OPINION

HUG, Circuit Judge:

The question presented is whether the district court abused its discretion in approving the parties' $9.5 million settlement agreement as "fair, reasonable, and adequate," either because a Facebook employee sits on the board of the organization distributing cy pres funds or because the settlement amount was too low. We hold that it did not.

I

Facebook is an online social network where members develop personalized web profiles to interact and share information with other members. The type of information members share varies considerably, and it can include news headlines, photographs, videos, personal stories, and activity updates. Members generally publish information they want to share to their personal profile, and the information is thereby broadcasted to the members' online "friends" (i.e., other members in their online network).

In November of 2007, Facebook launched a new program called "Beacon." Facebook described the purpose of the Beacon program as allowing its members to share with friends information about what they do elsewhere on the Internet. The program operated by updating a member's personal profile to reflect certain actions the member had taken on websites belonging to companies that had contracted with Facebook to participate in the Beacon program. Thus, for example, if a member rented a movie through the participating website Blockbuster.com, Blockbuster would transmit information about the rental to Facebook, and Facebook in turn would broadcast that information to everyone in the member's online network by publishing to his or her personal profile.

Although Facebook initially designed the Beacon program to give members opportunities to prevent the broadcast of any private information, it never required members' affirmative consent. As a result, many members complained that Beacon was causing publication of otherwise private information about their outside web activities to their personal profiles without their knowledge or approval. Facebook responded to these complaints (and accompanying negative media coverage) first by releasing a privacy control intended to allow its members to opt out of the Beacon program fully, and then ultimately by discontinuing operation of the program altogether.

Unsatisfied with these responses, a group of nineteen plaintiffs filed a putative class action in federal district court against Facebook and a number of other entities that operated websites participating in the Beacon program. The class-action complaint alleged that the defendants had violated various state and federal privacy statutes.1 Each of the plaintiffs' claims centered on the general allegation that Beacon participants had violated Facebook members' privacy rights by gathering and publicly disseminating information about their online activities without permission. The plaintiffs sought damages and a variety of equitable remedies for the alleged privacy violations.

Facebook denied liability and filed a motion to dismiss the plaintiffs' claims. Before the district court ruled on Facebook's motion, the parties elected to attempt settling their case through private mediation. The parties' initial settlement talks reached an impasse over whether Facebook should terminate the Beacon program permanently, but after two mediation sessions and several months of negotiations, Facebook and the plaintiffs arrived at a settlement agreement. In September of 2009, plaintiff Sean Lane submitted the parties' finalized settlement agreement to the district court for preliminary approval.

The terms of the settlement agreement provided that Facebook would permanently terminate the Beacon program and pay a total of $9.5 million in exchange for a release of all the plaintiffs' class claims. Of the $9.5 million pay-out, approximately $3 million would be used to pay attorneys' fees, administrative costs, and incentive payments to the class representatives. Facebook would use the remaining $6.5 million or so in settlement funds to set up a new charity organization called the Digital Trust Foundation ("DTF"). The stated purpose of DTF would be to "fund and sponsor programs designed to educate users, regulators[,] and enterprises regarding critical issues relating to protection of identity and personal information online through user control, and the protection of users from online threats." The parties' respective counsel arrived at the decision to distribute settlement funds through a new grant-making organization, rather than simply give the funds to an existing organization, at the suggestion of the private mediator overseeing their negotiations. Neither Facebook's nor the plaintiffs' class counsel was comfortable with selecting in advance any particular non-profit or non-profits to receive the entirety of the settlement fund, so they acceded to the mediator's suggestion that Facebook set up a new entity whose sole purpose was to designate fund recipients consistent with DTF's mission to promote the interests of online privacy and security.

According to DTF's Articles of Incorporation, DTF would be run by a three-member board of directors. The initial three directors were Larry Magrid, a member of the federal government's Online Safety and Technology Working Group and several other online safety organizations; Chris Hoofnagle, director of the Information Privacy Programs at the Berkeley Center for Law and Technology and former director for an office of the Electronic Privacy Information Center; and most relevant here, Timothy Sparapani, Facebook's Director of Public Policy and former counsel for the American Civil Liberties Union. The Articles of Incorporation further provided that all of DTF's funding decisions had to be supported by at least two members of the three-member board of directors but that the plan for succession of directors required unanimous approval. Finally, the Articles of Incorporation provided that DTF would be strictly a grant-making organization and could not engage in lobbying or litigation.

The settlement agreement also provided for the creation of a Board of Legal Advisors within DTF, which would consist of counsel for both the plaintiff class and Facebook. The purpose of the Board of Legal Advisors would be to advise and monitor DTF to ensure that it acted consistently with its mission as articulated in the settlement agreement.

After a hearing, the district court certified the plaintiff class for settlement purposes and preliminarily approved the parties' proposed settlement. The settlement class consisted of all Facebook members who had visited the website of a Beacon participant that transmitted information about the members' activity to Facebook during the relevant period. The district court ordered Facebook to identify all class members and to send the class notification of the settlement. Following that order, Facebook identified 3,663,651 class members, to whom it provided notice of the settlement in several ways. The principal method was to send an e-mail to the class members. Facebook also posted a notice of the settlement in the "Updates" section of members' personal Facebook accounts and published a separate notice in the national edition of the newspaper USA Today. All forms of notice directed class members to a website and toll-free number that contained information about the settlement.

Also pursuant to the district court's order, notice to class members informed them of their right to opt out of the lawsuit and settlement, and to file any written comments or objections with the district court before final approval. At the conclusion of the notice period, 108 class members had opted out of the settlement, and four had filed written objections. The four class members who decided to remain in the lawsuit but file objections to the settlement were Ginger McCall, Megan Marek, Benjamin Trotter, and Patricia Burleson (collectively "Objectors").

