’ damages expert.
To establish the defendant’s New York retail sales, the Amla plaintiffs had subpoenaed Information Resources, Inc. (“IRI”), a data clearinghouse that generates estimated nationwide and statewide retail sales reports for consumer products in certain channels of trade. (Nielsen is another leading provider of these types of reports.) Of course, IRI does not actually monitor every retail transaction at every store in the United States. Rather, it obtains cash-register-scan data from a subset of retail stores or groups of stores and then uses a proprietary methodology to extrapolate statewide and nationwide totals from that sample. IRI then sells reports based on those extrapolations to manufacturers, researchers, and other interested parties.
In response to the subpoena, IRI produced a report ostensibly reflecting the total New York retail sales of the hair product at issue during the class period. The plaintiffs then provided that report to Colin Weir, their expert (and a perennial favorite of the class-action plaintiffs’ bar). Weir, in turn, drafted a report that took the number of units sold in New York during the class period, as stated in the IRI report, and multiplied that number by $50, the per-violation statutory damage award purportedly available under New York’s consumer protection statute. Et voilà—a class-wide damages model!
Notably, there were several things that the plaintiffs did not do. They did not take any discovery from IRI to probe how its underlying data is collected or what proprietary calculations it performs to generate its statewide extrapolations. They did not rely on any data from individual retailers. And although the defendant had produced its own wholesale sales records earlier in the litigation, Weir did not incorporate any of that data into his damages analysis.
The defendant moved to exclude Weir’s testimony on the ground that he was merely acting as a conduit for the admission of “unauthenticated, multiple-hearsay projected sales estimates modeled from unknown data via [IRI’s] inscrutably secret methodologies.” As defense counsel put it at oral argument:
So what plaintiffs are trying to do, your Honor, … [is] to launder or clean this … hopelessly inadmissible evidence, let Weir click around on a spread sheet [from IRI], which we all can do, let him push the plus sign, pay him $700 an hour, and then suddenly he has done this voodoo over it, and the spreadsheet is magically admissible?
The plaintiffs defended Weir’s reliance on IRI data on the ground that “Fortune 500 companies use it” in their day-to-day business. They also invoked the maxims that damages need not be quantified with certainty and that “the burden of uncertainty as to the amount of damages is on the wrongdoer.”
Judge Jed Rakoff ruled from the bench that Weir’s testimony must be excluded. For one thing, Weir’s report “involve[d], frankly, very modest expertise”—he had “basically [served as] a human adding machine,” totaling up the figures in IRI’s spreadsheets and multiplying the result by $50. But what Judge Rakoff found most “troubl[ing]” was Weir’s “total absence of … knowledge” concerning IRI’s data sources and extrapolation methods. The “whole thrust of Daubert,” he observed, is that experts can’t “get away with a damages calculation that doesn’t expose or even purport to expose its basic methodology.” And since Weir’s inadmissible testimony was the only evidence that the plaintiffs had adduced on the issue of class-wide damages, Judge Rakoff decertified the class on the threshold of trial.
Mr. Geragos made one last Hail Mary attempt to change the Court’s mind, suggesting that even if Weir’s reliance on IRI data was improper, the plaintiffs could ask the jury to make an inference about the defendant’s New York retail sales based on the wholesale sales records that it had previously produced. Judge Rakoff nixed that proposal, however, noting that none of this was in Weir’s report, and that in our judicial system, “we don’t have trial by ambush.”
Hart v. BHH, LLC, 15-cv-4804 (S.D.N.Y. Apr. 1, 2019)
Things turned out differently in Hart v. BHH, LLC (“Hart”), which concerned allegations that the defendant’s pest-repeller device was ineffective. The plaintiffs had survived a number of dispositive motions and were heading to trial with a certified nationwide class on a fraud claim and a certified multi-state class (including purchasers from 25 states) on a warranty claim. Just as in Amla, the plaintiffs had retained Colin Weir to opine on class-wide damages at trial.
In Hart, however, Weir took a different tack. Instead of relying on third-party data from IRI or Nielsen, Weir did what Mr. Geragos had belatedly proposed to do, and used the defendant’s nationwide wholesale sales records as a “proxy” for the number of nationwide sales of its products at retail. Then, to calculate the portion of those sales that took place in the 25 states relevant to the warranty class, Weir relied on retail sales data subpoenaed from an apparently arbitrary set of seven chain stores, using the proportion of their sales that occurred in the relevant 25 states to extrapolate the proportion of all retail sales that occurred in those states.
