ORDER GRANTING IN PART AND DENYING IN PART DEFENDANT'S MOTION TO DECERTIFY

LUCY H. KOH, District Judge.

Before the Court is a Motion to Decertify brought by Defendant Dole Packaged Foods, LLC ("Dole"), a subsidiary of Dole Food Company, Inc. ECF No. 171 ("Mot."). Plaintiff Chad Brazil ("Brazil") opposes the motion, ECF No. 177 ("Opp'n"), and Dole has replied, ECF No. 202 ("Reply"). Having considered the submissions of the parties, the relevant law, and the record in this case, the Court hereby GRANTS in part and DENIES in part Dole's Motion to Decertify.

I. BACKGROUND

A. Factual Background

Dole, a limited liability corporation with principal place of business in Westlake Village, California, "is a leading producer of retail food products" that sells "to consumers through grocery and other retail stores throughout the United States." ECF No. 148, Third Amended Complaint ("TAC") ¶¶ 16-17.

Brazil is a California consumer who "cares about the nutritional content of food and seeks to maintain a healthy diet." TAC ¶¶ 15, 76. "All of the misconduct alleged [in the TAC]," Brazil claims, "was contrived in, implemented in, and has a shared nexus with California." Id. ¶ 18. From April 11, 2008, to the present, Brazil has spent over $25.00 on Dole's food products, which he contends are "misbranded" in violation of federal and state law. Id. ¶¶ 1, 5, 76. In particular, Brazil alleges that he purchased the following three food products:

Id. ¶ 2. Brazil refers to these products collectively as the "Purchased Products." Id.

The TAC also alleges claims based on seven additional products that Brazil did not purchase, but which are, according to Brazil, substantially similar to those that he did in that they "(i) make the same label representations . . . as the Purchased Products and (ii) violate the same regulations." They include:

Id. ¶ 4. Brazil refers to this second group of products as the "Substantially Similar Products." Id. ¶ 3.

Brazil alleges that Dole makes numerous representations on the labels of these ten products that are unlawful, as well as false and misleading, under federal and California law. TAC ¶¶ 8-14. According to Brazil, regulations issued by the Food and Drug Administration ("FDA") dictate that Dole may not claim that a product is "all natural" if it contains "unnatural ingredients such as added color, [or] synthetic and artificial substances." Id. ¶ 30; see also 21 C.F.R. § 101.22 (setting forth the circumstances under which added colors and artificial flavors must be disclosed on a package's label). Dole's products are mislabeled, Brazil alleges, because they "contain artificial ingredients and flavorings, artificial coloring and chemical preservatives," precluding Dole's use of the term "natural." TAC ¶¶ 36-39. Specifically, Brazil contends that all ten of the products listed above contain the label statement "All Natural Fruit," which Brazil alleges is misleading because all ten products contain ascorbic acid (commonly known as Vitamin C) and citric acid, both allegedly synthetic ingredients. ECF No. 142, Order Granting in Part and Denying in Part Brazil's Motion for Class Certification ("Class Cert Order") at 3.

B. Procedural History

Brazil filed his original Complaint on April 11, 2012, naming both Dole Packaged Foods, LLC, and Dole Food Company, Inc., as Defendants. ECF No. 1. Brazil brought claims under, inter alia, California's Unfair Competition Law ("UCL"), Cal. Bus. & Prof. Code § 17200 et seq.; Fair Advertising Law ("FAL"), id. § 17500 et seq.; and Consumers Legal Remedies Act ("CLRA"), Cal. Civ. Code § 1750 et seq. Defendants filed a Motion to Dismiss on July 2, 2012. ECF No. 16. Rather than responding to the motion, Brazil filed a First Amended Complaint ("FAC") on July 23, 2012. ECF No. 25. The Court then denied Defendants' Motion to Dismiss as moot. ECF No. 28.

On August 13, 2012, Defendants filed a Motion to Dismiss the FAC. ECF No. 29. Brazil opposed the motion on August 31, 2012, ECF No. 35, and Defendants replied on September 14, 2012, ECF No. 37. After holding a motion hearing on January 24, 2013, ECF No. 53, the Court granted in part and denied in part the Motion to Dismiss on March 25, 2013, ECF No. 59. The Court granted leave to amend, and Brazil filed a Second Amended Complaint ("SAC") on April 12, 2013. ECF No. 60. In response to the SAC, Defendants again filed a Motion to Dismiss on April 29, 2013. ECF No. 62. Brazil opposed the motion on May 20, 2013, ECF No. 68, and Defendants replied on June 3, 2013, ECF No. 71. The Court granted in part and denied in part the Motion to Dismiss on September 23, 2013. ECF No. 76. The parties then stipulated to the dismissal of several products that Brazil testified he had never purchased. ECF No. 88 at 2. In addition, the stipulation dismissed Defendant Dole Food Company, Inc., from the case. Id.

Brazil filed a Motion for Class Certification on January 31, 2014. ECF No. 96. Dole opposed the motion on March 6, 2014, ECF No. 105, and Brazil replied on March 27, 2014, ECF No. 117. The Court granted in part and denied in part the Motion for Class Certification on May 30, 2014. Class Cert Order at 1. In its May 30 order, the Court certified two classes. Id. at 35. The first, certified under Rule 23(b)(2) of the Federal Rules of Civil Procedure (the "Injunction Class"), includes:

Id. The second, certified under Rule 23(b)(3) (the "Damages Class"), includes:

Id.

