ORDER GRANTING RENEWED MOTION TO REMAND AND REMANDING ACTION TO STATE COURT

Re: Dkt. No. 41

JOSEPH C. SPERO, Chief Magistrate Judge.

I. INTRODUCTION

Defendants removed this action from State court under the Class Action Fairness Act ("CAFA"), which requires, inter alia, that the removing party demonstrate that at least $5 million is in controversy. On December 12, 2017, the Court denied without prejudice Plaintiff's Motion to Remand, finding that Defendants had failed to meet their burden in establishing that CAFA's amount-in-controversy requirement was met and permitting the parties to conduct limited jurisdictional discovery. Presently before the Court is Plaintiff's Renewed Motion for Remand ("Renewed Motion"). The only question before the Court is whether Defendants have met their burden as to CAFA's amount-in-controversy requirement. For the reasons stated below, the Court concludes that they have not. Accordingly, the Court GRANTS the Renewed Motion and remands this case to the Superior Court of California, County of San Mateo.1 The motion hearing and further case management conference set for May 11, 2018 are vacated.

II. ANALYSIS2

A. Whether Defendants Have Demonstrated by a Preponderance of the Evidence that CAFA's Amount-in-Controversy Requirement is Met

The crux of the dispute is whether Defendants have established by a preponderance of the evidence that the economic losses placed into controversy by Plaintiff's First Amended Complaint are sufficient to meet the amount-in-controversy requirement.3 The Court concludes that Defendants have not met their burden because the method they have used to estimate damages is based on unsupported assumptions and questionable methodology, rendering their estimate speculative. The Court also finds that Defendants' estimate does not accurately reflect the scope of the class claims Plaintiff asserts in his First Amended Complaint because it includes the estimated value of denied claims for repairs of preexisting damage, that is, damage that occurred outside of the rental period.

1. Legal Standard

Under CAFA, the burden to establish removability is on the defendant. Abrego Abrego v. The Dow Chemical Co., 443 F.3d 676, 684 (9th Cir. 2006). "[W]hen a defendant seeks federal-court adjudication, the defendant's amount-in-controversy allegation should be accepted when not contested by the plaintiff or questioned by the court." Dart Cherokee Basin Operating Co., LLC v. Owens, 135 S.Ct. 547, 553 (2014). Where the amount in controversy is challenged, "both sides submit proof and the court decides, by a preponderance of the evidence, whether the amount-in-controversy requirement has been satisfied." Id. at 554 (citing 28 U.S.C. § 1446(c)(2)(B)). To demonstrate the amount in controversy for the purposes of establishing that removal was (or was not) proper, the parties may submit evidence beyond the complaint such as affidavits, declarations, or other "summary-judgment type evidence relevant to the amount in controversy. . . ." Singer v. State Farm Mut. Auto. Ins. Co., 116 F.3d 373, 377 (9th Cir. 1997); see also Orr v. Bank of Am., NT & SA, 285 F.3d 764, 773 (9th Cir. 2002) ("A trial court can only consider admissible evidence in ruling on a motion for summary judgment"); Gaus v. Miles, Inc., 980 F.2d 564, 567 (9th Cir. 1992) (holding that where the jurisdictional amount in controversy is challenged, it is the removing party's burden to come forward with "competent evidence" that this requirement is satisfied).

To meet its burden a removing party is not required to "research, state, and prove the plaintiff's claims for damages." Coleman v. Estes Express Lines, Inc., 730 F.Supp.2d 1141, 1147-48 (C.D.Cal.2010) (citations omitted). On the other hand, "a defendant cannot establish removal jurisdiction by mere speculation and conjecture, with unreasonable assumptions" under CAFA. Ibarra v. Manheim Investments, Inc., 775 F.3d 1193, 1197 (9th Cir. 2015). Thus, where a damages estimate is based on assumptions that have no evidentiary support in the record, the estimate may be insufficient to establish that the amount-in-controversy requirement is met. See Roth v. Comerica Bank, 799 F.Supp.2d 1107, 1126 (C.D. Cal. 2010)(finding in wage and hour case removed to federal court under CAFA that defendant's estimate of amount in controversy was speculative because many of the assumptions that were "key to defendants' damages calculations" were not supported by summary judgment-type evidence and remanding to state court).

