Opinion by Circuit Judge BERZON. Dissenting opinion by Circuit Judge TROTT.
BERZON, Circuit Judge.ORDER
The panel majority opinion filed November 26, 2002, is withdrawn and the attached opinion is ORDERED filed.
With the filing of the attached opinion, a majority of the panel has voted to deny defendants/appellees’ petition for rehearing. Judge Berzon has voted to deny the plaintiffs/appellees’ and defendants/appel-lees’ petitions for rehearing en banc and Judge Lay has so recommended. Judge Trott has voted to grant the petition for rehearing and to grant the petitions for rehearing en banc.
The full court was advised of the petitions for rehearing en banc. A judge of the court requested a vote on whether to rehear the matter en banc. The en banc request failed to receive a majority of the votes of the nonrecused active judges in favor of en banc consideration. Fed. R.App.P. 35.
The petition for rehearing and the petitions for rehearing en banc are DENIED.
OPINION
This case involves a consent decree in an employment discrimination class lawsuit. The action was brought in 1998 by a class of approximately 15,000 African-American employees of the Boeing Company (“Boeing” or “the Company”) against the Company. The decree requires Boeing to pay $7.3 million in monetary relief to the class, less reversions and an opt-out credit,1 and releases Boeing from race discrimination-related and other claims. It further provides for certain injunctive relief, although much of this relief appears to be largely precatory in nature. Finally, the decree *945awards to the lawyers for the class (“class counsel”) $4.05 million in attorneys’ fees.2
A group of class members objected to the proposed consent decree, arguing that the class fails to meet' the certification requirements of Fed.R.Civ.P. 23(a) (“Rule 28(a)”) for class actions and that the settlement contained in the decree is unfair, inadequate and unreasonable under Fed. R.Civ.P. 28(e) (“Rule 23(e)”). The district court approved the decree despite the objections, and the objectors appealed to this court. After oral argument, we requested supplemental briefs from the parties concerning the attorneys’ fees issues.
We hold that the district court acted within its discretion in certifying the case as a class action pursuant to Rule 23(a). We agree with the objectors, however, that the district court should not have approved the settlement agreement under Rule 23(e), because of several considerations relating to the award of attorneys’ fees and because of the structure of the damages payments established by the decree.
The parties negotiated the amount of attorneys’ fees as part of the settlement between the class and the Company. They included as a term of the proposed decree the amount of attorneys’ fees that class counsel would receive. The action falls under the terms of two fee-shifting statutes. By negotiating fees as an integral part of the settlement rather than applying to the district court to award fees from the fund created, Boeing and class counsel employed a procedure permissible if fees can be justified as statutory fees payable by the defendant.
Boeing and class counsel did not, however, seek to justify the attorneys’ fees on this basis but instead made a hybrid argument: They maintained that the award is an appropriate percentage of a putative “common fund” created by the decree even though common funds, as opposed to statutory fee-shifting agreements, usually do not isolate attorneys’ fees from the class award before an application is made to the court. The district court approved the fees on that common fund basis.
The incorporation of an amount of fees calculated as if there were a common fund as an integral part of the settlement agreement allows too much leeway for lawyers representing a class to spurn a fair, adequate and reasonable settlement for their clients in favor of inflated attorneys’ fees. We hold, therefore, that the parties to a class action may not include in a settlement agreement an amount of attorneys’ fees measured as a percentage of an actual or putative common fund created for the benefit of the class. Instead, in order to obtain fees justified on a common fund basis, the class’s lawyers must ordinarily petition the court for an award of fees, separate from and subsequent to settlement.
To assess the reasonableness of the attorneys’ fees awarded by the decree, the district court compared the amount of the fees to the amount of the putative common fund and determined what percentage of this fund the fee amount constituted. This comparison is a permissible procedure when, a court is determining the reasonableness of fees taken from a genuine common fund. In conducting the comparison, however, the district court included in the value of the putative fund the parties’ inexact, and quite probably inflated, estimate of the value of the proposed injunctive relief. Such relief should generally be excluded from the value of a common fund *946when calculating the appropriate attorneys’ fees award, as the benefit of that relief to the class members is most often not sufficiently measurable. The fact that counsel obtained injunctive relief in addition to monetary relief for their clients is, however, a relevant circumstance to consider in determining what percentage of the fund is reasonable as fees. We hold further, therefore, that parties ordinarily may not include an estimated value of undifferentiated injunctive relief in the amount of an actual or putative common fund for purposes of determining an award of attorneys’ fees.
Finally, the decree sets up a two-tiered structure for the distribution of monetary damages, awarding each class representative and certain other identified class members an amount of damages on average sixteen times greater than the amount each unnamed class member would receive. At least one person not a member of the class was provided a damages award. The record before us does not reveal sufficient justification either for the large differential in the amounts of damage awards or for the payment of damages to a nonmember of the class. On this ground as well, the district court abused its discretion in approving the settlement.
I. BACKGROUND
A. Lawsuit Filed and Settled
In September 1997, a group of African-American employees of the Company who believed that they were victims of race discrimination by Boeing consulted class counsel. Prior to the fifing of this lawsuit, forty-three African-American Boeing employees filed a lawsuit in March 1998 in federal court in Seattle, Washington, alleging individual claims of race discrimination in violation of 42 U.S.C. § 1981 and the state anti-discrimination law, Wash. Rev. Code § 49.60 et seq. (the “Seattle individual action”). Several months later, in June 1998, sixteen Boeing employees, including twelve plaintiffs from the Seattle individual action, filed this class action in the same court. The employees again alleged violations of § 1981 and the state anti-discrimination law but sought to represent both themselves and other similarly-situated African-American Boeing employees. The action alleged that Boeing’s promotion, compensation, and career development decisions were systematically discriminatory and that Boeing created and permitted a racially hostile work environment.
The plaintiffs amended their complaint on November 4, 1998. In the amended complaint thirty-two named plaintiffs seek to represent all African-American Boeing employees. The amended complaint alleges violations of Title VII of the Civil Rights Act of 1964, 42 U.S.C. § 2000e et seq., as well as violations of § 1981; it omits the state anti-discrimination claim but includes other state tort and contract causes of action. The named plaintiffs and over two hundred other Boeing employees each signed retainer agreements with class counsel. They agreed to pay class counsel an initial fee of $800 and to follow that payment with monthly payments of $200 each; the record contains several letters from class counsel to these individuals urging that the payments be brought up to date. Approximately $150,000 was raised in this manner.
Meanwhile, in July 1998, seven named plaintiffs filed a similar class action against Boeing in Philadelphia (the “Philadelphia class action”). The plaintiffs in this case moved in October 1998 to consolidate the two actions.
Soon thereafter, in early November, Boeing filed a motion to dismiss plaintiffs’ class claims. Also in November, class *947counsel and Boeing began settlement negotiations. Class counsel had met with numerous African-American Boeing employees before filing suit but almost no formal discovery had taken place by November. At the beginning of December, class counsel indicated frustration to their clients about the Company’s lack of responsiveness, characterizing “much of what Boeing has provided thus far as ‘junk,’ ” and adding that “Boeing has been unwilling to provide us with numerous documents we believe are pertinent to proving Boeing’s unlawful treatment of African-Americans.” The negotiations nonetheless proceeded rapidly for such a large class action at this early stage of litigation, with the result that, in January 1999, Boeing and class counsel announced that they had agreed to settle the lawsuit. The parties filed motions in district court for class certification and for preliminary approval of a consent decree.
On January 25, 1999, the district court provisionally certified the class and preliminarily approved the consent decree. The preliminary approval order required the Company to provide approved notice of the proposed decree to class members through newspaper publication, distribution using the Company’s payroll system for present employees, and first-class mailings to former employees. Two distributions of the notices were ultimately required because the first notices published and mailed were improper and had to be corrected. The notices explained that Boeing would pay the attorneys’ fees and costs and reported the total sum of money Boeing would pay under the decree, the amount to be paid for monetary awards to members of the class, and the amount ascribed to injunc-tive relief. Neither version expressly identified the amount of attorneys’ fees provided in the proposed decree.
B. Proposed Consent Decree
The proposed consent decree purports to resolve this case, the Seattle individual action, and the Philadelphia class action. The decree releases Boeing from liability for claims brought by any of the Company’s African-American employees in exchange for certain monetary and injunctive relief. In particular, Boeing is released by all class members from all existing claims for race discrimination (under any of the various discrimination laws) and for “negligent misrepresentation, fraud, detrimental reliance, promissory estoppel, or breach of contract,” without regard to whether such claims are in any way related to the alleged race discrimination. The period covered by the release is set according to the statute of limitations period of the state in which a class member resides and extends until the preliminary approval date of the decree. As a result, any claims arising before January 25, 1999 (the preliminary approval date) of the types covered by the decree and timely under the relevant state statute of limitations are barred.
The decree goes on to certify a settlement class pursuant to Fed.R.Civ.P. 23(b)(2) for purposes of equitable relief. That class consists of all African-Americans employed by Boeing from the beginning of the applicable limitations periods until the expiration of the decree (including new employees hired after the preliminary approval date of the decree). No opt-outs are allowed from the equitable relief class. The effect of this provision may be that no African-American employed by Boeing during the pertinent period, including new hires, can obtain any injunctive relief — reinstatement, promotion, or change in working conditions, for example — even if he or she opts out of the class for purposes of monetary relief and proves race discrimination in a sepa*948rate action.3
For purposes of monetary relief, the decree approves a Settlement Class pursuant to Fed.R.Civ.P. 23(b)(3), consisting of African-American Boeing employees employed from the beginning of the applicable limitations periods until the preliminary approval date of the decree, and allows members of that class to opt out of the monetary relief provisions. By the cut-off date of April 30, 1999, about 500 class members had opted out, including six named plaintiffs.
The class receives a total monetary award of $7.3 million. Out of the approximately 15,000-member class, a group of 264 individuals 4 — less than two percent of the class — made up of the named plaintiffs and other class members identified by class counsel as having actively participated in the litigation (together, the “individually identified recipients” or “IIRs”) is to receive $3.77 million, more than half the monetary award. The $3.77 million will be distributed among the IIRs in amounts established by class counsel, who credit the assistance of an independent claims adjuster for consultation on many, but not all, of the claims. There is ample evidence in the record that before retaining this claims adjuster class counsel extensively discussed specific award amounts with some IIRs. Moreover, the record indicates that class counsel made the final decisions concerning many of these designated payments.
The individual awards for the IIRs range from $5,000 to $50,000, with most of the class representatives receiving higher awards than the other IIRs, and average approximately $16,500. Based on our examination of records relating to the Wichita-based IIRs, the individuals singled out for HR settlement payments are for the most part the same people who signed individual retainers with class counsel that obligated them to pay monthly fees.
The remaining $3.53 million of monetary relief is to be distributed to the rest of the class (the “unnamed class members”). To receive an award, unnamed class members must submit a claim form to an independent claims arbitrator (hired by class counsel and approved by the district court), who will verify the validity of the claims against Boeing’s records and designate awards according to a detailed point system laid out in the decree and applicable only to the unnamed class members. Some 3,400 class members filed claims, so the average payment each unnamed class member would receive is approximately $1,000.
Boeing also agreed to pay $3.75 million to $3.85 million to class counsel for fees and expenses, as follow:
—$3 million for attorneys’ fees and costs (the parties agreed that class counsel had incurred approximately $126,000 in costs as of the preliminary approval date);
—up to $100,000 for explaining the decree to class members;5 and
—$750,000 for monitoring, administering, implementing, and defending the decree.
The decree also grants $200,000 to objectors’ counsel for their role in representing *949the putative class in the Philadelphia class action.
Finally, within three years of receiving final judicial approval of the decree, Boeing must spend an additional $8.65 million on expenses related to the approval and implementation of the decree. This $8.65 million would go toward the cost of providing notice to class members of the proposed settlement and toward the injunctive relief provided for in the decree, discussed below. The decree further provides that:
Such credited expenditures shall also include money spent by Boeing on diversity training and other programs designed to improve the cooperation between members of Boeing’s diverse workforce, to facilitate the advancement of African-Americans into first-level and higher-level management positions at Boeing, to prevent and/or resolve racial harassment concerns among the workforce, and/or to otherwise advance equal employment opportunity for African-American employees of Boeing.
Nothing in the decree requires that the credited amount be in addition to any amount of money Boeing was already planning to spend on such matters. The timing of such expenditures is within Boeing’s discretion, although the parties expressed the expectation that half the funds would be spent between the preliminary approval date and the first anniversary of the final approval.