With respect to the Damages Class, Brazil's damages expert, Dr. Oral Capps, advanced three models for measuring "the price premium attributable to Dole's use of the `All Natural Fruit' label statements." Class Cert Order at 26. Of those three, the Court accepted only the model based on "econometric or regression analysis" (the "Regression Model"). Id. at 29. The Court concluded that the Regression Model, as presented to the Court then, "provide[d] a means of showing damages on a classwide basis through common proof," thus "satisf[ying] the Rule 23(b)(3) requirement that common issues predominate over individual ones." Id. at 34. In reaching this conclusion, the Court rejected Dole's argument that class certification should be denied "because Dr. Capps ha[d] not yet run his regressions." Id. at 31. The Court did so because Dole's failure to provide the necessary discovery was the reason Dr. Capps had been unable to run his regressions. See id. at 31-32. Declining to allow Dole to "use damages discovery as both a sword and a shield," id. at 31, the Court certified the proposed Damages Class even though Dr. Capps had yet to conduct the regression analysis.1

On June 6, 2014, Dole moved for reconsideration, ECF No. 145, and on June 12, 2014, Brazil filed his TAC, ECF No. 148. The Court denied Dole reconsideration on June 16, 2014. ECF No. 150 ("Order Denying Recons.") at 7. In its order doing so, the Court found that "although Dole raises potentially legitimate concerns about Brazil's ability to prove damages," Dole's motion was premature because "Dr. Capps has yet to serve his expert report." Id. at 6. If at the close of expert discovery "it turns out that Dr. Capps's Regression Model cannot adequately assess damages on a classwide basis," Dole, the Court concluded, would "have a basis to move for decertification." Id.

On August 21, 2014, Dole filed the instant Motion to Decertify. Mot. at 18.2 In the alternative, Dole requested a "non-testimonial `Daubert' review of the reliability and relevance of Dr. Capps' testimony." Id. at 17-18. Brazil filed his Opposition on September 11, 2014. Opp'n at 19. Dole filed a Reply on September 25, 2014. Reply at 11.

II. LEGAL STANDARD

"Even after a certification order is entered, the [Court] remains free to modify it in the light of subsequent developments in the litigation." Gen. Tel. Co. of Sw. v. Falcon, 457 U.S. 147, 160 (1982); see also Fed. R. Civ. P. 23(c)(1)(C) ("An order that grants or denies class certification may be altered or amended before final judgment."). "The standard used by the courts in reviewing a motion to decertify is the same as the standard when it considered Plaintiffs' certification motions." Ries v. Ariz. Beverages USA LLC, No. 10-01139 RS, 2013 WL 1287416, at *3 (N.D. Cal. Mar. 28, 2013). "On a motion for decertification, the burden remains on the plaintiffs to demonstrate `that the requirements of Rules 23(a) and (b) are met.'" Id. (quoting Marlo v. United Parcel Serv., Inc., 639 F.3d 942, 947 (9th Cir. 2011)); see also Negrete v. Allianz Life Ins. Co. of N. Am., 287 F.R.D. 590, 598 n.1 (C.D. Cal. 2012) ("To the extent that pre-Marlo cases conclude that a defendant bears the burden on a motion to decertify of demonstrating that `the elements [of] Rule 23 have not been established,' Gonzales v. Arrow Fin. Servs. LLC, 489 F.Supp.2d 1140[, 1153] (S.D. Cal. 2007), these cases are no longer good law."). The burden, therefore, remains with Brazil.

Federal Rule of Civil Procedure 23, which governs class certification, has two sets of distinct requirements that Brazil must establish before the Court may certify a class. Brazil must satisfy all of the requirements of Rule 23(a) and at least one of the prongs of Rule 23(b).

Under Rule 23(a), the Court may certify a class only where "(1) the class is so numerous that joinder of all members is impracticable; (2) there are questions of law or fact common to the class; (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class; and (4) the representative parties will fairly and adequately protect the interests of the class." Fed. R. Civ. P. 23(a). Courts refer to these four requirements, which all must be satisfied to maintain a class action, as "numerosity, commonality, typicality and adequacy of representation." Mazza v. Am. Honda Motor Co., 666 F.3d 581, 588 (9th Cir. 2012). Further, courts have implied an additional requirement: the class to be certified must be "ascertainable." Bruton v. Gerber Prods. Co., No. 12-CV-02412-LHK, 2014 WL 2860995, at *3 (N.D. Cal. June 23, 2014) (citing Marcus v. BMW of N. Am., LLC, 687 F.3d 583, 592-93 (3d Cir. 2012)).

In addition to meeting the requirements of Rule 23(a), the Court must also find that Brazil has satisfied "through evidentiary proof" one of the three subsections of Rule 23(b). Comcast Corp. v. Behrend, 133 S.Ct. 1426, 1432 (2013). As relevant here, the Court may certify a Rule 23(b)(2) class if "the party opposing the class has acted or refused to act on grounds that apply generally to the class, so that final injunctive relief or corresponding declaratory relief is appropriate respecting the class as a whole." Fed. R. Civ. P. 23(b)(2). The Court may certify a Rule 23(b)(3) class if it finds that "questions of law or fact common to class members predominate over any questions affecting only individual members, and that a class action is superior to other available methods for fairly and efficiently adjudicating the controversy." Id. 23(b)(3).

On the whole, "a court's class-certification analysis must be `rigorous' and may `entail some overlap with the merits of the plaintiff's underlying claim.'" Amgen Inc. v. Conn. Ret. Plans & Trust Funds, 133 S.Ct. 1184, 1194 (2013) (quoting Wal-Mart Stores, Inc. v. Dukes, 131 S.Ct. 2541, 2551 (2011)); see also Mazza, 666 F.3d at 588 ("Before certifying a class, the trial court must conduct a `rigorous analysis' to determine whether the party seeking certification has met the prerequisites of Rule 23." (internal quotation marks omitted)). Nevertheless, "Rule 23 grants courts no license to engage in free-ranging merits inquiries at the certification stage." Amgen, 133 S. Ct. at 1194-95. "Merits questions may be considered to the extent—but only to the extent—that they are relevant to determining whether the Rule 23 prerequisites for class certification are satisfied." Id. at 1195. Within the framework of Rule 23, the Court ultimately has broad discretion over whether to certify or decertify a class. United Steel Workers Int'l Union v. ConocoPhillips Co., 593 F.3d 802, 809-10 (9th Cir. 2010).

