OPINION

PAEZ, Circuit Judge:

Between 2004 and 2008, many online internet advertisers used Google, Inc.'s ("Google") AdWords program, an auction-based program through which advertisers would bid for Google to place their advertisements on websites. Pulaski & Middleman, LLC and several other named plaintiffs ("Pulaski")1 brought this putative class action under California's Unfair Competition and Fair Advertising Laws, alleging that Google misled them as to the types of websites on which their advertisements could appear. The putative class initially sought injunctive and restitutionary relief. After Google changed certain features of the AdWords program, Pulaski, upon filing a Third Amended Consolidated Class Action Complaint, abandoned the claim for injunctive relief. The only relief the putative class now seeks is the equitable remedy of restitution.

Pulaski appeals the district court's denial of class certification. The district court held that on the claim for restitution, common questions did not predominate over questions affecting individual class members. In denying certification, the court reasoned that it was not bound by our decision in Yokoyama v. Midland National Life Insurance Co., 594 F.3d 1087, 1094 (9th Cir.2010). It then explained that determining which class members are entitled to restitution and what amount each class member should receive would require individual inquiries that "permeate the class claims."

Pulaski argues that the district court erred in failing to follow Yokoyama. As explained below, we agree. We therefore reverse the denial of class certification and remand for further proceedings.

I. Background

A.

This case concerns Google's AdWords program, an auction-based program through which Google served as an intermediary between website hosts and advertisers. Through AdWords, internet advertisers provided advertisements to Google and its third party website-owner partners. To participate, advertisers entered Google-defined variables into the AdWords interface on Google's website, including the maximum price per ad they would be willing to pay and their overall budget. They also selected which Google-defined categories of websites they wanted to display the ad. Afterwards, using an auction-based algorithm, AdWords determined the online placement and price of the ad. Thus, during the class period, advertisers did not know in advance exactly where their ads would appear.

Advertisers paid a particular price to Google each time an Internet user "clicked" on their displayed ad. The price of a particular click depended on several factors: the maximum bids of other AdWords customers for clicks based on the same search term, a "quality score" of the advertisement, and a "Smart Pricing" discount applied to the website where the ad had been placed. Google created and instituted Smart Pricing, an internally-calculated price adjustment, to adjust the advertiser's bids to the same levels that a "rational advertiser" would bid if the rational advertiser had sufficient data about the performance of ads on each website. Smart Pricing is a ratio calculated by dividing the conversion rate2 for the lower-quality website by the conversion rate for the same ad on google.com.

There are several categories of websites in play. During the class period, an advertiser using AdWords could request that its ads appear on Search Feed sites, Content Network sites, or both. Search Feed sites display AdWords ads along with search results after a user searches for information using a particular search term. After entering a particular term, a user would be presented with both ordinary search results and ads related to the search term. Content Network websites, on the other hand, are full content sites, like nytimes.com, that publish information independent of search results. Ads would appear on these sites if the ad's keywords matched those of the website.

There are other categories of sites that did not appear in the AdWords registration process: parked domains and error pages. Parked domain pages are undeveloped domains whose pages appear when users type generic terms into a web browser. These are pages of ads without content. Error pages appear when a person inputs an unregistered web address, or something other than a web address, into a web browser's address bar. Typing this information into an address bar used to result in error messages, but during the class period inputting this information resulted in error pages that offered ads. Even though only Search Feed and Content Network websites were listed in the AdWords registration process, AdWords ads appeared on both parked domains and error pages.

B.

Pulaski alleges that Google misled advertisers, violating California's Unfair Competition Law ("UCL"), Cal Bus. & Prof.Code § 17200 et seq.,3 and California's Fair Advertising Law ("FAL"), § 17500 et seq., by failing to disclose the placement of AdWords ads on parked domains and error pages. The putative class consists of "[a]ll persons or entities located within the United States who, from July 11, 2004 through March 31, 2008 ... had an AdWords account with Google and were charged for clicks on advertisements appearing on parked domain and/or error page websites," with exclusions.4 Pulaski, on behalf of the putative class, seeks restitution of moneys Google wrongfully obtained from the putative class.