The decree’s injunctive provisions are to be in effect for the three years following final judicial approval of the decree. The injunctive relief provided for in the decree is as follows:
(1)Boeing will not discriminate based on race or retaliate against employees for opposing race discrimination or participating in efforts to eradicate it. These general provisions mirror statutory prohibitions. However, “Court enforcement of this Decree shall not be utilized as a method for class members to litigate entitlement to individual relief for claims of alleged Race Discrimination,” and individual complaints of race discrimination “shall not be considered to raise an issue of compliance or non-compliance with this decree.”
(2) Boeing will meet annually with a three-person advisory committee chosen from among the class members to discuss “Settlement Class members’ viewpoints and concerns.” The members of the committee will bear their own expenses for attending the meetings.
(3) Boeing will hire one or more consultants “to assist it in developing and assessing the success of alternative and/or supplemental human resources systems designed to accomplish the objectives in this Section [describing the injunctive relief] of the Decree.” The consultant is to be chosen by Boeing and class counsel.6 The consultant is to investigate the degree to which the decree successfully addresses various of the class members’ concerns and to report back to Boeing and class counsel. Nothing in the decree requires Boeing to take any action in response to these reports or otherwise to take any action suggested by the consultant.
(4) Boeing — unilaterally—will develop and implement systems for providing information to hourly employees *950about the Company’s promotion systems and will develop and “pilot” a program designed to enable hourly employees to learn who received a particular promotion. Boeing is required to meet and confer with class counsel about the effectiveness of these programs once implemented but is not required to adopt any suggestions class counsel make or, with regard to the “pilot” promotion information program, to do anything more than “determine the feasibility of implementing that program, or comparable programs” throughout the Company.7
(5) Similarly, Boeing will develop a system whereby qualified but unsuccessful candidates for discretionary promotions will receive feedback and be directed to training or other steps that would make the candidates more competitive. Class counsel are to “monitor” this process, with no provision for any dispute resolution mechanism should class counsel conclude that the system is inadequate or ineffective.
(6) With regard to filling opportunities for temporary promotions (useful in providing experience relevant to desirable positions), Boeing “shall identify informal systems” to permit candidates to know about and be considered for such opportunities, and shall “meet and confer” with class counsel regarding such informal systems and related complaints. There is no requirement that Boeing change its behavior in response to any suggestions or objections by class counsel or any class member.
(7) Boeing “presently plans” to expand its First Level Management Selection Process (FLMSP) to all its operations over the first two years of the decree. The FLMSP, thus far a pilot program at Heritage Boeing8 locations, attempts to create a standardized, fair process for selecting first-level managers. If “Boeing decides not to implement FLMSP in certain portions of the Company’s operations, Boeing will advise Class Counsel of the alternative selection methods which will be utilized in such operations, and Class Counsel will provide feedback to Boeing regarding any systemic concerns about such alternative methods which they believe may impact upon the Settlement Class members.” Boeing can modify the FLMSP or eliminate it altogether; if it does so, Boeing must advise class counsel “and consider feedback provided by Class Counsel regarding such changes.”
(8) In 1998, in part in response to this litigation, Boeing developed new “Company-wide EEO Investigation Guidelines,” which, among other things, improve the time period for addressing internal discrimination complaints. Boeing will accept “feedback” from class counsel on the guidelines generally and on any modifications the company makes and “may” use the consultant’s services to refine these procedures.
*951(9) Boeing will continue the provisions of its existing harassment policy concerning race, or implement amended policies “reasonably designed to achieve the same effect” as the existing policy. Class counsel will have the opportunity to provide “feedback” on any modifications to that policy. However, “[individual [harassment] complaints shall not be considered to raise an issue of compliance or noncompliance with this Decree.”
C. Objections and Their Resolution
In April 1999, some members of the class filed objections to the proposed consent decree. The district court allowed limited discovery by the objectors, reviewed motions by all parties, and held two fairness hearings (but did not take any evidence at those hearings).
Among other matters, the objectors complained that class counsel could not have meaningfully assessed the value of class claims because of insufficient discovery; that the monetary relief was inadequate and unfairly distributed; that the injunctive relief would not result in concrete benefits to the class; that the court should not approve a single broad class, since the members of the putative class have divergent interests; that plaintiffs’ counsel are not fairly representing the plaintiffs because, inter alia, individual class members were promised monetary relief in order to secure their support of the decree; that the notice provided to class members was deficient; and that the fees awarded to class counsel are too high.
In partial response, most of the named plaintiffs and Boeing submitted summaries of allegedly comparable average monetary awards in other employment class actions; declarations of several experts, including the Reverend Jesse Jackson, praising the proposed decree (largely on the understanding that the decree would provide individual class members with free legal representation with regard to their employment issues at Boeing);9 and evidence that Boeing had vigorously contested race discrimination cases brought to trial against the Company, with victorious results that led class counsel, as stated in a declaration to the district court, to be “hard pressed to find anything that would support a nationwide victory over Boeing.”
In September 1999, the district court certified a settlement class and approved the decree. In its order approving the decree, the district court concluded that “there are important advantages to class-wide resolution in this type of dispute.” The court cited Boeing’s past success in defending against individual claims of race discrimination; the court’s assessment of the effectiveness of the injunctive relief; and the cooperative nature of the settlement. It found no merit to the objectors’ qualms over the class’s certification. Plaintiffs’ “allegations clearly raise class-wide legal and factual issues sufficient to satisfy the [commonality] requirement.” Moreover, typicality was assured by the “broadly selected cross-section ... of Boeing employees” serving as named plaintiffs. The court proceeded to certify the class.
Concerning the fairness of the decree, the district court emphasized “a strong *952judicial policy favoring settlement of class actions,” noting the conservation of resources for all concerned that leaves “more to devote to the problems raised by the claim.” The court approved the notice procedure followed by the parties and then conducted an analysis of the fairness, adequacy and reasonableness of the decree.
The “heart of the matter” according to the district court was the amount of the settlement; it reviewed the decree’s components and found that the objectors “have not presented any evidence to suggest that the amount of payments appear [sic ] inadequate or unfair when compared with the other cases [cited by Boeing].” Further, the court decided that “the awards to the named parties are not excessive.” Without reviewing the proposed decree in any detail in its order, the district court concluded that the injunctive provisions are not “toothless,” but “present a novel and potentially effective response to the problem of race discrimination.” After rejecting categorically allegations of collusion between class counsel and Boeing, the court concluded, citing this court’s precedent, that “the mere possibility of a better settlement is not sufficient grounds for finding the agreement unfair.” As a final matter, the district court found the award of attorneys’ fees “to be reasonable given the nature of the case, the risks to the plaintiffs’ counsel’s firm, and the amount of pre-filing and post-settlement work performed.”
II. DISCUSSION
This case presents difficult questions regarding the appropriate role of the courts in approving class action settlement decrees. The governing principles are clear, but their application is painstakingly fact-specific and hampered by the much greater knowledge of the parties as to the give- and-take of the bargaining process. Judicial review also takes place in the shadow of the reality that rejection of a settlement creates not only delay but also a state of uncertainty on all sides, with whatever gains were potentially achieved for the putative class put at risk. We are mindful of the value dialogue and cooperation have played in attempting to resolve this litigation and in aspiring to foster a spirit of future goodwill in the wake of an alleged systemic pattern of race discrimination.
To vindicate the settlement of such serious claims, however, judges have the responsibility of ensuring fairness to all members of the class presented for certification. Especially in the context of a case in which the parties reach a settlement agreement prior to class certification, courts must peruse the proposed compromise to ratify both the propriety of the certification and the fairness of the settlement. First, the district court must assess whether a class exists; “[s]uch attention is of vital importance, for a court asked to certify a settlement class will lack the opportunity, present when a case is litigated, to adjust the class, informed by the proceedings as they unfold.” Amchem Prods. Inc. v. Windsor, 521 U.S. 591, 620, 117 S.Ct. 2231, 138 L.Ed.2d 689 (1997). Second, the district court must carefully consider “whether a proposed settlement is fundamentally fair, adequate, and reasonable,” recognizing that “[i]t is the settlement taken as a whole, rather than the individual component parts, that must be examined for overall fairness_” Hanlon v. Chrysler Corp. 150 F.3d 1011, 1026 (9th Cir.1998) (citations omitted). When, as here, the parties have entered into a settlement agreement before the district court certifies the class, reviewing courts “must pay ‘undiluted, even heightened, attention’ to class certification requirements.... ” Id. at 1019 (quoting Amchem, 521 U.S. at 620, 117 *953S.Ct. 2231). Moreover, concerns about the fairness of settlement agreements “warrant special attention when the record suggests that settlement is driven by fees; that is, when counsel receive a disproportionate distribution of the settlement. ...” Id. at 1021.
In this case, the objectors contend that the lawsuit does not qualify for class action status under Rule 23(a). We review under the abuse of discretion standard a district court’s decision to certify a case as a class action. Armstrong v. Davis, 275 F.3d 849, 867 (9th Cir.2001). Although we have some concerns, largely relating to litigation management, as to whether the case could be maintained as a class action if the litigation continues, the district court did not abuse its discretion in certifying the case for settlement purposes pursuant to Rule 23.
The objectors argue in the alternative that the district court should not have approved the settlement agreement under Rule 23(e). ‘We have repeatedly stated that the decision to approve or reject a settlement is committed to the sound discretion of the trial judge because he is exposed to the litigants, and their strategies, positions and proof.” Hanlon, 150 F.3d at 1026 (citation and internal quotation marks omitted). Nonetheless, the district court did in this case abuse that discretion. To repeat what this court had reason recently to state: “Although we are always cautious to reverse the ... approval of a settlement agreement because of the time and effort dedicated by the parties and the district court, we are compelled to do so in this case because of the unjust terms of the decree.” Molski v. Gleich, 318 F.3d 937, 956 (9th Cir.2003).
We address first the propriety of the class certification and then examine the district court’s decision to approve the settlement agreement.
A. Class Certification
Rule 23(a) establishes four prerequisites for class action litigation, which are: (1) numerosity, (2) commonality, (3) typicality, and (4) adequacy of representation. We examine each of these requirements in turn.
1. Numerosity
Rule 23(a)(1) requires that “the class is so numerous that joinder of all members is impracticable.” There is no dispute that the numerosity requirement is met in this case. The plaintiff class before us is approximately 15,000 in number.
2. Commonality
Rule 23(a)(2) requires that “there are questions of law or fact common to the class.” We stated in Hanlon that
Rule 23(a)(2) has been construed permissively. All questions of fact and law need not be common to satisfy the rule. The existence of shared legal issues with divergent factual predicates is sufficient, as is a common core of salient facts coupled with disparate legal remedies within the class.
150 F.3d at 1019.
The class in this case is broad and diverse. It encompasses some 15,000 employees, from a wide range of positions both salaried and hourly, who are employed at Boeing facilities located in 27 different states. Class counsel argue, and the district court found, that the large class is united by a complex of company-wide discriminatory practices against African-Americans.
Gen. Tel. Co. of Southwest v. Falcon, 457 U.S. 147, 157, 159, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982), emphasizes that the fact “that racial discrimination is by definition class discrimination,” cannot automati*954cally convert a single allegation of race discrimination into “an across-the-board attack.” Though the commonality claim in this case is ambitious and therefore especially worthy of scrutiny, it is far from the theoretical extrapolation sought in Falcon. That ease concerned only one named plaintiff who did not identify any other plaintiffs subjected to the treatment he claimed to have experienced.
By contrast, the record indicates that class counsel interviewed more than 1,300 Boeing employees from facilities across the country. Counsel have produced detailed documentation of discrimination experienced by more than 200 of these employees. Those named employees include salaried managers and hourly line-workers, union members and non-union members, employees from all of Boeing’s major locations, and employees from two firms that Boeing recently acquired. Appellees also point to the results of an internal “survey of Boeing’s affirmative action issues,” distributed in an e-mail to Boeing senior management. The e-mail identified “racial bias” in the categories of hiring practices and promotion practices and “race issues” in the category of peer working environment as issues of concern at the Company.