2. The Benn Declaration

As a preliminary matter, the Court notes that the evidence and methodology on which Defendants now rely to establish the amount in controversy is new. In opposition to the original motion to remand, Defendants relied on the declaration of Turo's president, Alex Benn, who stated that he had personal knowledge, based on his own "review of records" and "research," that allowed him to estimate the number of claims denied since 2012 and the average value of those claims. Henderson Decl., Ex. 1 (Benn Opposition Decl.). Subsequent jurisdictional discovery revealed that the handful of documents that Benn had purportedly "reviewed" to come up with these numbers — a spreadsheet and three graphs —lend little or no support for his figures. Moreover, as to the spreadsheet (which reflected a total value of denied claims in the amount of $10,125,000), see Henderson Decl., Ex. 6, Benn testified that he did not know where the figures in it came from or what they were based on, and that he only remembered seeing the document on the screen of his attorney's computer. Id., Ex. 7 (Benn Dep.) at 42-43. In his deposition, Benn was unable to offer any solid evidentiary basis for the numbers contained in his declaration.

Apparently, the figures in the Benn Declaration actually came from Turo's Senior Director of Risk, Tristam Hewitt, who states in a declaration submitted in opposition to the Renewed Motion that the numbers in the spreadsheet upon which Benn says he relied are "rough estimates" that Hewitt provided to Turo's counsel. Hewitt Decl. ¶ 9 & Ex. B (spreadsheet). Hewitt does not explain in his declaration how he derived the figures in the spreadsheet. Defendants have now abandoned these numbers, making no effort to defend them or even to show that they ever had a good faith basis for believing that they were reasonable estimates. The Court is troubled by Defendants' cavalier (at best) approach towards the facts. This problem persists in Defendants' new analysis of the amount-in-controversy, as discussed below.

3. Reliability of Estimate Offered By Defendants in Opposition to the Renewed Motion

Following the hearing on Plaintiff's original motion to remand, Hewitt "was asked to provide a more detailed analysis as to the number and value of first party claims using a conservative valuation method." Hewitt Decl. ¶ 10. He did so and on the basis of that analysis, Defendants now contend the economic losses that are placed in controversy by Plaintiff's First Amended Complaint amount to at least $3,358,379.79. See Opposition at 9 (citing Hewitt Decl. ¶ 18).

Hewitt's analysis involved a number of steps. First, he used Turo's Dunlop system, an internal database that was created by Turo and became operational in March 2015, to determine the total number of claims that were fully denied between March 2015 and July 14, 2017, when the action was removed to federal court.4 Hewitt Decl. ¶¶ 8, 10. He found that there were either 3,310 or 3,210 such claims. See Hewitt Decl. ¶ 10 (3,310), ¶ 18 (3,210). Next, because it would have been extremely labor intensive (and as to some denials, impossible) to determine the value of the denied claims based on actual appraisal amounts, Mr. Hewitt developed a method to estimate the total value of these denied claims. Id. ¶ 15.

Hewitt started by creating an Excel spreadsheet of the denied claims and populating it with the fair market values of the vehicles involved, which he obtained from the Edmunds online database. Id. ¶ 10. Hewitt then obtained from a Turo employee, Todd Armstrong, who previously worked as a field estimator for Progressive Insurance, a list with estimated repair costs for certain general categories of damages. Id. ¶ 11. The list gave a dollar amount for each category of repair and also gave a percentage for each type of repair that reflected the cost of the repair as a percentage of a vehicle's total market value. Armstrong Decl., Ex. A.5 Armstrong derived the percentages in his list using the average cost of repair for particular types of damage (apparently based on the cost of repair "across vehicle types") as compared to a car worth $15,000. Armstrong Decl. ¶ 5. He states in his declaration that he used the $15,000 average vehicle value because this was "the rounded average value of vehicles that were listed on Turo's marketplace close to the time that the estimates were made." Id. The list provided by Armstrong also includes the notation "Methodology[:] Parts more valuable for more expensive[;] Repairs: labor and paints time matter more[.] Percentages based on an average vehicle value of $15,000 and average cost of type of repair across vehicle types." Id., Ex. A. Armstrong does not address in his declaration the relationship between vehicle value and repair costs or provide any details as to the "vehicle types" he considered to obtain the average cost of repair.