Following a final settlement approval hearing in which the district court heard from both the parties and Objectors, the district court entered an order certifying the settlement class and approving the class settlement. The district court dismissed the plaintiffs' class action consistent with the settlement agreement, and it maintained jurisdiction over implementation of the settlement. The district court also awarded class counsel attorneys' fees in a separate order. The amount of the attorneys' fees was calculated at $2,322,763 under the "lodestar" method, meaning that the court multiplied the number of hours class counsel reasonably spent on the case by a reasonable hourly rate. That amount was combined with costs for a total attorneys' fees award of $2,364,973, which represented less than one-third of the full $9.5 million settlement amount.

Objectors now appeal, contending that the district court abused its discretion in approving the parties' settlement. We have jurisdiction pursuant to 28 U.S.C. § 1291, and we affirm.

II

A district court's approval of a class-action settlement must be accompanied by a finding that the settlement is "fair, reasonable, and adequate." Fed. R.Civ.P. 23(e). Appellate review of the district court's fairness determination is "extremely limited," and we will set aside that determination only upon a "strong showing that the district court's decision was a clear abuse of discretion." See Hanlon v. Chrysler Corp., 150 F.3d 1011, 1026-27 (9th Cir.1998) (holding that district court should have broad discretion because it "is exposed to the litigants, and their strategies, positions and proof") (internal quotations omitted).

Both the district court and this court must evaluate the fairness of a settlement as a whole, rather than assessing its individual components. See id. at 1026. As our precedents have made clear, the question whether a settlement is fundamentally fair within the meaning of Rule 23(e) is different from the question whether the settlement is perfect in the estimation of the reviewing court. See id. at 1027. Although Rule 23 imposes strict procedural requirements on the approval of a class settlement, a district court's only role in reviewing the substance of that settlement is to ensure that it is "fair, adequate, and free from collusion." See id.

A number of factors guide the district court in making that determination, including:

Id. at 1026 (hereinafter the "Hanlon factors"). Additionally, when (as here) the settlement takes place before formal class certification, settlement approval requires a "higher standard of fairness." See id. The reason for more exacting review of class settlements reached before formal class certification is to ensure that class representatives and their counsel do not secure a disproportionate benefit "at the expense of the unnamed plaintiffs who class counsel had a duty to represent." See id. at 1027; see also In re Gen. Motors Corp. Pick-Up Truck Fuel Tank Prods. Liab. Litig., 55 F.3d 768, 787 (3d Cir.1995) (explaining that "[w]ith less information about the class" at the early stage before formal class certification, the court "cannot as effectively monitor for collusion, individual settlements, buy-offs (where some individuals use the class action device to benefit themselves at the expense of absentees), and other abuses"). Accordingly, when reviewing a district court's approval of a class settlement reached before formal class certification, we will not affirm if it appears that the district court did not evaluate the settlement sufficiently to account for the possibility that class representatives and their counsel have sacrificed the interests of absent class members for their own benefit.

The settlement in this case provides for a cy pres remedy. A cy pres remedy, sometimes called "fluid recovery," Mirfasihi v. Fleet Mortg. Corp., 356 F.3d 781, 784 (7th Cir.2004), is a settlement structure wherein class members receive an indirect benefit (usually through defendant donations to a third party) rather than a direct monetary payment. As we recently recognized, the "cy pres doctrine allows a court to distribute unclaimed or non-distributable portions of a class action settlement fund to the `next best' class of beneficiaries." Nachshin v. AOL, LLC, 663 F.3d 1034, 1036 (9th Cir.2011). For purposes of the cy pres doctrine, a class-action settlement fund is "non-distributable" when "the proof of individual claims would be burdensome or distribution of damages costly." See id. at 1038 (quoting Six Mexican Workers v. Ariz. Citrus Growers, 904 F.2d 1301, 1305 (9th Cir. 1990)). The district court's review of a class-action settlement that calls for a cy pres remedy is not substantively different from that of any other class-action settlement except that the court should not find the settlement fair, adequate, and reasonable unless the cy pres remedy "account[s] for the nature of the plaintiffs' lawsuit, the objectives of the underlying statutes, and the interests of the silent class members...." Nachshin, 663 F.3d at 1036.

III

Objectors challenge the district court's conclusion that the settlement in this case was "fair, reasonable, and adequate" within the meaning of Rule 23(e). The district court arrived at that determination after considering Objectors' written statements and holding a fairness hearing where it provided Objectors an opportunity to be heard. The district court accompanied its fairness conclusion with findings of fact, which included the court's application of the eight Hanlon factors to the parties' settlement agreement.

Weighing those factors, the district court found that the settlement should be approved on the basis of the following: (1) reliance on novel legal theories and unclear factual issues undermined the strength of the plaintiffs' case; (2) the complex nature of the plaintiffs' claims increased the risk and expense of further litigation; (3) the class action could be decertified at any time, which "generally weighs in favor of approving a settlement"; (4) "[i]n light of [the] litigation risks and in the context of settlement claims involving infringment of consumers' privacy rights," the class's $9.5 million recovery was "substantial" and "directed toward a purpose closely related to Class Members' interests in this litigation"; (5) the parties had engaged in significant investigation and informal discovery and research, which in addition to information about Beacon that was already publicly known enabled the plaintiff class to "make an informed decision with respect to settlement, even though formal discovery" had not yet been completed; (6) the settlement was "only achieved after intense and protracted arm's-length negotiations conducted in good faith and free from collusion," and that class counsel had "reasonably concluded that the immediate benefits represented by the Settlement outweighed the possibility — perhaps remote — of obtaining a better result at trial"; (7) no government agencies voiced objections or otherwise announced actions arising out of Facebook's Beacon program; and (8) only four class members objected and "slightly more than 100" from a class of over 3.6 million opted out of the settlement.

Objectors raise two issues in opposition to the district court's fairness findings. The first relates to the settlement agreement's provision for a cy pres remedy. The second relates to the overall amount of the settlement. Objectors also raise the ancillary argument that notice to class members concerning the settlement was inadequate. We address each of these issues in turn.