In fact, the named plaintiffs and objectors share the contention that discriminatory practices at Boeing are widespread and entrenched. According to the district court, “both the supporters of the consent decree and the objectors spoke forcefully of institutional problems with race discrimination.” The court may not go so far, of course, as to judge the validity of these claims. “Although some inquiry into the substance of a case may be necessary to ascertain satisfaction of the commonality and typicality requirements of Rule 23(a), it is improper to advance a decision on the merits to the class certification stage.” Moore v. Hughes Helicopters, Inc., 708 F.2d 475, 480 (9th Cir.1983), citing Eisen v. Carlisle & Jacquelin, 417 U.S. 156, 177-78, 94 S.Ct. 2140, 40 L.Ed.2d 732 (1974). But the breadth and consistency of class counsel’s initial evidence places the district court’s finding of commonality well within that court’s discretion.
Objectors also dispute the finding of commonality on three more specific grounds. First, objectors claim that common issues cannot link all African-American Boeing employees when two of Boeing’s subsidiaries are recent acquisitions. Boeing acquired the defense and space operations of Rockwell International in December 1996 (renaming them Boeing North America, hereinafter “BNA”), and the McDonnell Douglas Corporation (“MDC”) in August 1997. Employees from both these subsidiaries are among the class representatives. Their reported experiences are similar to the others’.
Boeing is responsible for the employment practices of all its sub-parts. The Company, for example, issued an “Equal Employment Opportunity” policy in July 1998, applicable to all Boeing organizations. This document stated that the “primary responsibility for implementing this policy rests with the senior management of the company.” Some of the class’s historical evidence from BNA and MDC employees may extend to before the mergers, but the kernel of the class complaint is that a complex of discriminatory practices pervades Boeing today and in the recent past. The fact of the mergers does not bar a finding of commonality.
Objectors also contest commonality on the ground that some class members were subject to collective bargaining agreements that laid out objective criteria for promotions. Objectors point to a comment in a footnote from the Falcon opinion:
*955Significant proof that an employer operated under a general policy of discrimination conceivably could justify a class of both applicants and employees if the discrimination manifested itself in hiring and promotion practices in the same general fashion, such as through entirely subjective decisionmaking processes. In this regard it is noteworthy that Title VII prohibits discriminatory employment practices, not an abstract policy of discrimination.
Falcon, 457 U.S. at 159 n. 15, 102 S.Ct. 2364. This hypothetical class in Falcon’s footnote fifteen is both broader and narrower than the class before us. It is broader because the class here does not combine employees with applicants and does not allege discriminatory hiring. It is narrower because the class here complains of a complex of discriminatory practices that includes compensation, training, and work environment in addition to promotions. Union employees under objective promotion systems may have been immune to discrimination in promotion, but they could still have been affected by other alleged axes of discrimination.
We understand footnote fifteen of Falcon to present a demonstrative example rather than a limited exception to the overall skepticism toward broad discrimination class actions.10 That is, as we read Falcon, it does not generally ban all broad classes but rather precludes a class action that, on the basis of one form of discrimination against one or a handful of plaintiffs, seeks to adjudicate all forms of discrimination against all members of a group protected by Title VII, § 1981, or a similar statute.
Bowman v. Block, 940 F.2d 1211 (9th Cir.1991), confirms our reading of Falcon. Bowman was a post -Falcon case in which we upheld the certification of a class of female applicants for a sergeant’s position. Defendant Los Angeles County objected to the lack of an evidentiary hearing on the commonality question, citing Falcon. We noted that the district court had concluded “that there were common questions in that ‘plaintiff is attacking defendants’ discriminatory practices against females, and this is not just as it applied to plaintiff only.’ This statement identifies a common legal issue, discrimination against women, and a common factual problem, discrimination as applied in the Sheriffs Department.” Id. at 1232. We found the Falcon concern inapplicable because class status was not sought on the basis of a single discriminatory practice “as it applied to plaintiff only.” Id. For similar reasons, Falcon does not bar a commonality finding in this case.
Third, objectors contend that decision-making at Boeing is too decentralized to permit a class that combines plaintiffs from disparate locales. Objectors rely for this argument on Doninger v. Pac. Northwest Bell, 564 F.2d 1304 (9th Cir.1977), in which we upheld a district court’s denial of certification to an attempted class of female employees and applicants.
The primary reason class certification was inappropriate in Doninger was that the putative class would in large part have *956overlapped with the terms of a consent decree entered into by American Telephone and Telegraph and already applicable to Pacific Northwest Bell employees. “Substantial numbers” of the individuals who would have been class members had waived their claims and accepted relief under the preexisting decree. Id. at 1309.
Additionally, plaintiffs in Doninger wanted to rely, in the style that Falcon later rejected, on the experiences of a few individuals. Ruling on the commonality question, we found it significant that Pacific Northwest Bell was divided into six “establishments,” each with its own affirmative action program, while “[a]ll of the named plaintiffs are employed in one of three establishments.... ” Id. at 1310. We reasoned that “[s]ince different affirmative action programs, and thus possibly different patterns and practices, exist in each establishment, appellants would have considerable difficulty in adequately representing class members from the other three PNB establishments.” Id. at 1311 (footnote omitted).
The case before us does not present the problem of a preexisting consent decree. Moreover, its named plaintiffs come from a wide array of Boeing’s divisions.
As noted above, Boeing is responsible for the employment practices of all its units. Class counsel introduced evidence of centralized decisionmaking. The unsurprising fact that some employment decisions are made locally does not allow a company to evade responsibility for its policies. See, e.g., Bates v. United Parcel Serv., 204 F.R.D. 440, 446 (N.D.Cal.2001); Morgan v. United Parcel Serv. of Am., 169 F.R.D. 349, 356 (E.D.Mo.1996).
We conclude that the district court was within its discretion to find the commonality requirement of Rule 23(a)(2) met in this case. In so holding, we stress that we are applying an abuse of discretion standard. The district court in all likelihood could, also without abusing its discretion, have declined to certify the overall class in favor of certifying discrete sub-classes, so as to assure commonality. As the district court noted, however, the objectors in this case, while pointing to many aspects of the consent decree with which they disagree, did not demonstrate that the certification of a broad class rather than subclasses compromised the interests of one or more of the groups of employees that might have had sufficiently cohesive interests to have been certified as a subclass. We later conclude that the consent decree did unfairly distribute the available funds among the members of the plaintiff class. But that unfair distribution did not reflect the same fault lines of potential conflict that the objectors maintain undermined the class action determination, namely the potential conflict among employees who work at different levels in the corporate hierarchy, among employees who work in different locations, and among employees who work for Heritage Boeing as opposed to those who work for the recently-merged units.
Although for the most part the same standards apply under Rule 23 in judging the propriety of a settlement class as apply in determining whether a class for trial purposes is appropriate, Amchem, 521 U.S. at 620-21, 117 S.Ct. 2231, in judging the propriety of a settlement class, “close inspection of the settlement in that regard [is] altogether proper.” Id. at 620, 117 S.Ct. 2231. The presence or absence of settlement terms that differentially affect the sub-groups which objectors contend have potentially diverging interests is indicative of whether these alleged conflicts are in fact pertinent to the issues in the lawsuit. Cf. id. at 626-27, 117 S.Ct. 2231 (noting that in the asbestos products liability case at issue “the terms of the settlement reflect essential allocation decisions” *957regarding the distribution of funds, terms that did not take into account “[t]he disparity between the currently injured and exposure-only categories of plaintiffs” regarding the relative desirability of immediate payments versus “an ample, inflation-protected fund for the future”); see also Molski, 318 F.3d at 955-56 (looking at the consent decree to determine likelihood of collusiveness). Consequently, the district court acted within its discretion in declining to insist on subclasses and thereby potentially undo the settlement for reasons unlikely to have affected it.
3. Typicality
Rule 23(a)(3) requires that “the claims or defenses of the representative parties are typical of the claims or defenses of the class.” Falcon noted that “[t]he commonality and typicality requirements of Rule 23(a) tend to merge.” 457 U.S. at 157 n. 13, 102 S.Ct. 2364. In this case, the district court explained that:
The named plaintiffs ... include a very broadly selected cross-section of the different categories of Boeing employees. Salaried and hourly, management and line-worker, union and non-union are all represented, as are each of the major geographic hubs of Boeing’s operations and each of the pre-merger companies. Particularly in a case in which the requested relief applies evenly to the various sub-groups, this cross-section of Boeing employees suffices to insure that the interests of these sub-groups have been adequately represented, and meets the typicality requirement of Rule 23(a).
Objectors do not dispute the breadth of representation, but complain that class counsel did not provide clear documentation that each job category had a class representative for each type of discrimination claim alleged.
That level of specificity is not necessary for class representatives to satisfy the typicality requirement. In Hanlon, we stated that “[u]nder the rule’s permissive standards, representative claims are ‘typical’ if they are reasonably coextensive with those of absent class members; they need not be substantially identical.” 150 F.3d at 1020. Typicality “does not mean that the claims of the class representative[s] must be identical or substantially identical to those of the absent class members.” 5 Herbert B. Newberg & Alba Conte, Newberg on Class Actions, § 24.25 at 24-105 (3d ed.1992); see also Armstrong, 275 F.3d at 869. The district court here was within its discretion to find that the representatives’ claims are “reasonably coextensive with those of absent class members.” Hanlon, 150 F.3d at 1020.
4. Adequacy of Representation
Rule 23(a)(4) permits the certification of a class action only if “the representative parties will fairly and adequately protect the interests of the class.” To determine whether the representation meets this standard, we ask two questions: (1) Do the representative plaintiffs and their counsel have any conflicts of interest with other class members, and (2) will the representative plaintiffs and their counsel prosecute the action vigorously on behalf of the class? Hanlon, 150 F.3d at 1020; see also Molski, 318 F.3d at 955 (quoting Crawford v. Honig, 37 F.3d 485, 487 (9th Cir.1995), and stating a similar standard).
Counsel conducted broad research, assertedly interviewed some 1,300 employees, held many meetings with class members at various Boeing sites, and achieved some relief from a company that has historically been successful in defending against discrimination claims. See, e.g., Croker v. Boeing Co., 662 F.2d 975 (3d Cir.1981) (court found against class on all *958issues of liability; nominal damages granted to individual plaintiffs; decision overruled on other, procedural grounds). Although we later question whether the settlement agreement, as opposed to class counsel’s pre-settlement activity, was the result-of disinterested representation, that question is better dealt with as part of the substantive review of the settlement than under the Rule -23(a) inquiry. Otherwise, the preliminary class certification issue can subsume the substantive review of the class action settlement.11 The district court also permitted discovery on the allegations of outright collusion between class counsel and Boeing and concluded that there was no proof that collusion had occurred.12 The district court neither abused its discretion in finding that counsel’s representation was appropriately vigorous for purposes of class certification nor clearly erred in finding that there was no overt collusion.
With regard to the first of the two adequacy questions, objectors contend that,a conflict arises from the facts that the class cuts across the levels of authority of Boeing employees and that, in particular, some class members supervise some of their fellow class members. This concern about classes that involve both supervisors and rank-and-file workers can be a valid one in some circumstances. In Wagner v. Taylor, 836 F.2d 578 (D.C.Cir.1987), for example, the court affirmed a district court’s finding of inadequate representation because a representative plaintiff in a Title VII case purported to represent a broad class that included non-supervisory employees. Wagner was both a senior executive and the only representative of an attempted class of all grade GS-9 and above African-American employees of, and applicants to, the Interstate Commerce Commission. The court worried that “[sjuper-visory employees are often inappropriate representatives of nonsupervisory employees because the structure of the workplace tends to cultivate distinctly different interests between the two groups. Although each group shares the interest in freedom from discrimination, potential conflicts may and do arise within a class including both.” Id. at 595 (footnotes omitted).
Wagner did not,-however, adopt any per se rule concerning adequacy of representation where .the class includes employees at different levels of an employment hierarchy. We decline to do so as well. The question whether employees at different levels of the internal hierarchy have potentially conflicting interests is context-specific and depends upon the particular claims alleged in a case.
Here, we do not find these workforce structure concerns to be dispositive. The named plaintiffs who are class representatives in this case include both supervisors and non-supervisory employees. The district court found that objectors fail
*959to identify a substantive issue for which there is a conflict of interest between two or more sets of employees. Given that the named plaintiffs include representatives of each major employee subgroup, and that the requested relief applies equally throughout the class, the Court finds that there are no conflicts between class members sufficient to defeat certification.
“Plaintiffs attempting representation of nonsupervisory employees by supervisory employees ... must offer evidence of coextensive interests or at least allege the existence of a general discriminatory policy.” Newberg and Conte, supra, § 24.42 at 24-170-71. Class counsel have met this burden here. The finding of adequacy was within the district court’s discretion.