III. DISCUSSION

In its Motion to Decertify, Dole makes two chief contentions. First, Dole argues that the Damages Class certified under Rule 23(b)(3) should be decertified because Dr. Capps' Regression Model is fundamentally flawed, rendering it incapable of measuring only those damages attributable to Dole's alleged misbranding. Mot. at 6-17. Second, Dole contends that the Damages Class, as well as the Injunction Class certified under Rule 23(b)(2), should be decertified because neither is ascertainable. Id. at 18. The Court addresses each argument in turn.3

A. Rule 23(b)(3) Predominance

For a class action to be certified under Rule 23(b)(3), the class representative must show that "questions of law or fact common to the members of the class predominate over any questions affecting only individual members and that a class action is superior to other available methods for the fair and efficient adjudication of the controversy." Fed. R. Civ. P. 23(b)(3). Only the predominance requirement is at issue here.

"The Rule 23(b)(3) predominance inquiry" is meant to "test[] whether proposed classes are sufficiently cohesive to warrant adjudication by representation." Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 623 (1997). "When common questions present a significant aspect of the case and they can be resolved for all members of the class in a single adjudication, there is clear justification for handling the dispute on a representative rather than on an individual basis." Hanlon v. Chrysler Corp., 150 F.3d 1011, 1022 (9th Cir. 1998) (internal quotation marks omitted).

To satisfy the Rule 23(b)(3) predominance requirement, Brazil must present a damages model that is "consistent with [his] liability case." Comcast, 133 S. Ct. at 1433 (internal quotation marks omitted). More specifically, Brazil's regression "model purporting to serve as evidence of damages in this class action must measure only those damages attributable to [Dole's conduct]." Id. At this stage, however, "[c]alculations need not be exact." Id.

The type of damages that Brazil's model seeks to prove is restitution, a remedy whose purpose is "to restore the status quo by returning to the plaintiff funds in which he or she has an ownership interest." Kor. Supply Co. v. Lockheed Martin Corp., 29 Cal.4th 1134, 1149 (2003). The UCL, FAL, and CLRA authorize trial courts to grant restitution to private litigants asserting claims under those statutes. See Colgan v. Leatherman Tool Grp., Inc., 135 Cal.App.4th 663, 694 (2006). The proper measure of restitution in a mislabeling case is the amount necessary to compensate the purchaser for the difference between a product as labeled and the product as received. See id. at 700 (rejecting restitutionary award for products "Made in U.S.A." where expert "did not attempt to quantify either the dollar value of the consumer impact or the advantage realized by [the defendant]"). This calculation contemplates the production of evidence that attaches a dollar value to the "consumer impact" caused by the unlawful business practices. Id. Restitution is then determined by taking the difference between the market price actually paid by consumers and the true market price that reflects the impact of the unlawful, unfair, or fraudulent business practices. Werdebaugh v. Blue Diamond Growers, No. 12-CV-2724-LHK, 2014 WL 2191901, at *22 (N.D. Cal. May 23, 2014). Accordingly, Brazil must present a damages methodology that can determine the price premium attributable to Dole's use of the "All Natural Fruit" label statements.

1. The Regression Model

Dr. Capps' Regression Model, which the Court preliminarily approved in its May 30 order, see Class Cert Order at 29-34, purports to determine the price premium attributable to Dole's use of the "All Natural Fruit" label on its products. "Regression analysis involves the quantification of the relationship between a variable to be explained, known as the dependent variable and additional variables that are thought to produce or to be associated with the dependent variable, known as the explanatory or independent variables." ECF No. 178-10, Expert Report of Dr. Oral Capps, Jr. ("Capps Report") ¶ 6. "Variables," Dr. Capps explained, "often are quantitative measures like price, volume sold, or dollar sales," but they "also may be qualitative in nature to represent events such as seasonal fluctuations, year-to-year fluctuations, or presence or absence of a labeling claim." Id. The goal of regression analysis is "to isolate whether a particular relationship exists between the dependent and independent variables and for measuring the magnitude of this relationship while controlling for other factors that could also influence the dependent variable." Id.

As Brazil acknowledges, Opp'n at 2, Dr. Capps initially proposed using regression analysis to measure "the portion of sales" attributable to Dole's use of the "All Natural Fruit" label, ECF No. 101-9, Declaration of Dr. Oral Capps, Jr. ("Capps Decl.") ¶ 18. To do so, Dr. Capps anticipated that he would "examin[e] sales of the identified products before and after Dole placed the alleged misrepresentations on its product labels, using regression analysis to control for other variables that could otherwise explain changes in Dole's sales." Class Cert Order at 29; see Capps Decl. ¶¶ 21-22. In granting class certification, the Court relied on Dr. Capps' representation that he would be conducting this type of before-and-after analysis once Dole provided him the necessary discovery. See Class Cert Order at 30 ("[S]ignificantly, the Regression Model compares data on identical Dole products: the product before the label statement was introduced, and the same product after its label included the alleged misrepresentation.").

Dr. Capps was unable to do so, however. As it turned out, discovery revealed that the labels for nine of the ten4 products certified did not actually change during the class period. Order Denying Recons. at 5; ECF No. 177-3, Deposition of Dr. Oral Capps, Jr. ("Capps Dep.") at 33-34, 40-41. This fact, Dr. Capps assured the Court, did not doom the Regression Model as a means for determining damages. "[I]t is nevertheless possible to use econometric or regression analysis," he claimed. Capps Report ¶ 7. But Dr. Capps had to change his methodology as a result. Capps Dep. at 7-8. He instead employed a type of regression methodology known as "hedonic price analysis" or "hedonic regression." Capps Report ¶ 8. Dr. Capps explained hedonic regression in his expert report:

Id. By "controlling for other attributes that may impact product prices in the regression analysis," Dr. Capps affirmed that he could "ascertain the impact of the ["All Natural Fruit"] labeling claim on prices of [Dole] food products." Id.