Before trial, the defendant moved to exclude Weir. Among other things, it argued that Weir had “identifie[d] no scientifically valid basis for treating wholesale sales figures and retail sales figures as interchangeable,” given that retailers do not necessarily sell through to consumers all the stock that manufacturers sell at wholesale. In addition, the defendant argued that Weir “offer[ed] no reliable methodology regarding his allocation of what percentage of … nationwide sales occurred in each [relevant] state,” since he made no attempt “to show that the [seven-retailer sample] from which he extrapolate[d] his estimates of state-level sales [was] representative” of retailers overall. The plaintiffs opposed the motion, arguing that Weir had made a reasonable estimate based on the best available data and that nothing more was required.
On April 1, 2019, Judge William H. Pauley III denied the defendants’ motion from the bench. In a brief three-sentence explanation, he found that “Defendants’ criticisms go to the weight” of Weir’s testimony, “not its admissibility,” and that reliance on wholesale rather than retail data was a “minor error” that “should be tested by the adversary process.” Judge Pauley did not explain in his ruling why he believed the error was a “minor” one. From his statements during oral argument, however, it appears that he found insufficient record evidence of practices such as inventory stockpiling or channel stuffing “where units are just sitting parked in places [along the supply chain] and ultimately never get to the retail public.” Judge Pauley did not address the defendant’s objections to Weir’s state-by-state allocation methodology, either during oral argument or in his bench ruling.
Lessons from Amla and Hart
Taken together, Amla and Hart might be read to suggest that state-specific retail sales reports from a clearinghouse such as IRI or Nielsen are inadequate inputs for a class-action damages model, while a company’s own wholesale transaction records could be adequate inputs—or at least sufficient to reach a jury. Going forward, savvy plaintiffs will be unlikely to rely exclusively on IRI or Nielsen retail reports, and will presumably offer some variation on Weir’s “proxy” methodology to prove retail sales at trial.
Importantly, however, Hart is unlikely to be the last word on the admissibility of such “proxy” models. Some judges may be less willing to abide damages experts’ unexplained assumptions and refusals to answer methodological questions. (As the Hart defendants noted, the plaintiffs’ counsel instructed Weir not to answer questions on the specifics of his methodology “more than 60 times” at deposition.) And other cases may be distinguishable from Hart on their facts. Defendants may be able to provide case-specific reasons why wholesale data is a poor “proxy” for retail data, lifting this mismatch out of the realm of “minor error” and into the realm of fundamental unreliability. For example, there may be evidence of inventory stockpiling, channel stuffing, product recalls, or other supply chain phenomena absent in Hart. Defendants may also be able to make a more compelling case-specific demonstration that the plaintiff’s expert used unrepresentative data to allocate nationwide sales across relevant states.
Smart defendants will keep these issues in mind from the outset and start developing the record early on, so that they will be positioned to attack the plaintiff’s “proxy” model with concrete, case-specific evidence when the time comes.
 See In re AMLA Litig., 282 F. Supp. 3d 571 (S.D.N.Y. 2017) (certifying class of New York purchasers); In re AMLA Litig., 320 F. Supp. 578 (S.D.N.Y. 2018) (denying motion for summary judgment); In re AMLA Litig., 328 F.R.D. 127 (S.D.N.Y. 2018) (denying motion for decertification).
 Motion to Exclude Weir, In Re Amla Litig., No. 16-cv-6593, Dkt. No. 258 (S.D.N.Y. Jan 8, 2019).
 H’g on Motion to Exclude, In Re Amla Litig., No. 16-cv-6593 (S.D.N.Y. Jan 22, 2019).
 See Hart v. BHH, LLC, 2017 U.S. Dist. LEXIS 105187 (S.D.N.Y. July 7, 2017); Hart v. BHH, LLC, 323 F. Supp. 3d 560 (S.D.N.Y. 2018).
 Motion to Exclude Weir, Hart v. BHH, LLC, 15-cv-4804, Dkt. No. 220 (S.D.N.Y. Mar. 1, 2019).
 H’g on Mot. to Exclude at 4-5, 103, Hart v. BHH, LLC, 15-cv-4804 (S.D.N.Y. Apr. 1, 2019).