B. The Settlement
1. General Principles
“Fed.R.Civ.P. 23(e) requires the district court to determine whether a proposed settlement is fundamentally fair, adequate, and reasonable.” Hanlon, 150 F.3d at 1026 (citation omitted). The objectors contend that the settlement agreement fails to meet Rule 23(e)’s standards.
To determine whether a settlement agreement meets these standards, a district court must consider a number of factors, including: “the strength of plaintiffs’ case; the risk, expense, complexity, and likely duration of further litigation; the risk of maintaining class action status throughout the trial; the amount offered in settlement; the extent of discovery completed, and the stage of the proceedings; the experience and views of counsel; the presence of a governmental participant; and the reaction of the class members to the proposed settlement.” Molski, 318 F.3d at 953 (citation omitted); see also Officers for Justice v. Civil Serv. Comm’n of San Francisco, 688 F.2d 615, 625 (9th Cir.1982) (noting that the list of factors is “by no means an exhaustive list of relevant considerations, nor have we attempted to identify the most significant factors”).
Despite this guidance, assessing the fairness, adequacy and reasonableness of the substantive terms of a settlement agreement can be challenging. As “the very essence of a settlement is compromise, ‘a yielding of absolutes and an abandoning of highest hopes,’ ” id. at 624 (citation omitted), review of the substantive terms of a consent decree — the total amount of damages awarded for example, or the precise terms of injunctive provisions — is often not productive. Courts cannot know the strength of ex ante legal claims and so are not privy to the relative strengths of the parties at the bargaining table. Nor can courts judge with confidence the value of the terms of a settlement agreement, especially one in which, as here, the settlement provides for injunctive relief.
At the same time, and critically for present purposes, there are real dangers in the negotiation of class action settlements of compromising the interests of class members for reasons other than a realistic assessment of usual settlement considerations such as the strength of their legal claims, the desire for immediate rather than delayed relief, and the costs of litigation. Incentives inhere in class-action settlement negotiations that can, unless checked through careful district court review of the resulting settlement, result in a decree in which “the rights of [class members, including the named plaintiffs] may not [be] given due regard by the negotiating parties.” Id. The class members are not at the table; class counsel and counsel for the defendants are. Unlike in the non-class action context, most of class counsel’s clients cannot be consulted individually about the terms of the settlement, nor is the resulting decree submitted to the class *960members for approval (although there is an opportunity to object).
That the class representatives are available for consultation and approval is no solution, for two reasons: First, the class representatives have their own incentives to advance their interests at the expense of the class. Second, class counsel ultimately owe their fiduciary responsibility to the class as a whole and are therefore not bound by the views of the named plaintiffs regarding any settlement. See In re GMC Pick-Up Truck Fuel Tank Prods. Liab. Litig., 55 F.3d 768, 801 (3d Cir.1995) (“Beyond their ethical obligations to their clients, class attorneys, purporting to represent a class, also owe the entire class a fiduciary duty once the class complaint is filed.”).
We have characterized these inherent dangers of class settlements as encompassing the possibility that “the agreement ... is the product of fraud or overreaching by, or collusion between, the negotiating parties .... ” Officers for Justice, 688 F.2d at 625; see also Hanlon, 150 F.3d at 1027. By so stating, we do not mean to indicate concern only with overt misconduct by the negotiators. The incentives for the negotiators to pursue their own self-interest and that of certain class members are implicit in the circumstances and can influence the result of the negotiations without any explicit expression or secret cabals. That is why district court review of class action settlements includes not only consideration of whether there was actual fraud, overreaching or collusion but, as well, substantive consideration of whether the terms of the decree are “fair, reasonable and adequate to all concerned.” Officers for Justice, 688 F.2d at 625.
Still, both the difficulties of judicial assessment of a compromise settlement, discussed above, and the rule that “[t]he district court’s final determination to approve the settlement should be reversed ‘only upon a strong showing that the district court’s decision was a clear abuse of discretion,’ ” Hanlon, 150 F.3d at 1027 (quoting In re Pac. Enters. Sec. Litig., 47 F.3d 373, 377 (9th Cir.1995)), circumscribe our inquiry on appeal into the fairness question. As a practical matter we will rarely overturn an approval of a class action consent decree on appellate review for substantive reasons unless the terms of the agreement contain convincing indications that the incentives favoring pursuit of self-interest rather than the class’s interests in fact influenced the outcome of the negotiations and that the district court was wrong in concluding otherwise. See Molski 318 F.3d at 953-54.
Our inquiry therefore most usefully focuses primarily upon whether the particular aspects of the decree that directly lend themselves to pursuit of self-interest by class counsel and certain members of the class — namely, attorneys’ fees and the distribution of any relief, particularly monetary relief, among class members — strictly comport with substantive and procedural standards designed to protect the interests of class members. This is not to say that we do not consider the remaining terms of the settlement agreement as well. “It is the settlement taken as a whole, rather than the individual component parts, that must be examined for overall fairness,” and “[t]he settlement must stand or fall in its entirety.” Hanlon, 150 F.3d at 1026; see also Strong v. BellSouth Telecomms., 137 F.3d 844, 848 (5th Cir.1998) (“To be enforceable, the Agreement require[s] the final approval of each federal court, pursuant to [Rule] 23(e). Any modification to the Agreement, whether by a party or a court, would render the Agreement void.”).
Where the other terms of a settlement raise questions not recognized by the dis*961trict court concerning fairness and adequacy — as we conclude they do here — our scrutiny of the fees and damages distribution provisions should be all the more rigorous. But absent some glaring inequity in the remaining terms of the agreement missed in the district court’s inquiry, it will be rare that we will reverse a district court’s approval of a class action consent decree unless the fees and relief provisions clearly suggest the possibility that class interests gave way to self-interest.
2. The Settlement as a Whole
In this case, we are somewhat uneasy, reading the settlement as a whole, about whether in reaching the settlement, class counsel adequately pursued the interests of the class as a whole. Provisions giving rise to this unease include the extent of Boeing’s release from liability, which includes any breach of contract action by any class member; the stipulation that the prohibition on race discrimination cannot be enforced in individual cases; the numerous instances in which Boeing is permitted to develop its own remedial schemes (and, in some instances, unilaterally to abandon such schemes as infeasible), with an obligation only to consult with class counsel but with no obligation to submit to any enforcement or dispute resolution mechanism if the schemes are unsatisfactory; the limited role for the consultant Boeing is required to hire; and the incorporation in the agreement of promotion and complaint programs Boeing had already developed and implemented, with no obligation on the part of the Company to continue those programs in their present form or alternatively to substitute programs of the same efficacy.13
Further, the district court did not entirely appreciate the limited scope of many of the injunctive provisions of the decree. For example, the court opined that “changes will be made in the procedure for resolving discrimination related complaints.” (Emphasis added.) In fact, the changes in the complaint procedure had already been implemented, and, while the agreement indicates that the implementation appeared adequate, Boeing made no commitment in the decree to continue the same process in effect during the term of the agreement (or even to assure that any replacement process would be as effective as the present one).
We also note that, unlike the district court, we decline to rely in our assessment of the injunctive provisions upon “the approval of several disinterested experts in race discrimination, the Reverend Jesse Jackson first among them.” The experts’ positive assessments all rely heavily on the assertion that the decree provides all members of the class with three years of free legal assistance to, as one declaration put it, “review their employment history, review proposed or actual job opportunities, ... assist them with job applications, and ... challenge the selection of someone else for the jobs.” As noted above, Boeing has expressed its skepticism that the decree embodies any obligation on the part of class counsel to provide such free indi*962vidualized legal assistance or on the part of Boeing to respond to such individualized representation by attorneys. Reading the decree carefully, we share that skepticism.
The attorneys’ fees provision provides $750,000 to class counsel for “monitoring, administration, implementation and defense of the Decree,” including, in particular, “Class Counsel’s time and expenses involved in the processing of claims under Section XI(c)(4) and the distribution of all monetary awards ... including expenditures by Class Counsel in regard to compensating the Claims Arbitrator.... ” That language hardly encompasses the individualized representation for future claims of discrimination the experts’ declarations assume. Nor does any provision in the decree specifically require Boeing to confer with class counsel about individuals’ promotion applications. It may be that class counsel, commendably, intend to attempt to provide such individual representation, although nothing in the factual record indicates a commitment to do so. If so, $750,000 is unlikely to go very far in compensating class counsel for such representation, given the size of the class and the other representational duties for which the decree specifically earmarks the money. Since the' experts’ understanding of the proposed settlement appears less than precise, the district court should not have relied so heavily upon those assessments, and we do not do so.
Despite all of the foregoing concerns, we would not overturn the district court’s determination to approve the settlement as fair were the release and injunctive provisions the only aspects of the decree that are troublesome. As the district court noted, plaintiffs’ risk of losing the case on the merits was quite high; Boeing had an unbroken history of prevailing in diserimi-nation cases; maintaining the class action was not a foregone conclusion; promotion decisions, a primary focus of the litigation, are largely discretionary; and discovery was likely to be extremely expensive for the plaintiffs and class counsel, whose resources are undoubtedly more limited than those of Boeing. The total monetary relief provided in the proposed settlement agreement is not insubstantial, either in total or on a pro rata basis given the number of claimants, and the balance between retrospective and prospective relief is usually one for the litigants to determine. Nor do the injunctive provisions themselves raise any flags regarding favoritism for some members of the class over others. No class member is assured a promotion or any other future privilege not accorded to others, nor are certain groups of class members treated more favorably than others for purposes of future relief.
Additionally, we cannot say with the requisite certainty that the district court erred in concluding that “the cooperative process which led to the agreement, and the provisions for oversight of the decree by independent observers ... will lead to genuine improvements in Boeing’s internal race relations.” The district court was in a position to assess the level of good-will between the parties at the point of settlement, as we are not.14
Furthermore, we are told that the decree in large part incorporates already-existing Boeing programs rather than creating new ones because Boeing determined once the lawsuit was filed to devise new, more effective promotion and complaint policies so as to avoid similar charges in the future. Although the failure of the consent decree to assure continuation of these same or equally effective programs *963in the future remains troubling, we cannot reject out of hand this explanation for crediting existent programs as part of the prospective relief attained by the proposed decree.
In short, the injunctive aspects of the proposed settlement neither directly reflect pursuit of self-interest by favored members of the class nor, standing alone, strike us as being so beyond the pale as a compromise of claims to merit reversal of the district court’s fairness assessment. At the same time, the questionable factors we have noted do suggest the possibility that class counsel and the IIRs could have agreed to relatively weak prospective relief because of other inducements offered to them in the course of the negotiations. We therefore scrutinize with particular care the aspects of the proposed settlement that provide monetary benefits directly to class counsel and to the IIRs: the attorneys’ fees and damages distribution provisions.15
3. Attorneys’Fees
a. Necessity of Scrutiny: Attorneys’ fees provisions included in proposed class action settlement agreements are, like every other aspect of such agreements, subject to the determination whether the settlement is “fundamentally fair, adequate, and reasonable.” Fed.R.Civ.P. 23(e). There is no exception in Rule 23(e) for fees provisions contained in proposed class action settlement agreements. Thus, to avoid abdicating its responsibility to review the agreement for the protection of the class, a district court must carefully assess the reasonableness of a fee amount spelled out in a class action settlement agreement. See, e.g., Piambino v. Bailey, 610 F.2d 1306, 1328 (5th Cir.1980) (“the District Court abdicated its responsibility to assess the reasonableness of attorneys’ fees proposed under a settlement of a class action, and its approval of the settlement must be reversed on this ground alone”); Strong, 137 F.3d at 848-50; In re GMC, 55 F.3d at 819-20; Jones v. Amalgamated Warbasse Houses, Inc., 721 F.2d 881, 884 (2d Cir.1983) (holding with regard to attorneys’ fees that “[t]he presence of an arms’ length negotiated agreement among the parties weighs strongly in favor of approval, but such an agreement is not binding on the court.”).16
*964That the defendant in form agrees to pay the fees independently of any monetary award or injunctive relief provided to the class in the agreement does not detract from the need carefully to scrutinize the fee- award. Ordinarily, “a defendant is interested only in disposing of the total claim asserted against it ... the allocation between the class payment and the attorneys’ fees is of little or no interest to the defense....” In re GMC, 55 F.3d at 819-20 (internal quotation marks and citation omitted); see also Evans v. Jeff D., 475 U.S. 717, 732, 734, 106 S.Ct. 1531, -89-L.Ed.2d 747 (1986) (recognizing that “the possibility of a tradeoff between merits relief and attorney’s fees” is often implicit in class action settlement negotiations, because “[m]ost defendants are unlikely to settle unless the cost of the predicted judgment, discounted by its probability, plus the transaction costs of further litigation, are greater than the cost of the settlement package.") (Emphasis added).