What Dr. Capps did in this case may, at a general level, be characterized as a two-step process. See Opp'n at 3. First, he conducted a hedonic regression analysis to determine whether the label statement "All Natural Fruit" had any statistically significant impact on the prices of the Dole products certified by the Court. Capps Report ¶¶ 19-26. For each product, the analysis produced a coefficient, or multiplier, purporting to represent the percentage change in the product's price attributable to the statement "All Natural Fruit," trying to control for all other factors. Id. ¶¶ 27-28. Second, Dr. Capps multiplied the coefficient for each product by the total retail dollar sales in California associated with that product during the relevant time period and added the resulting values together. Id. ¶¶ 29-32. By doing so, Dr. Capps purported to estimate the total value of restitution (or damages) for all of the challenged products during the class period.5 Id. ¶ 34.

To conduct the hedonic regression analysis, Dr. Capps collected data from Information Resources, Inc. ("IRI"), which lists weekly retail sales information by brand and by Universal Product Code ("UPC"). Capps Report ¶ 15. From IRI, Dr. Capps was able to obtain 280 weeks' worth of data on Dole's retail sales for the challenged products, dating from January 2009 to mid-May 2014.6 Id. ¶¶ 15-16. This methodology is known as a "revealed preference approach" because it relies on data from actual retail transactions. Id. ¶ 19.

Using the IRI data, Dr. Capps performed a hedonic regression analysis using the following variables: (1) price, the "dependent variable"; and (2) a set of "explanatory variables," namely package size, seasonality, year, brand, and the presence or absence of the labeling claim "All Natural Fruit." Capps Report ¶ 21. Inflation was also taken into account by adjusting the dependent variable, which was expressed in logarithmic form. Id.; Capps Dep. at 57-58. Further, the dependent variable consisted of the prices of the nine challenged Dole products as well as the prices of comparator products,7 some of which were identified based on the declaration of David Spare, Dole's vice president of marketing. Capps Report ¶ 22; see also ECF No. 107, Declaration of David Spare ("Spare Decl.") ¶¶ 5-10.

Dr. Capps then ran the regressions for each of the nine products under two scenarios. Capps Report ¶ 25. In Scenario 1, Dr. Capps assumed that both Dole and Del Monte products made "All Natural" claims on their labels. Id. In Scenario 2, Dr. Capps assumed that only Dole products made the claim. Id. Dr. Capps modeled these two scenarios because he "wasn't certain" that the Del Monte products he had found in the IRI dataset actually made the "All Natural" claim on their labels. Capps Dep. at 26. Each of the hedonic regressions associated with each scenario generated estimated coefficients that were, according to Dr. Capps, "economically and statistically significant." Capps Report ¶ 27. Under each scenario, the coefficients associated with each product were then multiplied by the total retail dollar sales in California for that product during the relevant time period and added together to derive the estimated total restitution (or damages) attributable to the labeling claim, controlling for all other factors. Id. ¶ 32. In the end, Dr. Capps estimated the total restitution (or damages) to be [REDACTED] under Scenario 1 and [REDACTED] under Scenario 2. Id. ¶ 25.

Dr. Capps also presented a third damages scenario ("Scenario 3"), wherein he lifted the hedonic regression coefficient from a 2007 study by Prof. Jeffrey Anstine seeking to determine the price premium traceable to an "All Natural" label claim on yogurt. Capps Report ¶¶ 9, 25 (citing Jeffrey Anstine, Organic and All Natural: Do Consumers Know the Difference?, 26 J. Applied Econ. & Pol'y 15 (2007)). Using that coefficient, Dr. Capps derived a [REDACTED] estimate for restitution (or damages). Id. ¶ 25.

2. The Model's Alleged Flaws

Dole argues that the Regression Model is fatally flawed, and that the Court should decertify the Damages Class as a result. Mot. at 6-17. In particular, Dole identifies six flaws, each of which warrants decertification in Dole's view: (1) the Court approved a "sales" regression but Dr. Capps performed a "price" regression; (2) the Regression Model confuses "brand" and "label"; (3) the Regression Model improperly uses retail-level data; (4) the Regression Model does not control for other variables; (5) the Regression Model is full of data errors; and (6) the Regression Model's Scenario 3 fails Comcast. The Court addresses each herein. For the reasons stated below, the Court finds that Dr. Capps' Regression Model does not sufficiently isolate the price impact of Dole's use of the "All Natural Fruit" labeling statements. Dr. Capps' model therefore fails under Comcast to adequately tie damages to Dole's supposed misconduct.

a. Alleged Flaw #1: Dr. Capps Performed a "Price" Regression

The first flaw Dole advances is that Dr. Capps performed the wrong regression. Mot. at 6-8. Dole contends that Dr. Capps promised the Court he would perform a before-and-after "sales" regression, but instead delivered a very different "price" regression. Id. at 6-7. Instead of answering the question the Court approved, "What happened to Dole's sales when the labels changed?" Dr. Capps answered "How much more would a retailer charge if a label said `All Natural Fruit?'" Id. at 7. Dr. Capps, argues Dole, "measured the wrong thing." Id.

Dole is mistaken. What the Court requested was "a damages methodology that can determine the price premium attributable to Dole's use of the `All Natural Fruit' label statements." Class Cert Order at 26. This was in fact the question that the Regression Model sought to answer. See Capps Report ¶ 18 ("The purpose of my analysis is to calculate the value consumers place on the labeling statement `All Natural Fruit.'"); Capps Dep. at 82 ("The object was using regression, can you ascertain a premium attached to the presence of the labeling claim, and that's what is demonstrated here."). It is true that Dr. Capps, prior to obtaining the necessary discovery from Dole, initially proposed a regression analysis that differed from the one he ultimately performed. However, so long as Dr. Capps' analysis is capable of isolating the impact on price traceable to Dole's labeling claim, whether he does so via the precise methodology he anticipated using months before the close of expert discovery is of no moment.