1

Objectors' first and strongest objection to the settlement goes to the structure of DTF, the organization that would distribute cy pres funds under the settlement agreement. Objectors contend that the presence of Tim Sparapani, Facebook's Director of Public Policy, on DTF's board of directors creates an unacceptable conflict of interest that will prevent DTF from acting in the interests of the class. Citing Six Mexican Workers, Objectors claim that the settling parties' decision to disburse settlement funds through an organization with such structural conflicts does not provide the "next best distribution" of those funds and thus is categorically an improper use of the cy pres remedy.

We disagree. Objectors' argument misunderstands the cy pres doctrine and the principle from our case law that a cy pres remedy must provide the "next best distribution" absent a direct monetary payment to absent class members. We do not require as part of that doctrine that settling parties select a cy pres recipient that the court or class members would find ideal. On the contrary, such an intrusion into the private parties' negotiations would be improper and disruptive to the settlement process. See Hanlon, 150 F.3d at 1027. The statement in Six Mexican Workers and elsewhere in our case law that a cy pres remedy must be the "next best distribution" of settlement funds means only that a district court should not approve a cy pres distribution unless it bears a substantial nexus to the interests of the class members — that, as we stated in Nachshin, the cy pres remedy "must account for the nature of the plaintiffs' lawsuit, the objectives of the underlying statutes, and the interests of the silent class members...." 663 F.3d at 1036.2

The cy pres remedy in this case properly accounts for the factors outlined in Nachshin. Objectors concede that direct monetary payments to the class of remaining settlement funds would be infeasible given that each class member's direct recovery would be de minimis. Objectors also do not dispute that DTF's distribution of settlement funds to entities that promote the causes of online privacy and security will benefit absent class members and further the purposes of the privacy statutes that form the basis for the class-plaintiffs' lawsuit. Unlike the cy pres remedies we disapproved in Nachshin and Six Mexican Workers, there is no issue in this case about whether the connection between the cy pres recipients and the absent class members is too tenuous, either because the cy pres entities' missions are unrelated to the class's interests or because their geographic scope is too limited. See Six Mexican Workers, 904 F.2d at 1308; Nachshin, 663 F.3d at 1040. The cy pres remedy the settling parties here have devised bears a direct and substantial nexus to the interests of absent class members and thus properly provides for the "next best distribution" to the class.

We find no substance in Objectors' claim that the presence of a Facebook employee on DTF's board of directors categorically precludes DTF from serving as the entity that will distribute cy pres funds. As the "offspring of compromise," Hanlon, 150 F.3d at 1027, settlement agreements will necessarily reflect the interests of both parties to the settlement, including those of the defendant. Defendants often insist on certain concessions in exchange for monetary payments or other demands plaintiffs make, and defendants can certainly be expected to structure a settlement in a way that does the least harm to their interests. Here, in exchange for its promise to pay the plaintiff class approximately $9.5 million, Facebook insisted on preserving its role in the process of selecting the organizations that would receive a share of that substantial settlement fund by providing that one of its representatives would sit on DTF's initial board of directors, and the plaintiffs readily agreed to this condition. That Facebook retained and will use its say in how cy pres funds will be distributed so as to ensure that the funds will not be used in a way that harms Facebook is the unremarkable result of the parties' give-and-take negotiations,3 and the district court properly declined to undermine those negotiations by second-guessing the parties' decision as part of its fairness review over the settlement agreement.

We also reject Objectors' claim that the settlement agreement's cy pres structure is impermissible because the parties elected to create a new grant-making entity, DTF, rather than give cy pres funds to an already-existing online privacy organization. Again citing Six Mexican Workers, Objectors argue that DTF has "no substantial record of service" and is therefore inherently disfavored as a cy pres recipient. But we have never held that cy pres funds must go to extant charities in order to survive fairness review, and a settlement agreement that provides for the formation of a new grant-making organization is not subject to a more stringent fairness standard. The reason we found it relevant in Six Mexican Workers that the charity organization designated to receive cy pres funds had no "substantial record of service" was that there was no way of knowing whether the organization would use the funds to the benefit of class members. See Six Mexican Workers, 904 F.2d at 1308. Here, there is no such worry, because the settlement agreement and DTF's Articles of Incorporation tell us exactly how funds will be used — to "fund and sponsor programs designed to educate users, regulators[,] and enterprises regarding critical issues relating to protection of identity and personal information online through user control, and the protection of users from online threats."4 As we have explained, that mission statement provides the requisite nexus between the cy pres remedy and the interests furthered by the plaintiffs' lawsuit consistent with the principles we announced in Nachshin.

Objectors' contention that the settling parties were prohibited from creating DTF to disburse cy pres funds is without merit, and the district court did not abuse its discretion in so concluding.

2

Objectors' second argument on appeal is that the district court did not sufficiently evaluate the plaintiffs' claims and compare the value of those claims with the class's $9.5 million recovery in the settlement agreement. Objectors contend that the value of the plaintiffs' claims was in fact greater than the $9.5 million the plaintiffs settled for, in large part because some unidentified number of the class members may have a claim under the Video Privacy Protection Act ("VPPA"). The VPPA prohibits any "video tape service provider" from disclosing "personally identifiable information" about one of its consumers, and it provides for liquidated damages in the amount of $2,500 for violation of its provisions. 18 U.S.C. §§ 2710(b) and 2710(c)(2). Objectors contend that the district court was not sufficiently mindful of the possibility that the class's VPPA claims would yield a high recovery at trial, and that the court would not have approved a settlement of $9.5 million if it had paid the proper attention to that possibility.

As an initial matter, we reject Objectors' argument insofar as it stands for the proposition that the district court was required to find a specific monetary value corresponding to each of the plaintiff class's statutory claims and compare the value of those claims to the proffered settlement award. While a district court must of course assess the plaintiffs' claims in determining the strength of their case relative to the risks of continued litigation, see Hanlon, 150 F.3d at 1026, it need not include in its approval order a specific finding of fact as to the potential recovery for each of the plaintiffs' causes of action. Not only would such a requirement be onerous, it would often be impossible — statutory or liquidated damages aside, the amount of damages a given plaintiff (or class of plaintiffs) has suffered is a question of fact that must be proved at trial. Even as to statutory damages, questions of fact pertaining to which class members have claims under the various causes of action would affect the amount of recovery at trial, thus making any prediction about that recovery speculative and contingent.