Given these economic realities, the assumption in scrutinizing a class action settlement agreement must be, and has always been, that the members of the class retain an interest in assuring that the fees to be paid class counsel are not unreasonably high. If fees are unreasonably high, the likelihood is that the defendant obtained an economically beneficial concession with regard to the merits provisions, in the form of lower monetary payments to class members or less injunctive relief for the class than could otherwise have obtained. See Court Awarded Attorney Fees, Report of the Third Circuit Task Force, 108 F.R.D. 237, 266 (1985) (“When a large attorney’s fee means a smaller recovery to plaintiff, a significant conflict of interest between client and attorney is created. Even if the plaintiffs attorney does not consciously or explicitly bargain for a higher fee at the expense of the beneficiaries, it is very likely that this situation has indirect or subliminal effects on the negotiations.”). In other words, the negotiation of class counsel’s attorneys’ fees is not exempt from the truism that there is no such thing as a free lunch.
We have, in closely analogous contexts, recently so recognized. See Zucker v. Occidental Petroleum Corp., 192 F.3d 1323 (9th Cir.1999); Lobatz v. U.S. West Cellular, 222 F.3d 1142 (9th Cir.2000). In Zucker, for example, we stated that “[i]n a class action, whether the attorneys’ fees come from a common fund or are otherwise paid, the district court must exercise its inherent authority to assure that the amount and mode of payment of attorneys’ fees are fair and proper,” 192 F.3d at 1328 (emphasis added); see also id. at 1327 (“In a class action ... [t]he 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).”). Similarly, in Lobatz, where the defendant had agreed that it *965would not contest a fee request of $1 million that was apart from the settlement fund, we noted that “[s]uch an agreement has the potential of enabling a defendant to pay class counsel excessive fees and costs in exchange for counsel accepting an unfair settlement on behalf of the class.” 222 F.3d at 1148.
The district court was therefore obligated to assure itself that the fees awarded in the agreement were not unreasonably high, so as to ensure that the class members’ interests were not compromised in favor of those of class counsel.
b. Substantive Scrutiny of Statutory Fees: Generally, litigants in the United States pay their own attorneys’ fees, regardless of the outcome of the proceedings. In order to encourage private enforcement of the law, however, Congress has legislated that in certain cases prevailing parties may recover their attorneys’ fees from the opposing side. When a statute provides for such fees, it is termed a “fee-shifting” statute. Under a fee-shifting statute, the court “must calculate awards for attorneys’ fees using the ‘lodestar’ method,” Ferland v. Conrad Credit Corp., 244 F.3d 1145, 1149 n. 4 (9th Cir. 2001), which involves “multiplying the number of hours the prevailing party reasonably expended on the litigation by a reasonably hourly rate,” Morales v. City of San Rafael, 96 F.3d 359, 363 (9th Cir.1996) and, “if circumstances warrant, adjusting] the lodestar to account for other factors which are not subsumed within it,” Ferland, 244 F.3d at 1149 n. 4; see also Caudle v. Bristow Optical Co., 224 F.3d 1014, 1029 (9th Cir.2000). The rules governing both reduction and enhancement have become increasingly refined over time, and we have therefore required careful explanations by district courts of statutory fee determinations.17
Both Title VII, § 2000e, et seq., and § 1981 — the two federal statutes under which this suit was brought — have fee-shifting provisions. See § 2000e-5 (k) (“In any action or proceeding under this sub-chapter the court, in its discretion, may allow the prevailing party ... a reasonable *966attorney’s fee .... ”); 42 U.S.C. § 1988 (“In any action or proceeding to enforce a provision .of section[ ] 1981 ... the court, in its discretion, may allow the- prevailing party ... a reasonable attorney’s fee.... ”). The parties therefore could have negotiated an award of fees under § 2000e-5(k) and § 1988. Had they- done so, the district court’s review would have focused on the reasonableness of the fee request under the lodestar calculation method. Were the amount of fees Boeing agreed to pay in the settlement agreement distinctly higher than the fees class counsel could have been awarded by the district court using the lodestar method, the court would almost surely have had to find the fees unreasonable. Absent some unusual explanation, a defendant would not agree in a class action settlement to pay out of its own pocket fees measurably higher than it could conceivably have to pay were the fee amount litigated, unless there was some non-fee benefit the defendant received thereby.
In fact, no lodestar-based scrutiny of the fees awarded class counsel in the settlement agreement ever took place. Boeing and class counsel did not attempt to explain the award of fees provided in the consent decree as negotiated under the applicable fee-shifting statutes. Further, the record as it stands would not have been sufficient for such an inquiry, as it contains only the barest estimate of hours expended, with no detail. Not even a summary of the billing records was submitted.
Of course, in the context of a settlement, the fees provided for in the agreement are as subject to compromise as are the merits provisions. Consequently, as Evans, 475 U.S. at 734-35, 106 S.Ct. 1531, made clear, the fee amount in a class action settlement agreement can be less than would be awarded by a court. And, since the proper amount of fees is often open to dispute and the parties are compromising precisely to avoid litigation, the court need not inquire into the reasonableness of the fees even at the high end with precisely the same level of scrutiny as when the fee amount is litigated. But here, there was no such inquiry at all.. Nor is the record adequate for an inquiry, even one employing a less-than-stringent standard that recognizes the settlement context.
We are therefore in no position to determine whether the fees Boeing agreed to pay are reasonable lodestar fees under the applicable fee-shifting statutes and do not do so. On remand, the parties are free to attempt such justification, based on the principles outlined in this opinion and in the extensive lodestar fees case law.
c. The Common Fund Justification: Rather than justifying the attorneys’ fees provisions of the settlement agreement on the statutory fee-shifting basis that would properly have applied, the parties sought to justify the fee amount according to the principles applicable to common funds. They did so by constructing a hypothetical “fund” by adding together the amount of money Boeing would pay in damages to members of the class under the agreement, the amount of fees provided to various counsel, the cost of the class action notices paid for by Boeing, and a gross amount of money ascribed to all the in-junctive relief contained in the agreement. For clarity, we will call the total of all those monetary amounts the “putative fund,” for, as we shall see, it is not properly viewed as a common fund as that term is used in attorneys’ fees law. (We will continue, also for clarity, to call the doctrine by its usual name, “common fund.”) The parties portrayed the total fee award as 28% of the putative fund, and maintained that such a percentage is well within the percentage permitted under our common fund fee cases. The district court *967viewed the fee award as the parties requested and approved it, and the consent decree as a whole, on that basis. For several reasons, that approval was not appropriate.
i. Availability of common fund fees
Before we can decide whether the attempted common fund justification in this case was adequate, we must resolve whether the existence of potentially applicable fee-shifting statutory provisions precludes class counsel from recovering attorneys’ fees under the common fund doctrine. We conclude, as have the two other circuits that have addressed the issue,18 that there is no preclusion on recovery of common fund fees where a fee-shifting statute applies.
Under the “common fund” doctrine, “a litigant or a lawyer who recovers a common fund for the benefit of persons other than himself or his client is entitled to a reasonable attorney’s fee from the fund as a whole.” Boeing Co. v. Van Gemert, 444 U.S. 472, 478, 100 S.Ct. 745, 62 L.Ed.2d 676 (1980). The Supreme Court explained:
The doctrine rests on the perception that persons who obtain the benefit of a lawsuit without contributing to its cost are unjustly enriched at the successful litigant’s expense. Jurisdiction over the fund involved in the litigation allows a court to prevent this inequity by assessing attorney’s fees against the entire fund, thus spreading fees proportionately among those benefited by the suit.
Id. (citation omitted). Thus, the common fund doctrine ensures that each member of the winning party contributes proportionately to the payment of attorneys’ fees. In contrast to fee-shifting statutes, which enable a prevailing party to recover attorneys’ fees from the vanquished party, the common fund doctrine permits the court to award attorneys’ fees from monetary payments that the prevailing party recovered in the lawsuit. Put another way, in common fund cases, a variant of the usual rule applies and the winning party pays his or her own attorneys’ fees; in fee-shifting cases, the usual rule is rejected and the losing party covers the bill. See generally Wininger v. SI Mgmt. L.P., 301 F.3d 1115 (9th Cir.2002).
The procedures used to determine the amount of reasonable attorneys’ fees differ concomitantly in cases involving a common fund from those in which attorneys’ fees are sought under a fee-shifting statute. As in a statutory fee-shifting case, a district court in a common fund case can apply the lodestar method to determine the amount of attorneys’ fees. In common fund cases, however, the court can apply a risk multiplier when using the lodestar approach. See In re Wash. Pub. *968Power Supply Sys. Sec. Litig., 19 F.3d 1291, 1299 (9th Cir.1994) (‘WPPSS”) (“[The City of Burlington v. Dague, 505 U.S. 557, 112 S.Ct. 2638, 120 L.Ed.2d 449 (1992) ] rationale for barring risk multipliers in statutory fee,cases does not operate to bar risk multipliers in common , fund cases.”). A “multiplier” is a number, such as 1.5 or 2, by which the base lodestar figure is multiplied in order to increase (or decrease) the award of attorneys’ fees on the basis of such factors as the risk involved and the length of the proceedings.
Alternatively, in a common fund case, the district court -can determine the amount of attorneys’ fees to be drawn from the fund by employing a “percentage” method. See- Hanlon, 150 F.3d at 1029 (“In ‘common fund’ cases where the settlement or award creates a large fund for distribution to the class, the district court has discretion to use either a percentage or lodestar method.”). As its name suggests, under the percentage method, “the court simply awards the attorneys a percentage of the fund....” Id. “This circuit has established 25% of the common fund as a benchmark award for attorney fees.” Id.
That common fund fees can be awarded where statutory fees are available follows from the equitable nature of common fund fees. In Alyeska Pipeline Serv. Co. v. Wilderness Soc’y, 421 U.S. 240, 95 S.Ct. 1612, 44 L.Ed.2d 141 (1975), the Court explained that 28 U.S.C. § 1923, a version of which was first enacted in 1853, limits the amount of attorneys’ fees that a prevailing party may recover from the loser, but does not prohibit the award of fees under the common fund doctrine. The Court stated:
In Trustees v. Greenough, 105 U.S. 527, 26 L.Ed. 1157 (1882), the 1853 Act was read as not interfering with the historic power of equity to permit the trustee of a fund or property, or a party preserving or recovering a fund for the benefit of others in addition to himself, to recover his costs, including his attorneys’ fees, from the fund or property itself or directly from the other parties enjoying the benefit. That rule has been consistently followed.... These exceptions are unquestionably assertions of inherent power in the courts to allow attorneys’ fees in particular situations, unless forbidden by Congress....
Alyeska Pipeline, 421 U.S. at 257-59, 95 S.Ct. 1612. Thus, unless Congress has forbidden the application of the common fund doctrine in cases in which attorneys could potentially recover fees under the type of fee-shifting statutes at issue here, the courts retain their equitable power to award common fund attorneys’ fees.
Congress did not explicitly forbid the use of the common fund doctrine in cases potentially involving § 2000e-5 and § 1988, and we see no reason to infer that it did so implicitly. The intent of the fee-shifting provisions at issue here is not countered by the application of common fund principles.
The fees available under a fee-shifting statute are part of the plaintiffs recovery and are not dependent upon any explicit fee arrangements between the plaintiffs and their counsel. For that reason, contingent fee agreements between counsel and client are valid in cases where statutory fees are available. See Venegas v. Mitchell, 495 U.S. 82, 86-89, 110 S.Ct. 1679, 109 L.Ed.2d 74 (1990). Common fund fees are essentially an equitable substitute for private fee agreements where a class benefits from an attorney’s work, so the same general principles outlined in Venegas should apply.