The Court requested a damages model capable of estimating the "price premium" traceable to Dole's "All Natural Fruit" labeling statement. Class Cert Order at 26 (emphasis added). Although Dr. Capps could have looked at changes in Dole's sales had the labels actually changed during the class period, it was equally proper for him to examine how much consumers overpaid because of Dole's alleged misrepresentations. See id. ("Restitution can then be determined by taking the difference between the market price actually paid by consumers and the true market price that reflects the impact of the unlawful, unfair, or fraudulent business practices."); see also Kor. Supply, 29 Cal. 4th at 1149 ("The object of restitution is to restore the status quo by returning to the plaintiff funds in which he or she has an ownership interest."). Dr. Capps' model, which fails for other reasons explained in Part III.A.2.d, infra, at least purported to do just that, see Capps Dep. at 51 (finding a "17 percent overcharge in price . . . attributed to the labeling claim `all natural'" in Scenario 1 based on actual sales of Dole products).8

As a result, the Court will not decertify because Dr. Capps conducted a hedonic regression focusing on price.

b. Alleged Flaw #2: The Model Confuses "Brand" and "Label"

The second flaw Dole advances is that the Regression Model confounds the "brand" and "label" variables. Mot. at 8-11. Dole contends that Dr. Capps' model, in the abstract, assumes there is no variation in label status within any of the brands. Id. at 9-10. As explained in Part III.A.2.d, infra, the Court agrees with Dole that Dr. Capps made this assumption and that Dr. Capps did so without verifying whether the non-Dole products he tested actually made an "All Natural" label claim. Putting that aside for the moment, Dole argues that Dr. Capps' model fails at a more general level because, based on the model's assumptions, if one knows the brand, one should also know with certainty whether or not the label is present. Mot. at 9-10; See ECF No. 172-3, Rebuttal Expert Report of Professor Carol A. Scott ("Scott Rebuttal Report") ¶ 32. This phenomenon, called perfect collinearity, would ordinarily diminish a model's predictive power because the regression would be incapable of teasing apart the impact of "brand" from "label." Mot. at 9; Scott Rebuttal Report ¶¶ 28, 32; see also In re High-Tech Emp. Antitrust Litig., No. 11-CV-02509-LHK, 2014 WL 1351040, at *21 n.51 (N.D. Cal. Apr. 4, 2014) (explaining that "collinearity may cause regression estimates to become[] less precise").

To escape the collinearity problem, Dole argues that Dr. Capps engaged in "intellectually dishonest" legerdemain by manipulating the "brand" variable. Mot. at 10-11; see Reply at 4-5. In Scenario 1, for example, Dr. Capps assigned Dole products a value of "1" while all other brands (e.g., Del Monte, regional brands, or private labels) were coded as "0" for purposes of the "brand" variable. Scott Rebuttal Report ¶¶ 34-36. In that same scenario, Dr. Capps treated both Dole and Del Monte as if they had made the labeling claim, assigning each a "1" for purposes of the "label" variable. Id. All other brands received a "label" code of "0." Id.9 In so doing, Dole asserts that Dr. Capps elided very real differences in brands and the impact those differences have on pricing. Mot. at 10-11.

Brazil defends Dr. Capps' approach. "In dealing with brand," Brazil maintains, "the bundling of the other products into a single category (all other brands but Dole), is a scientifically accepted method to control for brand." Opp'n at 10. According to Brazil, "Dr. Capps' objective was to control for the impact on brand and to insure his mislabeling coefficient was not a result of brand name," not to "identify the impact of each brand name on price." Id. at 11 (emphasis omitted). Brazil asserts that Dr. Capps' use of "dummy variables" to control for brand was perfectly proper. Id. at 11-12.

In resolving the issue, the Court pauses to note that Brazil's defense of Dr. Capps' brand bundling is rather perfunctory. Brazil spends more time attacking Dole's expert than detailing for the Court what Brazil's own expert did and why that was not improper. See Opp'n at 10-12. Nevertheless, the Court agrees with Brazil that the critical variable to control for is "the impact of the Dole brand on price versus the impact of other brands on price." Id. at 11. In determining the price premium of Dole's "All Natural Fruit" labeling claim, which is what the Court requested, see Class Cert Order at 26, it is essential that Dr. Capps' model control for brand loyalty to Dole. Indeed, what the Court wants to know is that the proffered damages model isolates the price premium attributable to Dole's labeling claim only, rather than any premium attributable to consumers' loyalty to the Dole brand. As a general matter, bundling other dissimilar brands into the same category for coding purposes appears to simplify, rather than undermine, that task.

While the Regression Model fails in this particular case for the reasons explained below, the Court will not decertify on the basis of Dole's more general criticism about the model's coding for "brand" and "label."10

c. Alleged Flaw #3: The Model Improperly Uses Retail Data

The third flaw Dole advances is that Dr. Capps improperly relied on retail sales data in performing his regression analysis. Mot. at 11-12. Dole argues that the Regression Model cannot tie any price premium to Dole specifically because it relies on prices set by retailers, not by Dole. Id. at 11. Dole contends, moreover, that Dole uses "line pricing" anyway, meaning that Dole charges its wholesale customers the same price regardless of whether the label says "All Natural Fruit." Id. at 12. Citing Weiner v. Snapple Beverage Corp., No. 07 CIV. 8742 DLC, 2010 WL 3119452 (S.D.N.Y. Aug. 5, 2010), and Red v. Kraft Foods, Inc., No. CV 10-1028-GW AGRX, 2012 WL 8019257 (C.D. Cal. Apr. 12, 2012), Dole claims that its line pricing is powerful evidence that no premium exists at all for its "All Natural Fruit" labeling claim. Mot. at 12; Reply at 6.

The Court does not necessarily agree. Dole emphasizes the part of the Court's May 30 order mentioning "the benefit Dole received from its label statements." Class Cert Order at 33. Although restitution may be measured in terms of Dole's ill-gotten gains, see Colgan, 135 Cal. App. 4th at 700, the question the Court asked was what price premium, if any, can be attributed to Dole's use of the "All Natural Fruit" label on the challenged products, Class Cert Order at 26. For that reason, the Court defined restitution as the "difference between the market price actually paid by consumers and the true market price that reflects the impact of the unlawful, unfair, or fraudulent business practices." Id. (emphasis added); see also Colgan, 135 Cal. App. 4th at 695 ("[T]he remedy of restitution serves two purposes—returning to the plaintiff monies in which he or she has an interest and deterring the offender from future violations."). Dr. Capps' use of Dole's actual weekly retail sales data was therefore proper.