Pulaski moved for class certification pursuant to Rule 23 of the Federal Rules of Civil Procedure ("Rule 23") for a Rule 23(b)(3) class. Pulaski proposed three different methods for calculating restitution, all of which were based on a "but for" or "out-of-pocket loss" calculation: the difference between what advertisers actually paid and what they would have paid had Google informed them that their ads were being placed on parked domains and error pages. The first approach is based on Google's Smart Pricing formula as described above. The amount of restitution owed a class member would be the difference between the amount the advertiser actually paid and the amount paid reduced by the Smart Pricing discount ratio. The second method is the Content Pricing approach,5 which factors in the lower bidding that would have occurred had advertisers been allowed to bid separately on parked domains and error pages. Search Feed clicks were priced higher than Content Network clicks, which in turn were considered more desirable than parked domains and error pages. Accordingly, where the same ad appeared both in the Search Feed and on Content Network websites, those Content Network ad prices could serve as a conservative but-for price for Search Feed clicks on parked domains and error pages. The third method is the Full Refund approach, in which advertisers would receive full refunds for clicks on ads placed on parked domains and error pages. Because some methods may work better than others for certain subsets of class members, Pulaski presented these methods as possibly complementary.

In ruling on the class certification motion, the district court initially found that the proposed class satisfied all of the criteria under Rule 23(a): numerosity, commonality, typicality, and adequate representation. The court next turned to the predominance inquiry under Rule 23(b)(3). On that issue, it found that, even assuming the plaintiff class could prevail on liability, common questions did not predominate on the issues of entitlement to restitution and amount of restitution due each class member.

First, the court expressed concern that individual questions may arise in ascertaining entitlement to restitution. It observed that "the question of which advertisers among the hundreds of thousands of proposed class members are even entitled to restitution would require individual inquiries." In particular, the court was concerned with how to "systematic[ally] ... identify and exclude from Plaintiffs' proposed class the many advertisers who have no legal claim to restitution because they derived direct economic benefits from ads placed on parked domains and error pages."

Second, the court identified individual questions that would arise in determining the amount of restitution owed to the class and individual class members. The court explained that our decision in Yokoyama, which held that damages calculations alone cannot defeat class certification, did not control the outcome of this issue because Yokoyama cited to decisions that mentioned a "workable method for calculating monetary recovery." Here, the court held that the plaintiffs had not proposed a method that was workable. The court explained that different costs for each advertiser, each ad, and each click, overlaid with an auction process, make it "more difficult to calculate what AdWords customers would have paid `but for' the alleged misstatements or omissions." It concluded that Pulaski's proposed methods were insufficient to account for all of the intricacies involved, including benefits received from parked domain and error pages.

Concluding that individual questions predominated on the issue of restitution, the court denied Pulaski's motion for class certification without addressing whether class treatment was a superior method for resolving the dispute as required by Rule 23(b)(3). Thereafter, Pulaski filed a motion for reconsideration, which the district court denied.

We granted permission to appeal the order denying class action certification as authorized by Rule 23(e). We have jurisdiction under 28 U.S.C. § 1292(e).

II. Standard of Review

A district court's class certification ruling is reviewed for abuse of discretion. See Parra v. Bashas', Inc., 536 F.3d 975, 977 (9th Cir.2008). "A district court would necessarily abuse its discretion if it based its ruling on an erroneous view of the law or a clearly erroneous assessment of the evidence." United States v. Hinkson, 585 F.3d 1247, 1259 (9th Cir.2009) (en banc) (quoting Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 405, 110 S.Ct. 2447, 110 L.Ed.2d 359 (1990)). "[W]hen an appellant raises the argument that the district court premised a class certification determination on an error of law, our first task is to evaluate whether such legal error occurred." Yokoyama, 594 F.3d at 1091. "If the district court's determination was premised on a legal error, we will find a per se abuse of discretion." Id. Otherwise, "we will proceed to review the district court's class certification decision for abuse of discretion as we have always done." Id.