Application of the common fund doctrine to class action settlements does not com*969promise the purposes underlying fee-shifting statutes. In settlement negotiations, the defendant’s determination of the amount it will pay into a common fund will necessarily be informed by the magnitude of its potential liability for fees under the fee-shifting statute, as those fees will have to be paid after successful litigation and could be treated at that point as part of a common fund against which the attorneys’ fees are measured. Conversely, the prevailing party will expect that part of any aggregate fund will go toward attorneys’ fees and so can insist as a condition of settlement that the defendants contribute a higher amount to the settlement than if the defendants were to pay the fees separately under a fee-shifting statute.19
The district court did not, therefore, err in treating this case as one that could fall under the common fund doctrine rather than under the potentially applicable fee-shifting provisions, if the parties properly so agreed, the resulting fee was reasonable, and other requisites applicable to common fund fees were met.
The possibility that a prevailing party could recover fees either under the court’s equitable powers or under its statutory authority does not, however, give the parties or the court free rein once either the common fund or the statutory rubric is selected. Fees sought or awarded under a fee-shifting statute require the application of the standards and procedures crafted fer such statutes, discussed above. Similarly, if the parties invoke common fund principles, they must follow common fund procedures and standards, designed to protect class members when common fund fees are awarded. We turn next to the specific procedure employed in the negotiation and award of the attorneys’ fees in this case.
ii. Inclusion in the settlement of the attorneys’ fees
The parties negotiated the amount of attorneys’ fees awarded class counsel as a term of the settlement agreement and thus conditioned the merits settlement upon judicial approval of the agreed-upon fees. See Hanlon, 150 F.3d at 1026 (“Neither the district court nor this court ha[s] the ability to ‘delete, modify or substitute certain provisions.’ The settlement must stand or fall in its entirety.”) (citations omitted). By proceeding in this fashion with respect to attorneys’ fees and then attempting to justify the fees not as statutory fees but as common fund fees, the parties followed an irregular and, as we hold below, improper procedure.
Under regular common fund procedure, the parties settle for the total amount of the common fund and shift the fund to the court’s supervision. The plaintiffs’ lawyers then apply to the court for a fee award from the fund. See Paul, Johnson, Alston *970& Hunt v. Graulty, 886 F.2d 268, 271 (9th Cir.1989) (in a common fund case, “a court has control over the fund — even one created pursuant to a settlement, as here ... and assesses the litigation expenses against the entire fund so that the burden is spread proportionally among those who have benefited.”) (citing Van Gemert, 444 U.S. at 478, 100 S.Ct. 745); see also Vizcaino v. Microsoft Corp., 290 F.3d 1043, 1046 (9th Cir.2002) (after approval of the settlement, class counsel applied to the district court for an award of attorneys’ fees); Cook, 142 F.3d at 1011 (“In common fund cases, after attorneys obtain a settlement for the class, they petition the court for compensation from the fund.... ”); Florin, 34 F.3d at 563 (in a common fund case, “the defendant typically pays a specific sum into the court, in exchange for a release of its liability. The court then determines the amount of attorney’s fees that plaintiffs’ counsel may recover from this fund, thereby diminishing the amount of money that ultimately will be distributed to the plaintiff class.”).20
In setting the amount of common fund fees, the district court has a special duty to protect the interests of the class. On this issue, the class’s lawyers occupy a position adversarial to the interests of their clients. The reason for the usual insistence upon judge-conferred common fund fees is that, as we have explained,
“Because in common fund cases the relationship between plaintiffs and their attorneys turns adversarial at the fee-setting stage, courts have stressed that when awarding attorneys’ fees from a common fund, the district court must assume the role of fiduciary for the class plaintiffs.” WPPSS, 19 F.3d at 1302. Accordingly, fee applications must be closely scrutinized. Rubber-stamp approval, even in the absence of objections, is improper.
Vizcaino, 290 F.3d at 1052. See also In re Coordinated Prerial Proceedings in Petroleum Prods. Antitrust Litig., 109 F.3d 602, 608 (9th Cir.1997) (“In a common fund case, the judge must look out for the interests of the beneficiaries, to make sure that they obtain sufficient financial benefit after the lawyers are paid. Their interests are not represented in the fee award proceedings by the lawyers seeking fees from the common fund”).
When the ordinary procedure is not followed and instead the parties explicitly condition the merits settlement on a fee award justified on a common fund basis, the obvious risk arises that plaintiffs’ lawyers will be induced to forego a fair settlement for their clients in order to gain a higher award of attorneys’ fees. That risk is, if anything, exacerbated where, as here, the agreement provides for payment of fees by the defendant, as in a statutory fee-shifting situation, but the parties choose to justify the fee as coming from a putative common fund. Where that is the case, courts have to be alert to the possibility that the parties have adopted this hybrid course precisely because the fee award is in fact higher than could be supported on a statutory fee-shifting basis, yet the deal is so dependent upon class counsel receiving a greater-than-lodestar amount of fees that the parties were not willing to give the court supervisory discretion to *971determine the distribution of the total settlement package between counsel and the class.
We recognize that in Evans, 475 U.S. at 720,106 S.Ct. 1531, the Court held that the parties to a class action may simultaneously negotiate merits relief and an award of attorneys’ fees under a fee-shifting statute, and may condition the entire settlement upon a waiver of fees. The Court explained:
[A] general proscription against negotiated waiver of attorney’s fees in exchange for a settlement on the merits would itself impede vindication of civil rights, at least in some cases, by reducing the attractiveness of settlement.
Id. at 732, 106 S.Ct. 1531; see also id. at 733, 106 S.Ct. 1531 (“If defendants are not allowed to make lump-sum offers that would, if accepted, represent their total liability, they would understandably be reluctant to make settlement offers.”) (quoting Marek v. Chesny, 473 U.S. 1, 6-7, 105 S.Ct. 3012, 87 L.Ed.2d 1 (1985)). Thus, to facilitate settlement by providing defendants with assurances as to the limits of their liability exposure, the parties to a lawsuit may, in conjunction with their merits negotiation, properly negotiate statutory fees to be paid by the defendant.
The concern motivating the decision in Evans — that prohibiting simultaneous negotiations and agreements as to merits and fees will discourage settlements — simply does not exist, however, in a case, such as this one, in which the parties to the negotiations seek to justify attorneys’ fees as coming from a putative fund and to apply common fund principles. Usually, an agreement that provides lawyers fees on a common fund basis constitutes a “lump-sum” agreement, Evans, 475 U.S. at 733, 106 S.Ct. 1531, one that enables defendants to know the precise extent of their liability regardless of the amount of attorneys’ fees eventually awarded from the fund. Thus, the parties could have simply agreed upon the total amount of the putative fund, as well as the damages and injunctive relief, and left the division of that fund as between the class and counsel to the district court, as is usual in common fund cases.21 Requiring the parties to so proceed or, in the alternative, to agree to a fee award as part of the settlement agreement in an amount no higher than could be justified by statutory fee-shifting principles, fully serves the defendant’s only legitimate interest in class counsel’s fee award. See In Re GMC, 55 F.3d at 819-20 (“[A] defendant is interested only in disposing of the total claim asserted against it ... the allocation between the class payment and the attorneys’ fees is of little or no interest to the defense.” (internal quotation marks and citations omitted)); Florin, 34 F.3d at 562 n. 1 (“The parties agreed that attorney’s fees were to come out of the settlement fund. Defendants have satisfied their obligation to pay into the settlement fund, and thus have no interest in the amount of fees class counsel want to extract from the fund.”). That requirement thereby provides the requisite impetus to settlement on the defendant’s part while protecting against a maldistri-bution of the total settlement package between the class and its counsel.
Further, the effect of conditioning the settlement on a set amount of attorneys’ fees based on an actual or putative common fund can be to inhibit district courts from engaging in independent determinations of reasonable fees, as required by law. The parties’ all-or-nothing approach imposes pressure to approve otherwise ac*972ceptable and desirable settlements in spite of built-in attorneys’ fees provisions. While this same dynamic may exist where fees can be justified on a statutory fee basis, the more precise lodestar standards for adjudging the reasonableness of such fees, summarized above, make the influence of such pressure much less forceful.
We hold, therefore, that in a class action involving both a statutory fee-shifting provision and an actual or putative common fund, the parties may negotiate and settle the amount of statutory fees along with the merits of the case, as permitted by Evans. In the course of judicial review, the amount of such attorneys’ fees can be approved if they meet the reasonableness standard when measured against statutory fee principles. Alternatively, the parties may negotiate and agree to the value of a common fund (which will ordinarily include an amount representing an estimated hypothetical award of statutory fees) and provide that, subsequently, class counsel will apply to the court for an award from the fund, using common fund fee principles. In those circumstances, the agreement as a whole does not stand or fall on the amount of fees. Instead, after the court determines the reasonable amount of attorneys’ fees, all the remaining value of the fund belongs to the class rather than reverting to the defendant.
The parties in this case did not follow either of these procedures, or any other that adequately protected the class from the possibility that class counsel were accepting an excessive fee at the expense of the class.22 The district court therefore erred in approving the consent decree.
iii. Injunctive relief as part of the district court’s putative fund
Even if the fee award had been determined in a procedurally proper way, approval of the amount of the attorneys’ fees on common fund principles would still have been mistaken as a matter of law, because the actual percentage award was much higher than the 28% the district court recognized.
In order for attorneys to obtain an award of fees from a common fund, the court must be able to: (1) sufficiently identify the class of beneficiaries; (2) accurately trace the benefits; and (3) shift the fee to those benefiting with some exactitude. Van Gemert, 444 U.S. at 478-79, 100 S.Ct. 745. “[T]he criteria are satisfied when each member of a certified class has an undisputed and mathematically ascertainable claim to part of a lump-sum judgment recovered on his behalf,” whereas they are not satisfied when “litigants simply vindicate a general social grievance.” Id. at 479, 100 S.Ct. 745.
Under these requirements, the monetary relief for the plaintiff class (including attorneys’ fees) provided for in the consent decree could be converted as described above so as to qualify as a common fund from which class counsel could obtain an award of attorneys’ fees. The class consists of the approximately 15,000 African-American Boeing employees and so is sufficiently identifiable. The benefits from the monetary relief provided for in the decree would be distributed according to *973its terms and so can be accurately traced. Finally, the fees could be shifted with exactitude to the benefiting class if taken properly from the fund.
In approving the award of attorneys’ fees provided for in the consent decree, the district court employed the percentage method to determine that the award was fair. The court found that the fees constituted 28% of the putative fund, just above the 25% benchmark. To make this calculation, however, the court included in the amount of the putative fund an estimated value of $8.65 million for injunctive relief, the amount that the decree required Boeing to spend on approval and implementation of this component.
Although the injunctive relief falls somewhere between the permissible and the prohibited bases for fees set forth in Van Gemert — the relief neither produces “an undisputed and mathematically ascertainable” amount for each class member, nor merely “vindicate [s]” a general social grievance” — -we have no difficulty here deciding that the district court abused its discretion in counting the parties’ estimated value of that relief towards the putative fund.
The injunctive relief included in the consent decree requires Boeing only to “meet and confer” with class counsel or to discuss certain issues. Although Boeing must participate in such conferences and discussions, there is no requirement that Boeing take any action with respect to what the Company learns. The conferences and discussions may not result in tangible relief to class members. Moreover, a diversity consultant may not benefit the class to a degree commensurate with his or her cost.
Additionally, while the injunctive relief (along with the cost of obtaining approval of the decree) is to cost Boeing a fixed minimum amount, $3.65 million, some of the injunctive relief described in the consent decree consists of steps Boeing had apparently decided to take on its own, even before it entered the settlement. The decree also permits Boeing to credit expenditures towards'the injunctive relief amount without regard to whether such expenditures are in addition to the cost of Boeing’s prior outlays for administering similar programs. Thus, the true cost of the injunction to the defendant — and the true benefit to the plaintiff class — is a matter of speculation and may be far less than $3.65 million. That amount of money cannot be accurately traced to the decree, let alone to the beneficiaries making up the class. Without the estimated value of the injunctive relief, the fund is reduced to only $10.55 million, and the fee award of $4.05 million constitutes 38% — well above the 25% benchmark — of the putative fund.
We do not hold that a district court can never consider the value of injunctive relief in determining the reasonableness of a common fund fee. For instance, in Han-lon, 150 F.3d at 1029, we upheld the use of the common fund doctrine to award attorneys’ fees after the parties reached a settlement agreement under which Chrysler would replace defective latches on minivans that it had manufactured. Although the replacement of latches is injunctive in nature, the agreement bestowed upon each beneficiary a clearly measurable benefit: one replacement latch for each minivan owned. The court could therefore, with some degree of accuracy, value the benefits conferred. Even so, in Hanlon the district court used its valuation of the fund only as a cross-check of the lodestar amount, “rejecting] the idea of a straight percentage recovery because of its uncertainty as to the valuation of the settlement,” id, and it was on that basis that we affirmed the fee award.