Relatedly, the district court was not required to include among its findings specific commentary on each of the plaintiffs' five statutory claims. All of the plaintiffs' claims arise under similar privacy statutes, and as Facebook correctly points out, the plaintiffs' likelihood of success with regard to each of those claims depends on the same basic legal theories and factual issues. The district court acted properly in evaluating the strength of the plaintiffs' case in its entirety rather than on a claim-by-claim basis. See Hanlon, 150 F.3d at 1026.

Moreover, the record contradicts Objectors' general argument that the district court did not meaningfully account for the potential value of the plaintiffs' claims, including any claims under the VPPA. Both before and after the final settlement approval hearing, the district court specifically addressed the possibility that the presence of VPPA claims among some class members might affect the class settlement. In its order preliminarily approving the settlement, the district court notified the parties that "final approval will require a sufficient showing that terms of the settlement are reasonable, specifically in light of the claims under the VPPA, and the apparent availability of statutory penalties thereunder" (emphasis added). Following the district court's instructions, the parties did address the VPPA issue in their briefing and arguments at the final approval hearing. The district court also heard from Objectors at that hearing, who again argued that the settlement was too low in light of the possibility of recovery under the VPPA.

The district court rejected that argument. It first observed that Objectors had not "brought to the Court's attention any cases in which plaintiffs have been awarded multiple liquidated damages," which if available would likely increase the class's potential recovery under the VPPA substantially (even if only a small number of class members had VPPA claims). The district court further noted that bringing the VPPA claims to trial would involve significant risk for the class given that the plaintiffs' claims relied on "novel legal theories" and "vigorously disputed" factual issues concerning the Beacon program. And although the district court did not mention it in its approval order, the parties had presented evidence to the court that Blockbuster, one of the only defendants that might qualify as a "video tape service provider" and therefore be subject to liability under the VPPA, was on the verge of bankruptcy, likely making any substantial damages against it annihilative. Based on its consideration of these factors, the district court concluded that the "$9.5 million offered in settlement is substantial."

That conclusion was not an abuse of the district court's broad discretion. A $9.5 million class recovery would be substantial under most circumstances, and we see nothing about this particular settlement that undermines the district court's conclusion that it was substantial in this case. Objectors are no doubt correct that the VPPA claims of some class members might prove valuable if successful at trial, but that does not cast doubt on the district court's conclusion as to the fairness and adequacy of the overall settlement amount to the class as a whole. It is an inherent feature of the class-action device that individual class members will often claim differing amounts of damages — that is why due process requires that individual members of a class certified under Rule 23(b)(3) be given an opportunity to opt out of the settlement class to pursue their claims separately, as were the class members in this case. See Hanlon, 150 F.3d at 1024. But a class-action settlement necessarily reflects the parties' pre-trial assessment as to the potential recovery of the entire class, with all of its class members' varying claims. So even if some of the class members in this case would have successful claims for $2,500 in statutory damages under the VPPA, those individuals represent, to use the candid phrasing of Objectors, "only a fraction of the 3.6 million-person class." Their presence does not in itself render the settlement unfair or the $9.5 million recovery among all class members too low.5

Objectors rely significantly on Molski v. Gleich, 318 F.3d 937, 949 (9th Cir.2003) overruled on other grounds by Dukes v. Wal-Mart Stores, Inc., 603 F.3d 571 (9th Cir.2010), in claiming that the cy pres remedy here "did not adequately protect the interests of the class," but that case does not support Objectors' argument. Molski involved a settlement that required the defendant to pay $195,000 in cy pres funds in exchange for a release of all the disability-related claims of a large class. 318 F.3d at 943-44. The district court in Molski had certified a mandatory settlement class under Rule 23(b)(2) without providing class members an opportunity to opt out of the settlement. Id. at 947. In addition to holding that the inability to opt out of the settlement violated class members' due process rights, we held that "use of the cy pres award was inappropriate" under the circumstances because the parties had not made any showing that direct distribution of settlement funds to the class would be burdensome or costly. Id. at 954-55. We also found "troubling" that the class's recovery under the settlement was so low relative to the high number of potential class members. See id.

Unlike the $195,000 cy pres fund in Molski, the settlement in this case provides for a substantial $9.5 million pay-out by Facebook for the benefit of the class and thus does not present a situation in which class representatives and counsel accepted their respective fees as a quid pro quo for quietly going away while the class receives virtually nothing. See id. at 953-54. Also fundamentally different is that class members here received notice and were given the opportunity to opt out of the settlement. And, most essentially, there is no dispute that it would be "burdensome" and inefficient to pay the $6.5 million in cy pres funds that remain after costs directly to the class because each class member's recovery under a direct distribution would be de minimis. See id. at 955. These features distinguish the present case from Molski and help to account for why the latter was one of the "rare" cases where we have intruded into the discretion of the district court by setting aside its determination that a settlement agreement is fundamentally fair. See Staton v. Boeing Co., 327 F.3d 938, 960-61 (9th Cir.2003).

The record here convincingly establishes that the district court accounted for the potential value of the VPPA claims of some class members, and the district court's review of the circumstances surrounding the settlement was sufficiently comprehensive to ensure that class representatives and their counsel did not throw absent class members under the proverbial bus to secure a disproportionate benefit for themselves. See Hanlon, 150 F.3d at 1027. That review was accordingly compliant with this circuit's requirement that the district court apply heightened review to a class-action settlement reached before formal certification. See id. at 1026. This is particularly manifest in that the district court's detailed approval order included the specific factual finding that the settlement agreement "was only achieved after intense and protracted arm's-length negotiations conducted in good faith and free from collusion." Objectors have not made any showing, let alone a "strong" one, that this or any of the district court's other findings was erroneous or amounted to a "clear abuse of discretion." See id. at 1027.