That Dole employs line pricing and elects not to charge wholesalers a higher price for products labeled "All Natural Fruit" does not categorically mean that consumers do not value that label on their food products. Neither case Dole cites indicates that wholesaler line pricing in any way forecloses the existence of a consumer premium for products bearing the "All Natural Fruit" labeling claim. In Weiner, the court found the plaintiffs' expert unreliable because, inter alia, he had done "a cursory review of the underlying record in this action," had not even "attempted to identify any comparable products to be used in his analysis," and had failed to "explain how his approach would isolate the impact of the `All Natural' labeling from the other factors that purportedly affect the price." 2010 WL 3119452, at *7-8. Successfully or not, Dr. Capps has done far more. Additionally, the Weiner court rejected "Plaintiffs' assertion that causation and injury can be proven class-wide using wholesale, rather than retail, price data," finding the claim "purely speculative and unsupported by [their expert's] testimony." Id. at *10. Contrary to Dole's suggestion, the court in Weiner did not conclude that "Snapple's use of line pricing was compelling evidence of no premium." Mot. at 12. Rather, the court found only that the expert would have realized that Snapple used "line prices" with wholesalers had the expert bothered to "examine[] Snapple's wholesale pricing structure." 2010 WL 3119452, at *10 n.19. The court's footnote says nothing about whether line pricing itself implicates harm to consumers. Red, for its part, does not even mention line pricing. 2012 WL 8019257, at *11.

The Court, thus, is not convinced that Dr. Capps' reliance on retail data was improper and will not decertify on that basis.

d. Alleged Flaws #4, #5, and #6: The Model Does Not Control for Other Variables and Scenario 3, on Its Own, Fails Comcast

The fourth flaw Dole advances is that, despite Dr. Capps' assurances, his Regression Model fails to control for other variables. Mot. at 12-14. The Court preliminarily accepted Dr. Capps' model on the basis that it could "control[] for all other factors that may affect the price of Dole's fruit cups." Class Cert Order at 30. Specifically, the Court approved a model that could "control[] for variables such as Dole's advertising expenditures, the prices of competing and complementary products, the disposable income of consumers, and population." Id. The Regression Model, Dole argues, "controls for none of these." Mot. at 13. Further, the fifth flaw Dole alleges is that Dr. Capps' analysis is littered with errors, including Dr. Capps' failure to corroborate many of the assumptions he made as to whether certain non-Dole products actually made an "All Natural" labeling claim. Id. at 14-16. Finally, the sixth alleged flaw is that Scenario 3 in the Regression Model fails Comcast because Scenario 3 relied solely on a coefficient lifted from an unrelated 2007 study on yogurt. Mot. at 16-17. The Court addresses each flaw below.

The Court agrees with Dole that Brazil has failed to show how the Regression Model controls for other variables affecting price. With respect to advertising, Dr. Capps admits that he did not control for this variable. Capps Dep. at 70. His expert report never mentions why he chose not to do so. See generally Capps Report. The only answer Dr. Capps gives is found in Dr. Capps' deposition testimony, where he says there was no need to control for advertising because "it's reflected in part as part of the [retail] price." Capps Dep. at 70. That is no answer, however. It is precisely because advertising expenditures "would be reflected in the retail price" that a model would need to control for it. Id. How else could the Court know whether any price premium on the challenged products was due to Dole's "All Natural Fruit" labeling claim rather than to its advertising expenditures? Perhaps Dr. Capps could not control for advertising expenses because he could not obtain the data on a "week-to-week" basis. Id. "It is not likely at all," Dr. Capps opined, "that Dole would have weekly advertising expenditures attached to each and every UPC." Id. at 70-71. That is no answer, either. Indeed, insufficient data only favors a finding that the Regression Model is incapable of accomplishing its objective. See Jones, 2014 WL 2702726, at *20 (rejecting one of Dr. Capps' regression models where he had not "determined whether the data related to any or all of his proposed control variables exists").

Notwithstanding Dr. Capps' deposition testimony, Brazil contends that "[Dr. Capps] controlled for advertising via his quarterly and year dummy variables." Opp'n at 14. However, the variables for "year" and "seasonality" should affect all of the brands equally. Brazil does not explain how seasonal spikes in advertising—say, for the holiday season—affect Dole differently than any other brand. All brands may spend more on advertising during the holidays. If so, the seasonality variable may be incapable of differentiating across brands, and the model would erroneously treat Dole, which may spend much more on advertising than its competitors, the same as its competitors in this regard. The Court would not know, then, how much of the identified premium was due to Dole's "All Natural Fruit" labeling claim and how much was due to its advertising expenditures.

As to prices of competing products, Brazil asserts that Dr. Capps "controlled for prices of competing and complementary products." Opp'n at 14. In support of this claim, Brazil writes only: "Fundamentally, the dependent variable in the hedonic regressions consisted of a stacking of prices of not only the relevant Dole products, but also comparable products." Id. This conclusory response, submitted without citation or analysis, is unhelpful. What concerns the Court, moreover, is that many of Dr. Capps' assumptions about the competing products upon which his model relies are either false or untested. To start, the whole reason Dr. Capps performed regression analysis under two scenarios was because he "wasn't certain" that Del Monte, Dole's chief competitor, actually made the "All Natural" labeling claim on its products. Capps Dep. at 26. But why didn't Dr. Capps verify this information or have Brazil verify it for him? Brazil provides no explanation.

In fact, Dr. Capps chose not to corroborate many assumptions he made about the regional and private label brands he included in the model. With one exception, Dr. Capps did not check in person whether the non-Dole brands he assumed had no "All Natural" label actually had one. Capps Dep. at 29-32. Dr. Capps, instead, relied on his own experience and an Internet website that displayed visuals of some of the non-Dole products to conclude that all the non-Dole products, except for Del Monte in Scenario 1, contained no "All Natural" label claims. Id. at 27, 29-32. Yet the best Dr. Capps can say is "I tried to look at all of them," recognizing full well that he was "at the mercy of what will be displayed on the Hoovers website." Id. at 32.