*974Precisely because the value of injunctive relief is difficult to quantify, its value is also easily manipulable by overreaching lawyers seeking to increase the value assigned to a common fund. We hold, therefore, that only in the unusual instance where the value to individual class members of benefits deriving from injunctive relief can be accurately ascertained may courts include such relief as part of the value of a common fund for purposes of applying the percentage method of determining fees.23 See Van Gemert, 444 U.S. at 478-79, 100 S.Ct. 745. When this is not the case, courts should consider the value of the injunctive relief obtained as a “relevant circumstance” in determining what percentage of the common fund class counsel should receive as attorneys’ fees, rather than as part of the fund itself. See Vizcaino, 290 F.3d at 1049. Alternatively, particularly where obtaining injunctive relief likely accounted for a significant part of the fees expended, courts can use the common fund version of the lodestar method either to set the fee award or as a cross-check to assist in the determination of how the “relevant circumstance” of the injunctive relief should affect a percentage award. See id. at 1050 (“Calculation of the lodestar, which measures the lawyers’ investment of time in the litigation, provides a check on the reasonableness of the percentage award.”).24
The district court did not employ either of these procedures here. Nor can we determine on the record before us that considering the injunctive relief as a “relevant circumstance” or employing the common fund lodestar method would have justified the award of $4.05 million as a reasonable fee. On this ground, also, the district court erred in approving the proposed attorneys’ fees award.
iv. Treatment of Costs of Litigation
In assessing the reasonableness of the fee award, the parties and the district court (1) included the pre-settlement costs of litigation — fairly low in this case, as the settlement occurred early in the lawsuit— as part of the fees awarded; but (2) included the cost incurred by Boeing for providing notices to the class of the settlement in the value of the injunctive relief, and therefore as part of the putative fund against which the reasonableness of the fees was measured.
The parties to the proposed settlement agreed to the inclusion of costs in the amount attributed to fees and the objectors, understandably, have not protested that inclusion. As all of those affected are content with that method of calculation and no class member’s interests are adversely affected, the district court had no cause to disapprove the attribution, nor do we.
The district court also did not abuse its discretion by including the cost of provid*975ing notice to the class of the proposed consent decree as part of its putative fund valuation, although the cost of providing two notices rather than one should not have been included. We have said that “the choice of whether to base an attorneys’ fee award on either net or gross recovery should not make a difference so long as the end result is reasonable. Our case law teaches that the reasonableness of attorneys’ fees is not measured by the choice of the denominator.” Powers v. Eichen, 229 F.3d 1249, 1258 (9th Cir.2000). The post-settlement cost of providing notice to the class can reasonably be considered a benefit to the class. Also, where, as here, it is the defendant who pays for the notice, we may assume that the inherent incentives to minimize the cost involved are sufficient. Additionally, the court’s supervision of the form of notice and the method of communication assures that the costs expended are contained. We con-' elude that where the defendant pays the justifiable cost of notice to the class — -but not, as here, an excessive cost — it is reasonable (although certainly not required) to include that cost in a putative common fund benefiting the plaintiffs for all purposes, including the calculation of attorneys’ fees.
4. Awards to Named and Unnamed Class Members
One other aspect of the settlement agreement also raises serious concerns as to its fairness, adequacy and reasonableness. The 237 people who are HRs (and have not opted out) would each receive, on average, sixteen times greater damages than each of the unnamed class members; the subgroup of IIRs who are class representatives would receive even more relative to the class as a whole. While a point-based formula is used to distribute the damages fund among the rest of the class members who make claims, no such objective standards were applied in determining who would be paid as an HR or in what amount. We find no sufficient justification in the record for this differential in the amount of damage awards and the process for awarding them.
The district court “considered this disparity carefully because excessive payments to named class members can be an indication that the agreement was reached through fraud or collusion.” Indeed, “[i]f class representatives expect routinely to receive special awards in addition to their share, of the recovery, they may be tempted to accept, suboptimal settlements at the expense of the class members whose interests they are appointed to guard.” Weseley v. Spear, Leeds & Kellogg, 711 F.Supp. 713, 720 (E.D.N.Y.1989); see also Women’s Comm. for Equal Employment Opportunity v. Nat’l Broad. Co., 76 F.R.D. 173, 180 (S.D.N.Y.1977) (“[W]hen representative plaintiffs make what amounts to a separate peace with defendants, grave problems of collusion are raised.”).
Although the district court expressed concern* about, the differential in damage awards, the court ultimately concluded, that the settlement was reasonable, explaining that the named plaintiffs are “the group of class members identified by plaintiffs’ counsel as having the strongest claims and thus as providing the strongest foundation for the lawsuit. Because they have the strongest, claims it is fair that they would receive the largest awards.” The record does not, however, support this explanation.
Class counsel represent that an individual became an IIR due to his or her willingness to step forward, risk retaliation, contribute to the attorneys’ costs, and assist *976in coordination of the lawsuit.25 The contention is that the individuals with the strongest claims were the most likely to participate in this manner. But that generalization greatly oversimplifies. Some further, considerably more direct evidence regarding the strength of the HRs’ claims is necessary to justify the large disparities in damages.
The need for direct evidence supporting the representation that the IIRs had particularly strong claims is heightened because the record suggests an alternative explanation for the selection of particular individuals as IIRs — their presettlement retention of class counsel and promise to contribute to counsel’s costs. Without contrary proof, the possibility that class counsel were simply rewarding with higher damages amounts those class members who had promised to contribute toward their costs during the pendency of the suit gains considerable plausibility.
Such special rewards for counsel’s individual clients are not permissible when the case is pursued as a class action. Generally, when a person “joints] in bringing [an] action as a class action ... he has disclaimed any right to a preferred position in the settlement.” Officers for Justice, 688 F.2d at 632. Were that not the case, there would be considerable danger of individuals bringing cases as class actions principally to increase their own leverage to attain a remunerative settlement for themselves and then trading on that leverage in the course of negotiations.
In In re Cont’l Ill. Sec. Litig., 962 F.2d 566, 571 (7th Cir.1992) (hereinafter “Continental Illinois ”), the Seventh Circuit approved of “incentive fees” to compensate named plaintiffs for the risks they take and their vanguard role in the class action. See id. (“Since without a named plaintiff there can be no class action, such compensation as may be necessary to induce him to participate in the suit could be thought the equivalent of the lawyers’ nonlegal but essential case-specific expenses, such as long-distance phone calls, which are reimbursable.”). Continental Illinois would not justify the damages distribution in this case, as the much higher awards in the consent decree went to a large group of class members, not only to the class representatives. The two hundred-odd IIRs who were not class representatives were not essential to the litigation, although they may have been helpful to it.
We have, however, approved incentive awards of $5,000 each to the two class representatives of 5,400 potential class members in a settlement of $1,725 million. See In re Mego Fin. Corp. Sec. Litig., 213 F.3d 454, 463 (9th Cir.2000); see also In re U.S. Bancorp Litig., 291 F.3d 1035, 1038 (8th Cir.2002) (approving $2,000 incentive awards to five named plaintiffs out of a class potentially numbering more then 4 million in a settlement of $3 million); Cook, 142 F.3d at 1016 (approving, in the context of a recovery of more than $14 million, an incentive payment of $25,000 to one named plaintiff who “spent hundreds of hours with his attorneys and provided them with ‘an abundance of information’ ”); Continental Illinois, 962 F.2d at *977571-72 (upholding a district court’s rejection of a proposed $10,000 award to a named plaintiff “for his admittedly modest services”); In re SmithKline Beckman Corp. Sec. Litig., 751 F.Supp. 525, 585 (E.D.Pa.1990) (approving $5,000 awards for one named representative of each- of nine plaintiff classes involving more than 22,000 claimants in a settlement of $22 million). In the proposed consent decree in this case, 29 named class representatives are designated to receive payments totaling $890,000. Compared to the three cases just mentioned, the different orders of magnitude in the present case concerning the number of named plaintiffs receiving incentive payments, the proportion of the payments relative to the settlement amount, and the size of each payment— here up to $50,000, with an average of more than $80,000 — are obvious. Nevertheless, named plaintiffs, as opposed to designated class members who are not named plaintiffs, are eligible for reasonable incentive payments. The district court must evaluate their awards individually, using “relevant factors including] the actions the plaintiff has taken to protect the interests of the class, the degree to which the class has benefitted from those actions, ... the amount of time and effort the plaintiff expended in pursuing the litigation ... and reasonable] fear[s of] workplace retaliation.” Cook, 142 F.3d at 1016.
Additionally, class members can certainly be repaid from any cost allotment for their substantiated litigation expenses, and identifiable services rendered to the class directly under the supervision of class counsel can be reimbursed as well from the fees awarded to the attorneys.26 See Missouri v. Jenkins, 491 U.S. 274, 285, 109 S.Ct. 2463, 105 L.Ed.2d 229 (1989). Similarly, if class members other than the named plaintiffs demonstrate that they were in fact retaliated against — or at least make some credible allegation of past or possible future retaliation — based on their role in the lawsuit, higher damages awards for such individuals than for other members of the class would be justified. But any plaintiff assumes the risk of retaliation, and we hesitate to single out non-named plaintiff IIRs as entitled to special payments simply for undertaking such risk.
Further, singling out a large group of non-named plaintiff class members for higher payments without regard to the strength of their claims eliminates a critical check on the fairness of the settlement for the class as a whole. Such individual class members who have actively participated in the litigation are the ones likely to be most aware of the dynamic at the negotiating table, the strength of the class claims, and the costs of pursuing the litigation. If they support the settlement agreement and are treated equally in that agreement with other class members making similarly strong claims, the likelihood that the settlement is forwarding the class’s interests to the maximum degree practically possible increases. If, on the other hand, such members of the class are provided with special “incentives” in the settlement agreement, they may be more concerned with maximizing those incentives than with judging the adequacy of the settlement as it applies to class members at large.
All these concerns about incentive or risk payments to certain class members are exacerbated in this case by the allegation, and in one case (that of Cordell Bolder) the apparent reality, that HR awards went to individuals who were not proper *978members of the class.27 Again, if those individuals rendered compensable services to the lawyers, then the lawyers should pay for those services from the amount of the fund properly awarded for costs or fees, as appropriate. But if one or more of these individuals is clearly not a member of the class and therefore not entitled to any damages award, any proposed decree should be approved only if the provision awarding that person or those persons damages is deleted. If, alternatively, an individual’s class membership is debatable, then the award to him or her can be considered an element of the compromise.
We conclude that the aspects of the settlement agreement pertaining to the distribution of the damages fund cannot stand on the present record. Because the very large differential in the amount of damage awards between the named and unnamed class members is not justified on this record, the district court abused its discretion in finding the settlement agreement to be fair, adequate and reasonable under Rule 23(e). We therefore reverse the judgment on this ground as well.
III. CONCLUSION
Having determined that reversal is appropriate on several grounds, we need not consider the other arguments of the objectors.
On remand, the parties will have a choice concerning whether to attempt to justify the present proposed agreement under the principles outlined above or, instead, to renegotiate the aspects of the agreement yve have indicated are questionable. If they choose the former course and are able to justify the damages distribution (which on the present record appears quite unlikely), they will then also have to substantiate the fee award using a lodestar calculation under the applicable fee-shifting statutes rather than on a common fund basis.28 For the reasons stated, the fee as it stands cannot be justified on a common fund basis, and the court can only approve or disapprove the present agreement in its entirety. Thus, if the fee cannot be justified on the fee-shifting statutory basis, the entire agreement will have to be renegotiated.
If, alternatively (and more likely, given the seeming difficulty of justifying the current proposed decree), the parties decide to renegotiate and are able to present a revised settlement agreement to the district court, the court will of course need to reassess carefully the fairness of the settlement in accord with the standards we have explicated. If the parties negotiate a settlement fund and leave the attorneys’ fees to be determined by the court on a common fund basis, then the district court, upon petition by class counsel, will determine these fees under the common fund principles discussed above.
The decision of the district court approving the proposed consent decree is REVERSED and REMANDED for proceedings consistent with this opinion.
. After applying the reversion and opt-out provisions, the damages awarded by the decree amount to approximately $6.5 million.