Finally, the litigants devote several pages of briefing to a dispute over whether the settlement agreement's provision mandating the permanent termination of the Beacon program provided any meaningful relief to the plaintiff class. Specifically, Objectors argue that Facebook's promise to terminate Beacon is "illusory" because the original program was non-operational at the time of the settlement agreement and thus already "effectively terminated." In light of our holding affirming the district court's conclusion that the $9.5 settlement award substantially furthers the interests of the class, Objectors' argument that Facebook's promise to terminate Beacon provides no meaningful relief is of little moment, and in any event we find that it is without merit. Even assuming Objectors' premise that Beacon was already effectively terminated, absent a judicially-enforceable agreement, Facebook would be free to revive the program whenever it wanted. It is thus false to say that Facebook's promise never to do so was illusory.

We affirm the district court's holding that the settlement was fundamentally fair.

IV

Objectors argue additionally that the notice provided to class members during the opt-out period was insufficient because it did not describe the value of the plaintiffs' statutory claims and "did not accurately describe what the class members would receive in exchange for the release" of those claims. Objectors argue in particular that the notice should have included a description of the VPPA statute, that it should have alerted class members that a Facebook employee would be on the board of the organization distributing cy pres funds, and that its reference to Facebook's promise to terminate Beacon was misleading because Beacon was already dormant.

We disagree. Notice provided pursuant to Rule 23(e) must "generally describe[] the terms of the settlement in sufficient detail to alert those with adverse viewpoints to investigate and to come forward and be heard." Rodriguez v. West Publ'g Corp., 563 F.3d 948, 962 (9th Cir. 2009) (internal quotations omitted). That standard does not require detailed analysis of the statutes or causes of action forming the basis for the plaintiff class's claims, and it does not require an estimate of the potential value of those claims. See id. (notice need not include "expected value of fully litigating the case"). Nor is there any particular requirement that notice in a class-action settlement involving a cy pres remedy name the individuals sitting on the cy pres recipient's board of directors, even if one of those individuals has some association with the defendants in the case. Finally, for the same reasons we reject Objectors' argument that Facebook's promise to terminate Beacon was illusory, there was nothing misleading about referencing that promise in the class notice.

We agree with the district court that the notice in this case adequately apprised class members of all material elements of the settlement agreement and therefore complied with the requirements of Rule 23(e).

V

Ultimately, we find little in Objectors' opposition to the settlement agreement beyond general dissatisfaction with the outcome. That dissatisfaction may very well be legitimate insofar as Objectors would have acted differently had they assumed the role of class representatives. But while Objectors may vigorously disagree with the class representatives' decision not to hold out for more than $9.5 million or insist on a particular recipient of cy pres funds, that disagreement does not require a reviewing court to undo the settling parties' private agreement. The district court properly limited its substantive review of that agreement as necessary to determine that it was "fair, adequate, and free from collusion." See id.

AFFIRMED.

KLEINFELD, Senior Circuit Judge, dissenting:

I respectfully dissent. This settlement perverts the class action into a device for depriving victims of remedies for wrongs, while enriching both the wrongdoers and the lawyers purporting to represent the class.

A. The Facts.

1. "Beacon."

Millions of people connect themselves to their "friends" on Facebook. Some Facebook "friends" are friends in the traditional sense, people we know and like. Some are more in the nature of contacts, or acquaintances, or people we think may want to see what we post. For people who regularly use Facebook to communicate, "friends" may merely be their address book. The lead plaintiff in this case, Sean Lane, had over 700 Facebook "friends." Facebook operates like a bulletin board, so that "friends" can see whatever a user chooses to post and not make private.

Facebook is "free," furnished without a subscription price. The company makes money by selling advertising. To make such sales more lucrative, Facebook started a program called "Beacon" in November 2007. Like an actual beacon, the program shone light to make something easier to see: in this case, a user's "friends" could see whatever he had bought from companies that paid Facebook to participate in Beacon. Over forty companies signed up for Beacon, including Blockbuster, a movie retailer, Zappos, a shoe and clothing retailer, and Overstock.com, a discounter. If a Facebook user rented a movie from Blockbuster, for example, Facebook told all his friends what movie he had rented. Facebook told retailers, "Facebook Beacon enables your brand or business to gain access to viral distribution within Facebook. Stories of a user's engagement with your site.... will act as word-of-mouth promotion for your business and may be seen by friends who are also likely to be interested in your product."

Many Facebook users strongly objected to losing the privacy of their purchases. After all, people ordinarily post on their Facebook page only what they want to post, and they had not elected to tell all their "friends" what they had just bought. Some people buy things on the internet precisely because they want more privacy than they would have at a local store. Beacon took away their privacy, and broadcast their purchases to people who users wanted to remain in the dark.

Worse, Facebook made it very hard for users to avoid these broadcasts. The user had to actively opt out. And opting out required video game skills. The user would get a pop-up on his screen asking whether he wanted to opt out, but the pop-up would disappear in about ten seconds. Too slow reading the pop-up or clicking the mouse, and all a user's "friends" would know exactly what he had bought. Since the pop-up disappeared so quickly, someone looking at another window, or answering the phone, or just not paying attention, would likely not even be aware of the opt-out option before it disappeared.

Plaintiff Sean Lane alleges in the complaint that he bought a ring from Overstock.com as a surprise for his wife, but before he gave it to her, Facebook ruined the surprise by spreading the news to his over 700 "friends," including many alumni in his college class. Ginger McCall states that her video rentals at Blockbuster were disclosed to all her "friends." Of the vast number of people whose purchases were broadcast, no doubt some suffered embarrassment, and some suffered damage to employment, business, or personal relationships. Some Blockbuster rentals doubtless included erotica, some Overstock.com purchases probably included gifts meant to look more expensive than they were, and some Zappos purchases were probably more extravagant than purchasers' spouses were aware. Someone who had told her college classmate that she could not attend her wedding because she could not afford the plane fare could lose a friend when Facebook told her classmate that she'd bought $400 shoes. Mr. Lane complains that his wife asked him about his ring purchase before he gave it to her, ruining his Christmas gift to her. His wife might also have been less impressed by the ring than he had hoped, since she and all his other friends could click a link and see that he had bought it cheaply — good for advertising Overstock.com, bad for advertising Mr. Lane's generosity.