This methodology cannot survive Comcast. The whole objective of Dr. Capps' model is to isolate the price premium attributable to Dole's "All Natural Fruit" label claim. As Dr. Capps explains, "In hedonics, you have to have some products where a labeling claim was made—in this case, all natural—and other products that didn't have the claim." Capps Dep. at 42-43. Exactly. So if the model is unsure whether the non-Dole products actually made an "All Natural" labeling claim, then how can the Court know whether the price premium the model generates is based on Dole's labeling claim rather than on some other factor? Put simply, it cannot. For if another brand's label makes the claim, and Dole indicates that some of them do, see Mot. at 15, then whatever price difference exists between that brand's product and Dole's cannot be attributed to the "All Natural Fruit" claim.

Brazil's response is unpersuasive. Brazil does not dispute Dole's contention that Dr. Capps did not sufficiently verify which non-Dole products contained the "All Natural" label. Instead, Brazil merely responds that the private label products identified by Dole contain only the phrase "Natural," rather than "All Natural." Opp'n at 15-16. Brazil offers no basis for why this distinction matters. Moreover, Dr. Capps' failure to adequately verify the labels of non-Dole products is fatal to his analysis.

Equally concerning is the model's failure to account for the possibility that some products might make multiple labeling claims. Consider the example raised by Dr. Carol Scott, Dole's expert, where a product contains both an "All Natural Fruit" claim as well as a "No Sugar Added" claim. Scott Report ¶ 56; see Mot. at 14. If such a product were then compared to a regional or private label that makes no claim whatsoever, then how can the Court know how much of the resulting price premium derives from which labeling claim? Brazil provides no answer.

The Regression Model also overlooks differences in how the products are packaged. As Dole points out, consumers might be willing to pay a premium for fruit products packaged in a certain way. Mot. at 13. Brazil agrees. Opp'n at 14. But the model does not account for this factor. Many of the challenged products, such as the "Pineapple Tidbits," come in "four packs," or four, 4-oz. cups packaged together. Scott Report ¶ 53 n.51. A competitor pineapple product used by Dr. Capps in his regression, by contrast, appears to come in a single can. Id. The model makes no distinction, except to control for package "size"—i.e., a package's total ounces of fruit. Capps Report ¶ 21. In other words, Dr. Capps' model treats a "four pack" as equal to a 16-oz can. There is no control for packaging convenience in the model, even though consumers might well pay a premium for the convenience of four individual fruit cups. See Scott Report ¶ 11 n.10 ("As anyone who has ever packed school lunches can attest, a 4-pack of products that are 4 ounces each is quite different than a single 20-ounce multiple serving can. . . .").

In the end, it is not enough for Brazil or Dr. Capps to just say that the Regression Model controls for other factors; Brazil must show the Court that the model can. Brazil has not done so. Thus, Brazil has not met his burden to show that the model he proposes is capable of controlling for all other factors and isolating the price premium attributable to Dole's "All Natural Fruit" label only. Cf. Class Cert Order at 28 (rejecting Dr. Capps' "Price Premium Model" due to its "inability to account for . . . any factors that may cause consumers to prefer the identified Dole products over other identical products"). As such, Comcast requires the Court to find that the Rule 23(b)(3) predominance requirement has not been satisfied. Comcast, 133 S. Ct. at 1433 ("[A] model purporting to serve as evidence of damages in this class action must measure only those damages attributable to [defendant's conduct].").11

The Court's conclusion is only bolstered by other troubling aspects of Dr. Capps' model. For instance, Dole highlights a reply declaration Dr. Capps submitted for another product labeling case in the district, Lanovaz v. Twinings North America, Inc. Mot. at 7. This is not the first time the Court has expressed unease with regards to the apparently contradictory opinion Dr. Capps offered in that case regarding the efficacy of regression modeling when the label statement on a product has not changed during the class period. See Order Denying Recons. at 5-6 (expressing concern over Dr. Capps' "somewhat contradictory testimony" on this point). In rejecting Dr. Capps' proposed damages models in Lanovaz, Judge Whyte quoted Dr. Capps' reply declaration: "[T]he antioxidant claims have been on the labels over the entire class period. Hence, it is not possible in this case to invoke a regression analysis approach because of the lack of any variable in sales or units sold attributed to the antioxidant claims." Lanovaz v. Twinings N. Am., Inc., No. C-12-02646-RMW, 2014 WL 1652338, at *6 (N.D. Cal. Apr. 24, 2014).

How is it that regression analysis was "not possible" in Lanovaz, but remains eminently so here? Dr. Capps attempted to explained the discrepancy in his deposition testimony:

So in [Lanovaz], with that particular regression analysis approach, it wasn't possible to do a before and an after. But with the hedonic pricing analysis, still focusing on the regression approach, but now focusing on prices, one can do it.

Capps Dep. at 42. In other words, because the dependent variable in Lanovaz was "units sold," Dr. Capps could not employ a hedonic price analysis, which only examines the impact a particular attribute has on a product's price. Even so, it is not clear to the Court why units sold, rather than price, had to be the dependent variable in Lanovaz. There, as here, the damages model needed to show the "price premium attributable to [the defendant's] use of [a certain] label." Lanovaz, 2014 WL 1652338, at *6. All things being equal, it would seem that a hedonic regression analysis could have been used in Lanovaz to generate the price premium traceable to Twinings' antioxidant label claim. Neither Brazil nor Dr. Capps ever provides a satisfactory explanation on this score. The Court is left only to wonder.

Finally, the Court agrees with Dole that Scenario 3 in the Regression Model fails Comcast because Dr. Capps simply borrowed the hedonic regression coefficient from Prof. Anstine's 2007 study evaluating the price premium of the "All Natural" label claim on yogurt. Capps Report ¶¶ 9, 25 (citing Anstine, supra). Using that coefficient, Dr. Capps derived a damages estimate more than twice as large as under either of Dr. Capps' two other scenarios. Id. ¶ 25. Dr. Capps, however, offers no reason to believe that packaged fruit consumers value an "All Natural" label claim to the same extent yogurt consumers might. Indeed, Dr. Capps offers no explanation at all for why he created a damages estimate based on the Anstine study other than that the study involved "an `all natural' claim." Capps Dep. at 54. As nothing in the Anstine study purports to address the price premium attributable to Dole's labeling claim, Comcast would bar the Court from considering Scenario 3 had the Court not already ruled that the entire model does not satisfy Comcast.