III. Discussion

To obtain certification, a putative class must satisfy four prerequisites:

Rule 23(a). Additionally, the proposed class must qualify as one of the types of class actions identified in Rule 23(b). Here, Pulaski sought to certify a class under Rule 23(b)(3). Under Rule 23(b)(3), the court must find 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."

Pulaski must "affirmatively demonstrate... compliance with the Rule." Wal-Mart Stores, Inc. v. Dukes, ___ U.S. ___, 131 S.Ct. 2541, 2551, 180 L.Ed.2d 374 (2011). The question of certification requires a "rigorous analysis." Id. Courts may have to "probe behind the pleadings before coming to rest on the certification question." Id.

The district court denied certification because it found that the putative class did not meet the predominance requirement. It explained that questions regarding which advertisers are entitled to restitution in the first instance, and the amount of restitution owed to each advertiser, both defeat predominance. We disagree.

A.

Entitlement to restitution is a separate inquiry from the amount of restitution owed under California's UCL and FAL. To the extent that the district court rested its holding that common questions do not predominate on the putative class's entitlement to restitution, it committed legal error.

The UCL prohibits "any unlawful, unfair, or fraudulent business act or practice." § 17200. The FAL prohibits "untrue or misleading" statements in the course of business. § 17500. This language is "broad" and "sweeping" to "protect both consumers and competitors by promoting fair competition in commercial markets for goods and services." Kwikset Corp. v. Super. Ct., 51 Cal.4th 310, 320, 120 Cal.Rptr.3d 741, 246 P.3d 877 (2011).

To state a claim under the UCL or the FAL "based on false advertising or promotional practices, it is necessary only to show that members of the public are likely to be deceived." In re Tobacco II Cases, 46 Cal.4th 298, 312, 93 Cal.Rptr.3d 559, 207 P.3d 20 (2009);6 see also Stearns v. Ticketmaster Corp., 655 F.3d 1013, 1020 (9th Cir.2011) (holding that a district court erred in denying class certification by requiring individualized proof of reliance and causation, and remanding in light of In re Tobacco II Cases), cert. denied, ___ U.S. ___, 132 S.Ct. 1970, 182 L.Ed.2d 819 (2012). This inquiry does not require "individualized proof of deception, reliance and injury." In re Tobacco II Cases, 46 Cal.4th at 320, 93 Cal.Rptr.3d 559, 207 P.3d 20; Stearns, 655 F.3d at 1020 (same). "[I]n effect, California has created what amounts to a conclusive presumption that when a defendant puts out tainted bait and a person sees it and bites, the defendant has caused an injury; restitution is the remedy." Stearns, 655 F.3d at 1021 n. 13.7

Under the UCL:

§ 17203. This language, as well as "nearly identical" language under the FAL, see § 17535, grants a court discretion to order restitution. Cortez v. Purolator Air Filtration Prods. Co., 23 Cal.4th 163, 173, 96 Cal.Rptr.2d 518, 999 P.2d 706 (2000).

Thus, a court need not make individual determinations regarding entitlement to restitution. Instead, restitution is available on a classwide basis once the class representative makes the threshold showing of liability under the UCL and FAL. Accordingly, the district court erred in holding that such individual questions would predominate.

B.

We held in Yokoyama that "damage calculations alone cannot defeat certification." 594 F.3d at 1094. By concluding that it was not bound by Yokoyama under the circumstances presented in this case, the district court erred.

Yokoyama concerned the Hawaii Deceptive Practices Act, Haw. Rev. Stat. § 480-2. We concluded that the district court erred when it held that this law required individualized showings of reliance because Hawaii courts' caselaw "look[ed] to a reasonable consumer, not the particular consumer." Id. at 1092. As we noted, the case, at the liability stage, would "not require the fact-finder to parse what oral representations each broker made to each plaintiff." Id. at 1093. Rather, the liability portion would be uniform, as it "will focus on the standardized written material given to all plaintiffs to determine whether those materials are likely to mislead consumers acting reasonably under the circumstances." Id. Because it committed legal error, its denial of class certification was a "[p]er [s]e [a]buse of [d]iscretion." Id.