Many users' private purchases were exposed, and over 50,000 complained. Within a few weeks (long before this lawsuit was filed), Facebook eliminated the opt-out Beacon program. Facebook changed it to an opt-in program, so that users did not need to maintain video game alertness to avoid disclosure to all their friends. In the opt-in version of Beacon, purchases made in private stayed private unless the user expressly allowed Facebook to publicize them. One of the objectors to the settlement, Ginger McCall, says her movie rentals were disclosed even after Beacon had supposedly changed to an opt-in, and no findings have been made on whether the opt-in worked or was tricky to operate.

2. The Settlement.

This lawsuit was filed in August 2008, about eight months after the opt-out version of Beacon had ended. The complaint challenged only the opt-out program that had lasted for a few weeks, not the opt-in version that had been in place since then. The parties mediated and settled, all before any class was certified. They agreed to end Beacon, both opt-in as it then was, and opt-out as it had been originally.

The settlement agreement approved by the district court (mistakenly, in my view) greatly changed the class aspect of the case. First, the parties agreed to certify the class for purposes of settlement. Second, they agreed to expand it far beyond what the complaint had sought. The complaint sought damages only for users affected during the few weeks when they had to opt out, but the settlement expanded the class to include everyone affected during the much longer opt-in period. Since the members of the class got no money from the settlement, the effect of certification and expansion was to bar any claims the expanded class might have, not to provide more people with recompense. In exchange for nothing, class members were barred from suing Facebook, Blockbuster, Overstock.com, or any of the other defendants for any claims arising from or relating to Beacon, "including, without limitation, arising from or related to data gathered from Beacon."

The majority states that Facebook promised never to revive the Beacon program, but this is not quite right. Facebook remained free to revive the program, even the cancelled version under which the subscriber had only a few seconds to opt out. The only limitation the settlement imposed was that Facebook had to call the Beacon program by some other name. The agreement said that Facebook would terminate "the Beacon Program," and defined "Beacon" to mean "the program launched by Facebook on November 6, 2007 and all iterations thereof bearing the `Beacon' name" (emphasis added). The district judge asked about this term, and plaintiffs' attorney expressly conceded that Facebook was free to reinstitute the same program under a different name. "[T]he problem was when you tried to describe the functionality and you preclude Facebook from using that functionality going forward, it becomes truly problematic and becomes impossible to reach an agreement because you're limiting their ability to run their business.... At the end of the day, we could not reach agreement with defendants regarding limiting their future actions as a corporation." That was an on the record concession that the injunction meant as little as it said, and Facebook remained free to do what it had done before, under a different name. The injunctive relief the class received was no relief at all, not even a restriction on future identical conduct.

Facebook users who had suffered damages from past exposure of their purchases got no money, not a nickel, from the defendants. Even those who had rented videos, and were arguably entitled to statutory damages of $2,500 for each disclosure,1 got nothing. Class counsel, on the other had, got millions. Plaintiffs' lawyers and Facebook agreed that Facebook would not object to attorneys' fees up to one third of what they called the "settlement fund." One third would be a fee of $3,166,667. The fee would come out of the "settlement fund" and would not be in addition to it, so Facebook had no economic interest in reducing the amount. The fee actually approved by the district court was $2,322,763 plus costs of $42,210.58, 25% of the "settlement fund." That $2.3 million payment was for getting their clients nothing and barring all the claims of a vastly broadened class.

Not a cent of the remaining "settlement fund" money would go to the Facebook users on whose behalf class counsel purportedly settled. The only exceptions were $10,000 to Mr. Lane, $5,000 each to two others, and $1,000 each to the other 19 named plaintiffs, amounting to $39,000 for the few people in the class who presumably had personally agreed to have class counsel represent them.

The remaining millions were to go to a new "privacy foundation" that did not yet exist. The board of the new foundation would be three directors to be agreed upon by Facebook and class counsel, or if they disagreed one chosen by each and the third chosen by those two. Under the agreement, all three directors could come from the Facebook advertising and sales staff if class counsel and Facebook so chose. The board of directors of this "privacy foundation" was to be advised by Facebook's own lawyer and class counsel. The agreement provided that the "privacy foundation" was to use its millions to "fund projects and initiatives that promote the cause of online privacy, safety, and security" however its Facebook-friendly board chose.

B. Analysis.

The class action rule2 was designed to facilitate lawsuits where individuals' or small groups' judgments would not add up to enough money to justify hiring lawyers, but judgments for large numbers of similarly situated victims of misconduct would. "The policy at the very core of the class action mechanism is to overcome the problem that small recoveries do not provide the incentive for any individual to bring a solo action prosecuting his or her rights. A class action solves this problem by aggregating the relatively paltry potential recoveries into something worth someone's (usually an attorney's) labor."3

This procedural device has obvious attendant risks, because class counsel's "clients" are not clients at all in the traditional sense; they do not hire the lawyer, they do not agree on a fee with him, and they do not control whether he settles their case. They are in no position to prevent class counsel from pursuing his own interests at their expense.4 The named plaintiffs, those who actually have some chance of directing their lawyers, typically get amounts of cash without much relation to their individual damages, so their incentives align more with class counsel than with their fellow class members.