Next, Hewitt "generated an Excel formula to conduct keyword searches for various types of common damage." Hewitt Decl. ¶ 12. These searches encompassed only the portion of the denied claims for which the Claims Team had entered annotations describing the type of damage at issue — about 60% of the denied claims. Id. ¶ 10. According to Hewitt, the search formula would flag references in the claim annotations to the various categories of damages and then, for each claim, the percentages corresponding to these different types of repairs (as reflected in the list provided by Armstrong) would be added together and multiplied by the market value of the car to obtain the value of that claim. See Henderson Decl., Ex. 8 (Hewitt Dep.) at 66-68. Hewitt testified at his deposition that even though an annotation might contain multiple references to the same type of damage, that category would be counted only once.6 Id. at 67.

Finally, Hewitt used this information to obtain the average value of each fully denied claim for which there was an annotation and then used that number to estimate the value of the fully denied claims for which there were no annotations to come up with the estimated total value of all fully denied claims. Hewitt Decl. ¶18.

Hewitt's method is complicated and he came up with vastly different estimates based on the search formulas he used. When Plaintiff pointed out in his Renewed Motion that the claim in row 13 of the spreadsheet, for a 3-inch scratch, was miscategorized as an overheated engine claim and valued at over $11,000, see Renewed Motion at 15-16, Hewitt revisited his search formulas and concluded that at least one of them was incorrect. Hewitt Declaration ¶ 17. He states in his declaration that he then "corrected the error in the formula," which "materially reduced" his damages estimate. Id. & Ex. J. Indeed, whereas he had previously concluded that the average value of fully denied claims was $2,167.35, the value he came up with after modifying the formula was only $1,097.71, leading his overall estimate of the value of denied claims to drop by almost half. See Hewitt Decl. ¶¶ 12, 18 & Ex. J, F. Subsequently, Hewitt found another error in the formula related to the "dents" search term and again revised his search formulas, resulting in a further reduction of the average claim amount to $1,046.22 (the amount used for the estimate Defendants now rely upon in opposing the Renewed Motion). Hewitt Decl. ¶ 18 & Ex. K.

While Hewitt has gone to extensive effort to estimate the value of denied claims, the Court is not persuaded that his methodology is reliable. The Court is particularly concerned about the "search formulas" he used to flag categories of repairs, which can dramatically affect the estimate of the value of denied claims. Hewitt has modified these formulas at least twice to correct for obvious problems with the results he obtained. These changes were made in response to arguments in the Renewed Motion, after Hewitt was deposed, thus depriving Plaintiff of the opportunity to address the reliability of the Excel formulas Hewitt used to obtain his current estimate.7 Moreover, at his deposition, Hewitt was not able to adequately explain how the Excel search worked or what some of the notations on his spreadsheet meant, making it even more difficult for Plaintiff — and by extension, the Court — to determine whether the method Hewitt used was reliable.

Further, the spreadsheet and declarations that Defendants have provided do not fully explain Hewitt's methodology. For example, the chart in Hewitt's declaration listing keyword searches and the corresponding spreadsheet columns lists the same columns (BH through BJ) for a small dent (with a repair cost estimated at 5% of the vehicle's value) and a large dent (with a repair cost estimated at 17% of the vehicle's value). Hewitt Decl. ¶ 12. It is not clear how (or if) the searches differentiate between the two types of damage. The spreadsheet itself seems to reflect that the percentage Hewitt used for dents was 11%. See Hewitt Decl., Ex. C at TURO_000929. It is possible that there is a reasonable explanation for this discrepancy. Perhaps Hewitt was unable to capture the difference between a small dent and a large dent in his search formulas and therefore he used the average of the two numbers. The Court is left guessing — both as to how the Excel formulas were actually used and whether Hewitt's approach was reasonable. Likewise, the current record does not provide the Court with any basis to conclude that the formulas that Hewitt now says he has corrected are, in fact, sound. Nor has he offered any specific evidence explaining how he created these formulas or showing that he is qualified to conduct searches that will produce reliable results.