The district court in Yokoyama also erroneously concluded that the "damages calculation involved highly individualized and fact-specific determinations," a conclusion to which the district court's premise of subjective reliance may have contributed. Id. In examining predominance for class certification purposes, the district court had considered factors such as:

Id. at 1093-94. We held that, even though all these variables impacted damages calculations, the individualized calculations did not defeat predominance. Id. at 1093; see also Stearns, 655 F.3d at 1026 ("We have held that the mere fact that there might be differences in damage calculations is not sufficient to defeat class certification." (citing Yokoyama)).

Google argues that Comcast Corp. v. Behrend, ___ U.S. ___, 133 S.Ct. 1426, 185 L.Ed.2d 515 (2013), called Yokoyama's holding into question. There, in analyzing a putative antitrust class, the Court held that the plaintiffs' proposed damages model fell "far short of establishing that damages are capable of measurement on a classwide basis." Id. at 1433. The district and circuit courts had failed to inquire into whether the model translated the "legal theory of the harmful event into an analysis of the economic impact of that event." Id. at 1435 (emphasis omitted). The Court reasoned that "a model purporting to serve as evidence of damages in [a] class action must measure only those damages attributable to that theory. If the model does not even attempt to do that, it cannot possibly establish that damages are susceptible of measurement across the entire class for purposes of Rule 23(b)(3)." Id. at 1433. In such a situation, "[q]uestions of individual damage calculations will inevitably overwhelm questions common to the class." Id.

Since Comcast, we have continued to apply Yokoyama's central holding. In Leyva v. Medline Industries, Inc., we reaffirmed that damage calculations alone cannot defeat class certification. 716 F.3d 510, 513-14 (9th Cir.2013) (citing Yokoyama). We explained that Comcast stood for the proposition that "plaintiffs must be able to show that their damages stemmed from the defendant's actions that created the legal liability." Id. at 514; see also Roach v. T.L. Cannon Corp., 778 F.3d 401, 407 (2d Cir.2015) ("Comcast held that a model for determining classwide damages relied upon to certify a class under Rule 23(b)(3) must actually measure damages that result from the class's asserted theory of injury; but the Court did not hold that proponents of class certification must rely upon a classwide damages model to demonstrate predominance."). The putative class's problem in Comcast was that the damages model "did not isolate damages resulting from any one theory of antitrust impact." Leyva, 716 F.3d at 514 (quoting Comcast, 133 S.Ct. at 1431). Following this discussion, we reversed a denial of class certification in part because the "damages could feasibly and efficiently be calculated once the common liability questions are adjudicated." Id.

We reaffirmed the proposition that differences in damage calculations do not defeat class certification after Comcast in Jimenez v. Allstate Insurance Co., 765 F.3d 1161, 1167 (9th Cir.2014) (quoting Leyva, including the portion quoting Yokoyama). As we explained, our sister circuits have adopted "[s]imilar positions" since Comcast. See id. at 1167-68 (citing cases from the Sixth, Seventh, and Fifth Circuits); see also Roach, 778 F.3d at 407-08 (citing cases from the First, Tenth, Fifth, Seventh, and Sixth Circuits, as well as Leyva and Yokoyama, to support the proposition that Comcast did not hold that Rule 23(b)(3) requires a classwide basis for damages calculation).

In sum, Yokoyama remains the law of this court, even after Comcast. Because "[d]amages calculations alone ... cannot defeat certification" under Yokoyama, the district court erred in concluding that Yokoyama "does not apply to the facts here." Thus, it abused its discretion in denying class certification on this basis. See Yokoyama, 594 F.3d at 1090-92.8

C.