3. Conclusion

Accordingly, because Brazil's proposed damages model fails to provide a means of showing damages on a classwide basis through common proof, the Court concludes that Brazil has not satisfied the Rule 23(b)(3) requirement that common issues predominate over individual ones. The Court must therefore decertify the Damages Class.

In light of the Court's ruling, it need not address Dole's request for a non-testimonial Daubert review, which is DENIED as moot. The Court also DENIES Brazil's request to keep the Damages Class certified on the basis that nominal damages might still be available for class members. See Opp'n at 18 (citing Cal. Civ. Code § 3360). Brazil cites no authority to suggest that a damages class should remain certified solely because nominal damages may be available, even though the class would otherwise be properly decertified. Existing authority suggests that nominal damages are not even available in this case, where Brazil seeks damages only under the CLRA. See Jones, 2014 WL 2702726, at *23 (declining to certify a CLRA class for nominal damages in a food labeling case).

B. Ascertainability

"As a threshold matter, and apart from the explicit requirements of Rule 23(a), the party seeking class certification must demonstrate that an identifiable and ascertainable class exists." Sethavanish v. ZonePerfect Nutrition Co., No. 12-2907, 2014 WL 580696, at *4 (N.D. Cal. Feb. 13, 2014) (internal quotation marks omitted). A class is ascertainable if the class is defined with "objective criteria" and if it is "administratively feasible to determine whether a particular individual is a member of the class." See Wolph v. Acer Am. Corp., No. 09-1314, 2012 WL 993531, at *1 (N.D. Cal. Mar. 23, 2012).

Here, Brazil has adequately defined the class based on an objective criterion: purchase of the identified Dole fruit products within the class period. The class definition "simply identifies purchasers of Defendant's products that included the allegedly material misrepresentations." Astiana v. Kashi Co., 291 F.R.D. 493, 500 (S.D. Cal. 2013). Likewise, "[b]ecause the alleged misrepresentations appeared on the actual packages of the products purchased, there is no concern that the class includes individuals who were not exposed to the misrepresentation." Id. In the Court's view, "[t]his is enough to satisfy Rule 23(a)'s implied ascertainability requirement." Forcellati v. Hyland's, Inc., No. CV 12-1983-GHK MRWX, 2014 WL 1410264, at *5 (C.D. Cal. Apr. 9, 2014) (certifying class of consumers who purchased "Defendants' children's cold or flu products within a prescribed time frame").

In the instant motion, Dole argues that both the Injunction and Damages Classes are no longer ascertainable because "the `All Natural Fruit' statement was not added to the Wildly Nutritious Signature Blends Mixed Fruit label until 2009." Mot. at 18; see also Spare Decl. ¶ 4 ("The `Wildly Nutritious Signature Blends Mixed Fruit' label did not even state `All Natural Fruit' until 2009."). This fact alone, however, does not justify decertification. There is evidence in the record suggesting that the "All Natural Fruit" label appeared on the "Frozen Mixed Fruit" product as early as September 29, 2008. See ECF No. 107-4 (photograph of product showing "All Natural Fruit" label statement dated "09.29.2008"); Capps Report ¶ 14. Moreover, Brazil suggests that the class period can simply be amended to begin on January 1, 2009, the date on which IRI data became available. See Opp'n at 18 n.26 (proposing an amended class start date of January 1, 2009); see also Capps Report ¶ 16. As Dole does not object to this amended date in its Reply, the Court will adopt the amended date for the Injunction Class.

Furthermore, Dole's citations to Bruton and Jones do not support decertification on ascertainability grounds. Unlike here, Bruton involved "69 different types of Gerber baby food products." 2014 WL 2860995, at *4. More importantly, there, unlike here, "Plaintiff's proposed method for identifying class membership require[d] consumers to recall much more than whether or not they purchased" the challenged products. Id. at *8. The consumers in Bruton had to recall "(1) whether they purchased a Gerber 2nd Foods product; (2) whether they purchased a 2nd Foods product in a qualifying flavor; (3) whether the product was in the appropriate packaging; and (4) whether the product was labeled with a challenged label statement." Id. Here, in contrast, class members would only have to recall whether they purchased any challenged products, all of which bore the labeling claim, during the revised class period. Jones is distinguishable for the same reason, because the class members there would have had to remember whether they purchased the challenged products and whether those products contained the allegedly misleading label statement. 2014 WL 2702726, at *10-12. Class members here need only remember the former.

Because the proposed Injunction Class is sufficiently definite to identify putative class members, the Court finds it ascertainable. Having granted Dole's decertification motion as to the Damages Class, the Court need not reach the ascertainability of that proposed class.

IV. CONCLUSION

For the foregoing reasons, the Court GRANTS in part and DENIES in part Dole's Motion to Decertify. Brazil has satisfied the requirements of Rule 23(a) and (b)(2), but he has failed to satisfy the predominance requirement of Rule 23(b)(3).

The Court DENIES Dole's Motion to Decertify the Injunction Class. The Court CERTIFIES the following class under Rule 23(b)(2): "All persons in the United States who, from January 1, 2009, until the date of notice, purchased a Dole fruit product bearing the front panel label statement `All Natural Fruit' but which contained citric acid and ascorbic acid. Excluded from the class are (1) Dole and its subsidiaries and affiliates, (2) governmental entities, and (3) the Court to which this case is assigned and its staff."

The Court GRANTS Dole's Motion to Decertify the Damages Class. The Court therefore DENIES Brazil's request to keep the Damages Class certified for purposes of pursuing nominal damages.

The Court DENIES as moot Dole's request for a non-testimonial Daubert review.

IT IS SO ORDERED.