Defendants also have not offered any evidence to show that the percentages provided by Armstrong, valuing repairs as a percentage of the value of the vehicle, represent a reliable method for estimating the likely cost of repairs for denied claims. Armstrong's chart contains a somewhat cryptic notation suggesting that parts or labor may be more expensive for cars with a higher market value, but he does not address this question in his declaration. Nor does Hewitt offer any testimony to support this approach. In response to a declaration by Cory Liang submitted by Plaintiff stating that this is not a reliable means of estimating the cost of repairs, see Henderson Decl., Ex. 16 (Liang Decl.), Defendants criticize Liang for failing to opine as to whether the percentages are likely to give rise to estimates that are too high or too low, and note that Liang does not "provide any basis for how he would calculate the value of the denied claims in this case." Opposition at 17 (emphasis added). Defendants do not, however, explain why the methodology used by Hewitt and Armstrong to estimate the cost of repairs is reliable.

Hewitt is not testifying as a lay witness about facts within his personal knowledge; rather, he is offering expert opinions based on his analysis of the evidence available to him related to the amount in controversy. Accordingly, under Rule 702 of the Federal Rules of Evidence, his opinions must be based on a reliable methodology to be admissible. See Ellis v. Costco Wholesale Corp., 657 F.3d 970, 982 (9th Cir. 2011) ("Under [Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 597 (1993)], the trial court must act as a `gatekeeper' to exclude junk science that does not meet Federal Rule of Evidence 702's reliability standards by making a preliminary determination that the expert's testimony is reliable."). Further, it is Defendants' burden to show that Hewitt's methodology is sound. SFF-TIR, LLC v. Stephenson, 250 F.Supp.3d 856, 1000 (N.D. Okla.), reconsideration denied, 264 F.Supp.3d 1148 (N.D. Okla. 2017) (citations omitted) ("The proponent of expert testimony has the burden of establishing by a preponderance of the evidence that the pertinent admissibility requirements are met."). Defendants have not met that burden.

The Court appreciates that Defendants are faced with a difficult task. Many claims that were denied did not proceed to an appraisal and attempts to determine the amounts that were actually at stake for these claims — especially for the claims that were denied several years ago — would be extremely time-consuming and expensive, and perhaps, for some claims, impossible. The problem here is not that Defendants resorted to an approach that relied on certain assumptions to estimate the value of the denied claims; it is that they have not provided an evidentiary basis for the assumptions they used or demonstrated that the methodology was reliable. As a result, the Court concludes that while it may be plausible that Defendants' denied claims during the relevant period exceeded $3 million, Defendants have not established this amount (or any other amount) by a preponderance of the evidence.

4. Whether Defendants' Estimate of Economic Losses Was Overbroad

The parties dispute whether it was appropriate for Defendants to include the amount of all denied claims in their amount-in-controversy estimate. Plaintiff contends his First Amended Complaint makes clear that he is not seeking to assert class claims based on denials of coverage under the specific exclusions that were disclosed by Defendants but instead, only to assert claims based on Defendants' allegedly false and misleading representation that "the owners' vehicles will be covered if a renter damages their car, even if the renter intentionally damages their car." See Renewed Motion at 10-11; FAC ¶ 27. He points out that the common allegations of the First Amended Complaint specifically reference the coverage exclusions, arguing that he was not required to also allege in the claims themselves that denials based on policy exclusions were excluded. Motion at 10-11 (citing FAC ¶ 24).8