Defendant and class counsel, in any class action, have incentives to collude in an agreement to bar victims' claims for little or no compensation to the victims, in exchange for a big enough attorneys' fee to induce betrayal of the interests of the purported "clients." The defendant's agreement not to oppose some amount for the fee creates the same incentive as a payment to a prizefighter to throw a fight. A real client may refuse a settlement that is bad for him but benefits his lawyer, but a large class of unknown individuals lacks the knowledge or authority to say no. It is hard to imagine a real client saying to his lawyer, "I have no objection to the defendant paying you a lot of money in exchange for agreement to seek nothing for me." "The absence of individual clients controlling the litigation for their own benefit creates opportunities for collusive arrangements in which defendants can pay the attorneys for the plaintiff class enough money to induce them to settle the class action for too little benefit to the class (or too much benefit to the attorneys, if the claim is weak but the risks to the defendants high)."5

Rule 23 protects against these risks much as the courts have traditionally protected against similar risks when attorneys represent children, estates of deceased persons, and unknown persons, by requiring judicial approval of settlements. Approval and review, though, are a weak substitute for real clients, because judges know little about the case beyond what the lawyers tell them. That works much better when the lawyers are on different sides than when they are on the same side. Judges also may face an incentive problem, where a heavy docket cannot easily withstand the additional weight of a huge lawsuit that does not settle. Objectors provide a critically valuable service of providing knowledge from a different point of view, but one that is too often not used effectively. Our review process is supposed to assure that settlement of a class action, despite the risk of perverse incentives, is "fair, reasonable, and adequate"6 and that notice is given "in a reasonable manner"7 so that those bound by the settlement have an opportunity to be heard.

In this case, the process has failed. The attorneys for the class have obtained a judgment for millions of dollars in fees. The defendant, Facebook, has obtained a judgment that bars claims by millions of people victimized by its conduct. So have the other companies involved in Beacon. The victims, on the other hand, have obtained nothing. Under the settlement, Facebook even preserved the right to do the same thing to them again.

1. The Settlement is Unfair, Unreasonable, and Inadequate.

The factors for evaluating class action settlements8 are multifarious and indeterminate, but the cases have become less tolerant of settlements not beneficial to class members. We used to be extremely deferential when district courts approved settlements, as in Hanlon v. Chrysler Corp.,9 the 1998 case on which the majority relies. We have in the last few years become much less so, as in our recent decisions In re Bluetooth,10 Nachshin v. AOL, LLC,11 and Dennis v. Kellogg Co.12 We still exercise deferential review for abuse of discretion, but do so in light of what we rejected in Bluetooth, Nachshin, and Dennis. Review for abuse of discretion has never meant that we will affirm whatever a district court does.13

An extremely important qualification even in Hanlon was a "higher standard of fairness"14 when settlement is reached before a class is certified. In this case, not only was settlement reached before class certification, but the class certified for settlement purposes was far broader than the one sought when the case was filed. The Hanlon "higher standard of fairness" matters because of "the dangers of collusion between class counsel and the defendant."15 Bluetooth emphasizes the need for greater scrutiny of precertification settlement on behalf of a class.16 "Collusion may not always be evident on the face of a settlement, and courts therefore must be particularly vigilant not only for explicit collusion, but also for more subtle signs that class counsel have allowed pursuit of their own self-interests and that of certain class members to infect the negotiations."17

Collusion is far more likely before certification, and exponentially higher if the class is expanded as part of the settlement. Here is why. If a lawsuit is only on behalf of named plaintiffs, damages are limited to what they may properly receive, so if a case is reasonably defensible, a defendant may make a sound financial decision to defend. But if a vast class is certified, then even a meritless case may require a defendant to settle or bet all the money it has or can borrow for attorneys' fees, because even a very small chance of a very large verdict is too much to risk. Plaintiffs' counsel want certification, to make the damages enough to be worth the time and expense of the litigation. Defense counsel oppose it, to keep the risk down to a level where they can afford the risk of litigation. Because certification of a class may turn even a meritless plaintiff's case into a bet-the-company defendant's case, defendants usually vigorously oppose class certification, giving courts the benefit of adversarial presentations.

Once the parties agree to settle, and agree to certify a class, defendant's interests are reversed. Plaintiffs' counsel still have an interest in keeping a large class certified, because the larger the class, the higher the attorneys' fees are likely to be. But if the defendant will get a bar against claims, almost always a term of any settlement, the more people whose claims are barred the better. The risk of having to pay out a huge amount of money gets converted, by class certification, into a certainty that vast numbers of people will be unable to sue the defendant. So when settling before class certification, and agreeing upon class certification as part of the settlement, both sides have the same incentive, to certify the class and make it as vast and all-encompassing as possible. It is a bonanza for the defendant if it can bar the claims not only of everyone in the class described in the complaint, but also of a much larger class on whose behalf more and different claims might have been asserted.

And that is just what happened here. The complaint claims wrongdoing against and damages to Facebook users during the few weeks of the opt-out period of "Beacon." The settlement bars claims of all the users during that period and during the much longer opt-in period. When they settled, Facebook and class counsel shared the same interest, as broad a class certification as possible. Ideally, from both the point of view of both sides' interests (attorneys' fees for one side, protection from claims for the other) the class would include everyone in the world, and bar all claims of any kind from the beginning of time to the present day. They came about as close to that as they plausibly could.

Bluetooth emphasizes that "clear sailing" agreements on attorneys' fees are important warning signs of collusion.18 We have a version of a clear sailing agreement here: Facebook's agreement not to oppose an attorneys' fees claim of up to $3,166,667. If, as here, the defendant agrees not to oppose an attorneys' fees claim, and defendants payout will be the same no matter how high the fee is, then both sides have an incentive to make the fee large enough to induce plaintiffs' counsel to sacrifice class interests to plaintiffs' attorneys' interests. Bluetooth holds that caution is especially necessary when, as here, members of the class receive no money, but class counsel receive a great deal of it.19 As the amount of the fee to which no objection will be made grows, especially if the fee will not affect the cost to the defendant, it makes economic sense (though not ethical sense) for plaintiffs' counsel to throw the fight for the money.