Google argues that the district court properly denied Pulaski's motion for certification under Comcast because the proposed method for calculating restitution was "arbitrary," and thus does not satisfy Rule 23(b)(3)'s predominance requirement. See Comcast, 133 S.Ct. at 1433. We disagree.

Restitution is "the return of the excess of what the plaintiff gave the defendant over the value of what the plaintiff received." Cortez, 23 Cal.4th at 174, 96 Cal.Rptr.2d 518, 999 P.2d 706. Restitution has two purposes: "to restore the defrauded party to the position he would have had absent the fraud," and "to deny the fraudulent party any benefits, whether or not for[e]seeable, which derive from his wrongful act." Nelson v. Serwold, 687 F.2d 278, 281 (9th Cir.1982) (citing the Restatement of Restitution).

Restitution under the UCL and FAL "must be of a measurable amount to restore to the plaintiff what has been acquired by violations of the statutes, and that measurable amount must be supported by evidence." Colgan v. Leatherman Tool Grp., Inc., 135 Cal.App.4th 663, 698, 38 Cal.Rptr.3d 36 (2006). Where a defendant has wrongfully obtained a plaintiff's property, "the measure of recovery for the benefit received ... is the value of the property at the time of its improper acquisition ... or a higher value if this is required to avoid injustice" where the property has changed in value. Id. at 698-99, 38 Cal.Rptr.3d 36 (quoting the Restatement of Restitution). Where plaintiffs are "deceived by misrepresentations into making a purchase, the economic harm is the same: the consumer has purchased a product that he or she paid more for than he or she otherwise might have been willing to pay if the product had been labeled accurately." Kwikset, 51 Cal.4th at 329, 120 Cal.Rptr.3d 741, 246 P.3d 877 (emphasis in original). As the California Supreme Court explained while discussing economic harm in the context of standing, this measure "is the same whether or not a court might objectively view the products as functionally equivalent:"

Id. at 329-30, 120 Cal.Rptr.3d 741, 246 P.3d 877. Applying these concepts to other forms of fraudulent omission, UCL and FAL restitution is based on what a purchaser would have paid at the time of purchase had the purchaser received all the information.

In calculating damages, here restitution, California law "requires only that some reasonable basis of computation of damages be used, and the damages may be computed even if the result reached is an approximation." Marsu, B.V. v. Walt Disney Co., 185 F.3d 932, 938-39 (9th Cir. 1999). "[T]he fact that the amount of damage may not be susceptible of exact proof or may be uncertain, contingent or difficult of ascertainment does not bar recovery." Id. at 939.

We conclude that Pulaski's proposed method was not "arbitrary," as Google argues. The calculation need not account for benefits received after purchase because the focus is on the value of the service at the time of purchase. Instead, in calculating restitution under the UCL and FAL, the focus is on the difference between what was paid and what a reasonable consumer would have paid at the time of purchase without the fraudulent or omitted information. See Kwikset, 51 Cal.4th at 329, 120 Cal.Rptr.3d 741, 246 P.3d 877.

Here, the harm alleged is Google's placement of ads on lower-quality web pages without the advertisers' knowledge. Pulaski's principal method for calculating restitution employs Google's Smart Pricing ratio, which directly addresses Google's alleged unfair practice by setting advertisers' bids to the levels a rational advertiser would have bid if it had access to all of Google's data about how ads perform on different websites. Because restitution under the UCL and FAL measures what the advertiser would have paid at the outset, rather than accounting for what occurred after the purchase, using a ratio from Google's data that adjusts for web page quality is both targeted to remedying the alleged harm and does not turn on individual circumstances. Thus, the Smart Pricing method measures the monetary loss "resulting from the particular ... injury" alleged. See Comcast, 133 S.Ct. at 1434.9

IV. Conclusion

The district court erred by conflating restitution calculation with the liability inquiry for UCL and FAL claims, and by failing to follow our rule in Yokoyama. Further, the proposed method for calculating restitution was not "arbitrary" under Comcast.

REVERSED and REMANDED.