Defendants, on the other hand, point out that Plaintiff did just that in his original complaint, alleging in his claim under Cal. Bus. & Prof. Code §§ 17200 that the class he sought to represent was composed of "all persons who [ ] owned a vehicle which was listed on RelayRides's market place, but [were] denied coverage, or not fully covered for damages, after a customer damaged the owner's vehicle, wherein the denial was not based on an express policy exclusion." Opposition at 20 n. 18; Complaint ¶ 30 (emphasis added). In contrast, they note, Plaintiff removed this language from his First Amended Complaint. Id. at 20. In the First Amended Complaint, Defendants contend, the two classes proposed by Plaintiff are defined broadly to include any renter who was "denied coverage . . . after a renter damaged the owner's car during the rental period." Opposition at 20 (quoting FAC ¶35); see also FAC ¶ 34 ("The requested class membership includes those individuals and entities who listed their vehicles on the Relayrides' marketplace and had their cars damaged during the rental period and/or paid out of pocket expenses."). Moreover, Defendants assert, if Plaintiff's claims do not include denials based on Defendants' Terms of Service (which include the coverage exclusions discussed above), he himself has no claim because his own claim for coverage also was denied under the Terms of Service, namely, for renting out an unregistered vehicle that was not in "safe and roadworthy condition." Opposition at 22.

Plaintiff's First Amended Complaint is not a model of clarity. While Plaintiff may have intended to limit his class claims to denials of coverage that were not based on specific disclosed exclusions, the mere fact that those exclusions were recited in his common allegations is not enough to achieve that result where Plaintiff did not state as much in the claims themselves or his class definitions. Given the broad language Plaintiff used in the First Amended Complaint, it was reasonable for Defendants to conclude that Plaintiff's class claims do not categorically exclude denials based on the express exclusions in Defendants' Terms of Service. On the other hand, one particular exclusion, for "pre-existing damage" clearly does fall outside of the scope of Plaintiff's claims. Plaintiff's first proposed class is composed of individuals whose claims were denied "after a renter damaged the owner's vehicle during the rental period." FAC ¶ 35 (emphasis added). Similarly, his second class is composed of individuals who "paid out of pocket expenses after the owner's vehicle was damage during a rental period." Id. (emphasis added). In other words, claims for coverage that were denied on the basis that the damage was "pre-existing" fall outside the scope of Plaintiff's claims.9

Defendants did not separate out the claims that were denied under specific exclusions, but the spreadsheet of claims they provided reflects that almost a thousand claims listed on it were denied on the basis that the damage was pre-existing. In other words, it is likely that by including in their estimate claims that were denied under the pre-existing damage exclusion, Defendants significantly inflated their amount in controversy estimate. For this additional reason, the Court concludes Defendants have failed to establish by a preponderance of the evidence that CAFA's amount-in-controversy requirement is satisfied.

B. Whether the Court Should Award Attorneys' Fees and Costs Incurred in Connection with Removal

Pursuant to 28 U.S.C. § 1447(c), where removal is found to be improper the Court has discretion to "require payment of just costs and any actual expenses, including attorney fees, incurred as a result of the removal." 28 U.S.C. § 1447(c); see Moore v. Permanente Med. Grp., Inc., 981 F.2d 443, 448 (9th Cir. 1992). The Court declines to exercise its discretion to award fees and costs. Defendants may have had a good-faith basis for believing that the requirements for CAFA jurisdiction were satisfied at the time they removed the action to federal court. Furthermore, the allegations and claims in Plaintiff's First Amended Complaint were less clear than they could have been, making it more difficult for Defendants to determine the scope of the class claims. Under these circumstances, it is reasonable for each side to bear its own fees and costs.

III. CONCLUSION

For the reasons stated above, the Court GRANTS in part and DENIES in part Plaintiff's Renewed Motion. In particular, the Court concludes that Defendants have not shown by a preponderance of the evidence that CAFA's amount-in-controversy requirrement is satisfied and therefore REMANDS this action to the Superior Court of California, County of San Mateo. The Court denies Plaintiff's request for attorneys' fees and costs incurred in connection with the removal of this action to federal court.

IT IS SO ORDERED.