. This amount includes $3.85 million in fees and costs to class counsel and $200,000 to objectors’ counsel.
.It is possible that new hires are not so barred. The decree provides, in a separate provision, that nothing in the decree "bars any claims of members of the Settlement Class ... based on or arising out of events occurring after the Preliminary Approval date.”
. After opt-outs this group was reduced to 237 persons.
. The full $100,000 has already been paid to class counsel, by order of the district court.
. The decree appears to provide — but is not lucid in this respect — that Boeing is to designate a group of three nominees from which the consultant will be selected, taking into consideration nominees proposed by plaintiffs.
. If Boeing decides that there are "significant impediments to implementing ... programs” providing information about who received a particular promotion, Boeing is to "meet and confer with Class Counsel regarding alternative approaches.” Again, there is no requirement that Boeing actually implement any such alternative approaches.
. "Heritage Boeing” refers to the portion of the Company that existed before its purchase of the McDonnell Douglas Corporation and parts of Rockwell International.
. Similarly, the district court read the decree . as providing that "free legal advice will be available to assist class members who feel that they have been discriminated against by the company.” In its memorandum in support of class certification, Boeing wrote that it "does not subscribe to as expansive an interpretation of Class Counsel's post-decree monitoring responsibilities.”
. One Fifth Circuit opinion may reflect the latter approach. In Vuyanich v. Republic Nat’l Bank, 723 F.2d 1195 (5th Cir. 1984), that court over-turned the certification of a class of women and African-American employees and applicants (which was divided into five subclasses). The decision relied in part on the existence of objective factors in the hiring process, referring to the comment in footnote fifteen as a " 'general policy of discrimination’ exception.” "The district court’s finding that the Bank relied on two objective inputs — • education and experience — in its necessarily subjective hiring process ... precludes reliance on this ‘general policy of discrimination' exception.” Id. at 1199-1200.
. In Molski, we found inadequate representation under Rule 23(a)(4) primarily because of the different circumstances of the named plaintiff and some members of the class. We did look to the terms and circumstances of the ultimate agreement as confirmation of inadequate representation, but did not base the finding of inadequacy of representation on the substance of the agreement alone. 318 F.3d at 955-56.
. Even when there is no direct proof of explicit collusion, there is always the possibility in class action settlements that the defendant, class counsel, and class representatives will all pursue their own interests at the expense of the class. For that reason, the absence of direct proof of collusion does not reduce the need for careful review of the fairness of the settlement, particularly those aspects of the settlement that could constitute inducements to the participants in the negotiation to forego pursuit of class interests. See p. 960, infra.
. The racial harassment provisions of the proposed consent decree illuminate the defects in many of the other remedial sections: Unlike the promotion and complaint provisions, the section covering the harassment policy does contain some enforceable standards should Boeing choose to change its policy — "the amended policy [must have] the same import as the [present] Harassment Policy and [be] reasonably designed to achieve the same effect as the Harassment Policy in respect to preventing and remedying racial harassment.” Similar language is noticeably absent from all the other sections concerning the programs Boeing is to develop or has developed to remedy discrimination against the class.
. We stress once again that, as this case has not been litigated, we have no way of knowing whether there was in fact race discrimination in employment at Boeing.
. Also contributing to our determination to scrutinize the attorneys' fees provisions with special care is the nature of the notice to the class with respect to fees. The class notice did not break out the amount of attorneys' fees provided for in the settlement agreement, although an alert class member could have calculated those fees from the information provided.
Appellees admit that “with benefit of hindsight it would have been preferable if the formal notice to [the] class had included more specificity with respect to the fee award.” They have not explained, however, why the decision was made to leave out that particular figure from the class notice while spelling out the total fund available to the class, the amount of monetary relief, and the value assigned to the injunctive relief.
Notice of the amount of fees serves as "adequate notice of class counsel’s interest in the settlement.” Torrisi v. Tucson Elec. Power Co., 8 F.3d 1370, 1375 (9th Cir.1993); see also Goldenberg v. Marriott PLP Corp., 33 F.Supp.2d 434, 441 (D.Md.1998) (“Notice of the potential extent of attorneys fee awards is deemed essential because it allows class members to determine the possible influence of the fees on the settlement and to make informed decisions about their right to challenge the fee award.”). Where the class was informed of the amount of fees only indirectly and where the failure to give more explicit notice could itself be the result of counsel's self-interest, the courts must be all the more vigilant in protecting the interests of class members with regard to the fee award.
. Jones held that a district court could, after reviewing the attorneys' fees awarded as part of a class action settlement, reduce the amount of fees, apparently while retaining the *964binding nature of the remainder of the agreement. Such a procedure is not consistent with Evans v. Jeff D., 475 U.S. 717, 726-27, 106 S.Ct. 1531, 89 L.Ed.2d 747 (1986), decided after Jones. Evans held that, with respect to attorneys' fees provisions as with respect to any other provision, "the power to approve or reject a settlement negotiated by the parties before trial does not authorize the court to require the parties to accept a settlement to which they have not agreed.... Rule 23(e) does not give the court the power, in advance of trial, to modify a proposed consent decree and order its acceptance over either party’s objection.” (Footnote omitted). See also Hanlon, 150 F.3d at 1026 (stating, without exception, that "[t]he settlement must stand or fall in its entirety”).
. See, e.g., City of Burlington v. Dague, 505 U.S. 557, 112 S.Ct. 2638, 120 L.Ed.2d 449 (1992) (prohibiting pure contingency enhancements of the lodestar under fee-shifting statutes); Ferland, 244 F.3d at 1149, 1151 (holding that where the district court cuts substantially the number of hours compensated because of perceived inefficiency, the court must either “calibrate the number [of hours] chosen to demonstrable inefficiency in carrying out particular tasks’’ or provide an explanation of the level of reduction chosen); Caudle, 224 F.3d at 1029, 1029 n. 11 (finding an abuse of discretion where the district court did not "explicitly follow [lodestar] procedures” and noting that a modest amount of recovery cannot be used to reduce a fee award below the lodestar); Van Gerwen v. Guar. Mut. Life Co., 214 F.3d 1041, 1049 (9th Cir.2000) (concluding that the district court did not abuse its discretion in refusing to award fees for hours spent on discovery unrelated to the record); Guam Soc’y of Obstetricians & Gynecologists v. Ada, 100 F.3d 691, 697 (9th Cir.1996) {“Dague left undisturbed earlier Supreme Court case law allowing a fee applicant to recover more than the lodestar figure where the applicant has met the burden of showing that such an adjustment is necessary to the determination of a reasonable fee.” (internal quotation marks and citations omitted)); Morales, 96 F.3d at 365 (vacating attorneys' fee award because the exception to lodestar calculation for nominal damages cases in which the plaintiff's success is de minimis did not apply); Gates v. Deukmejian, 987 F.2d 1392, 1398, 1400 (9th Cir. 1992) (rejecting the district court's reduction of the lodestar based on the absence of "concise but clear” explanatory language to show that it did not "uncritically accept[ ] plaintiffs' suggested reductions and fail[ ] independently to review the record”); Cunningham v. County of Los Angeles, 879 F.2d 481, 487, 489 (9th Cir. 1988) (holding four justifications for adjusting the lodestar improper because they are subsumed in the lodestar determination itself).
. See Brytus v. Spang & Co., 203 F.3d 238, 246-247 (3d Cir.2000) (holding that common fund fees can be appropriate in both settled and litigated cases where statutory fees are available); Cook v. Niedert, 142 F.3d 1004 (7th Cir. 1998) (approving fees measured by common fund rather than statutory principles where statutory fees were available); Florin v. Nationsbank, 34 F.3d 560, 564 (7th Cir. 1994) (common fund principles "properly control a case which is initiated under a statute with a fee shifting provision, but is settled with the creation of a common fund.”); Skelton v. General Motors COrp., 860 F.2d 250, 256 (7th Cir. 1988) ("[W]hen a settlement fund is created in exchange for release of the defendant’s liability both for damages and for statutory attorneys’ fees, equitable fund principles must govern the court’s award of the attorneys’ fees.”); 1 Mary Francis Derfner and Arthur D. Wolf, Court Awarded Attorney Fees, ¶ 2.05[7] at 2-81 (2001) (”[T]he mere fact that a fee-shifting statute is implicated in the action does not ensure that fees will be awarded under that statute.... [F]ees may be taxed against the [settlement] fund under the common fund doctrine.” (citing Skelton and Florin)).
. The Seventh Circuit reasoned similarly in Florin:
The settlement agreement approved by the court provides that the defendants are released from potential liability for statutory attorney’s fees and that class counsel may instead petition the court for an award of fees from the settlement fund. Thus, the settlement agreement seems to anticipate that the amount paid by the defendants into the fund includes an unspecified sum for class counsel's fees. An award of attorney's fees from the fund would therefore be consistent with the goal of the fee-shifting provision to allow the offending party [to] bear the costs of the award ... Furthermore, an award of fees from the settlement fund comports with the fee-shifting policy of enabling meritorious plaintiffs who would not otherwise be able to afford to bring a lawsuit ... to pursue their claims.
34 F.3d at 564. See also Brytus, 203 F.3d at 246 (where there is a settlement in a case in which statutory fees are available, "consideration of the attorney’s fees was likely factored into the amount of settlement”).
. On one occasion, we permitted a carefully conceived procedure to substitute for completely independent judicial determination of lodestar-based common fund fees. Hanlon, 150 F.3d at 1029 (fees were negotiated only after the merits agreement was concluded and a mediator present at the negotiations provided assurance to the court "that the fee was not the result of collusion or a sacrifice of the interests of the class” before the district judge reviewed the award using the lodestar, not percentage, method, "requiring class counsel to submit detailed evidence of their work on behalf of the class”).
. The description of the total amount of the fund need not take any particular form and could result from adding up separately-enumerated amounts in the agreement.
. By spelling out these alternatives, we do not mean to preclude all others. Rather, the parties have flexibility in negotiating class action settlement agreements, including the attorneys' fee provisions. The alternatives outlined in the text are paradigms. Any variants, to be reasonable, would have to provide equivalent assurance that the inherent tensions among class representation, defendant’s interests in minimizing the cost of the total settlement package, and class counsel’s interest in fees are being adequately policed by the court.
. Appellees cite cases in which fees have been awarded as a percentage of what they refer to as "non-monetary” benefits. But there is no appellate case cited that supports the fee award here. Two of the cases cited concerned not a common fund but the inapplicable common benefit doctrine. Mills v. Electric Auto-Lite Co., 396 U.S. 375, 392, 90 S.Ct. 616, 24 L.Ed.2d 593 (1970); Loring v. City of Scottsdale, 721 F.2d 274, 275 (9th Cir. 1983). And in Hanlon, supra, and Wing v. Asarco Inc., 114 F.3d 986, 990 (9th Cir. 1997), attorneys’ fees were calculated (as opposed to cross-checked) not by a common fund percentage approach but using the lodestar method.
. As Vizcaino notes, where attorneys’ fees are awarded on a common fund basis, "[t]he bar against risk multipliers ... does not apply,” so the lodestar approach can include a risk multiplier. 290 F.3d at 1051; see also Florin, 34 F.3d at 564; WPPSS, 19 F.3d at 1299-1300.
. For instance, a declaration by one of the class lawyers states that the named plaintiffs were those who were willing to take the risk of appearing as named plaintiffs; being team leaders; assisting the regional groups in gathering documents, evidence, and identifying other witnesses; coordinating the information flow to and from class counsel; assisting in efforts to raise sums to pay for the cost of this litigation; and/or agreeing to come forward with their claims and be witnesses. In our minds, each of these people was in effect a class representative or at least prepared to be a class representative.
. According to class counsel, counsel will return to the IIRs the monies they paid under retainer agreements, so the damages awards do not cover those funds.
. Cordell Bolder is the son of a class representative whose support of the decree was important. Objectors also allege irregularities with respect to another IIR, Mr. Bolder’s stepfather, Jimmy Dean, as well as a class representative, Charles Jones. The district court had a responsibility to address each of these individuals' proposed payments. Instead, its analysis did not go beyond stating that the "award to Mr. Bolder [was] unfortunate," but not fraudulent. Having found that "the parties now admit that he is not a member of the class,” the court should have made approval of the decree contingent on its amendment to eliminate "this single error.”
. In that case, the limitation on risk multipliers announced in Dague, supra, would apply.