Strikingly, the settlement here goes even further than coupon settlements, where class members get only discounts if they buy again from the defendant claimed to have wronged them before, while their purported lawyers get huge amounts of money. Here the Facebook users get nothing at all, not even coupons. Every nickel of the remainder of the $9,500,000 after class counsel's cut, administrative costs, and incentive payments to the named plaintiffs, goes not to the victims, but to an entity partially controlled by Facebook and class counsel. The new entity, dressed to look good in old law French with its "cy pres" award and "non-profit" status, can spend the money to "educate" people about privacy on the internet, perhaps via some instructional videos on how to use all the privacy features available in Facebook.

Arguably, no harm would be done if all claims of wrongdoing to Facebook users from the Beacon program were frivolous. If their claims were worthless, then no wrong is done to them when those claims are barred and $9.5 million gets transferred to some lawyers they never met and a new entity not likely to benefit them. But that would denigrate the claims too far. There is reason to believe that Facebook needed the shield its $9.5 million bought. Facebook got customer complaints and bad publicity from the opt-out Beacon program. The class had colorable claims. Facebook had a good argument that it was not itself a "video tape service provider" under the federal statute entitling a customer to liquidated damages of $2,500 for disclosure of what videotape someone had rented from Blockbuster,20 but still had a risk of some sort of vicarious, joint, or "civil conspiracy" liability.21 If found liable, it was a deep pocket target for the punitive damages for which the statute expressly provides.22 And at least one federal district court has taken an expansive view of who is a "video tape service provider" prohibited from making disclosures.23 The facts alleged in the complaint stimulate a concern about the privacy of people's purchases on the internet and the use of customer information by Facebook.

Tort law tends to evolve to make actionable conduct widely seen as harmful, especially when the conduct is willful, as it was here. The plaintiffs' claims and the risk of that evolution of tort law were worth money to avoid, for Facebook. We cannot reasonably say that a risk worth $9.5 million to Facebook to avoid nevertheless had no value whatsoever to the potential claimants whose claims presented that risk. If Facebook users had no colorable claims, why would Facebook have paid $9.5 million to bar them?

2. The Settlement does not Meet our Standards for Cy Pres Awards.

Even if the $9.5 million number, the attorneys' fees, and the absence of any relief whatsoever to class members all were "fair, reasonable and adequate," the new foundation would still not satisfy the standards for cy pres awards. We held in Dennis v. Kellogg Co.24 quoting Staton v. Boeing Co.,25 that cy pres distributions present "a particular danger" that "incentives favoring pursuit of self-interest rather than the class's interests in fact influenced the outcome of negotiations."26

Cy pres traditionally was a means by which, say, a bequest to a charity no longer existing when a testator died might be given instead to a similar charity doing similar work. Thus a bequest to the Boys' Club might go to its replacement, the Boys' and Girls' Club. The doctrine has never meant simply that money for harm to someone would be given to someone else preferred by the defendant and plaintiff's attorney and perhaps by the court. We cautioned in Nachshin v. AOL that "When selection of cy pres beneficiaries is not tethered to the nature of the lawsuit and the interests of the silent class members, the selection process may answer to the whims and self interests of the parties, their counsel, or the court."27

The rules of judicial ethics have in many forms for over a hundred years prohibited judges from endorsing charities, because of the risk that lawyers and litigants will feel compelled to contribute to them.28 Too liberal an approach to cy pres means that a court may simply order, and not merely encourage, someone subject to its jurisdiction to give to a preferred charity. A defendant may prefer a cy pres award to a damages award, for the public relations benefit. And the larger the cy pres award, the easier it is to justify a larger attorneys' fees award. The incentive for collusion may be even greater where, as here, there is nothing to stop Facebook and class counsel from managing the charity to serve their interests and pay salaries and consulting fees to persons they choose.

Nachshin holds that the district court must ensure that a cy pres award targets the plaintiff class.29 Here it does not. Six Mexican Workers v. Arizona Citrus Growers30 holds that a district court must reject awards that provide "no reasonable certainty that any member will be benefitted."31 This one does not. We require an established record of performance by the charity of acts beneficial to people in the wronged class.32 The cy pres award in this case goes to a new entity with no past performance at all. For all we know it will fund nothing but an "educational program" amounting to an advertising campaign for Facebook. That would appear to satisfy the articles and bylaws, and Facebook, after all, together with class counsel and their nominees, will run it.

3. Notice.

We review adequacy of notice de novo, not deferentially.33 This is because notice is a matter of due process of law.34 If a person owns a claim, it is property, and the owner of the claim is constitutionally entitled not to have it taken from him except with reasonable notice and an opportunity to be heard. Notice in this case was inadequate, most obviously because the class was not sufficiently informed that Facebook itself might be in control of the money purportedly awarded on account of wrongs it committed against class members. The articles of incorporation and bylaws of the purportedly charitable foundation were posted online for the class to see only a week before the deadline to opt out of the settlement. Those documents said that "Tim Sparapani" would be on the three-person board, but failed to mention who he was, Facebook's own Director of Public Policy. Nor did the notice say that Facebook's counsel, Michael Rhodes, would sit on the foundation's legal advisory board. Class members would have had to look carefully at the settlement agreement and figure out that Mr. Rhodes, the man designated as a legal advisor on page twelve of the settlement agreement, was the same man listed as Facebook's attorney on page five. Class members dependant on the notice would have no idea that the money supposedly paid for wrongs to them was to be spent by agents of the purported wrongdoer.

Conclusion

The majority approves ratification of a class action settlement in which class members get no compensation at all. They do not get one cent. They do not get even an injunction against Facebook doing exactly the same thing to them again. Their purported lawyers get millions of dollars. Facebook gets a bar against any claims any of them might make for breach of their privacy rights. The most we could say for the cy pres award is that in exchange for giving up any claims they may have, the exposed Facebook users get the satisfaction of contributing to a charity to be funded by Facebook, partially controlled by Facebook, and advised by a legal team consisting of Facebook's counsel and their own purported counsel whom they did not hire and have never met.

Facebook deprived its users of their privacy. And now they are deprived of a remedy.