OPINION OF THE COURT
BECKER, Circuit Judge.This is an appeal from an order of the District Court for the Eastern District of Pennsylvania approving the settlement of a large class action following its certification of a so-called settlement class. Numerous objectors challenge the fairness and reasonableness of the settlement. The objectors also challenge: (1) the district court’s failure to certify the class formally; (2) its denial of discovery concerning the settlement negotiations; (3) the adequacy of the notice as it pertained to the fee request; and (4) its approval of the attorneys’ fee agreement between the defendants and the attorneys for the class, which the class notice did not fully disclose, thereby (allegedly) depriving the class of the practical opportunity to object to the proposed fee award at the fairness hearing.
The class members are purchasers, over a 15 year period, of mid- and full-sized General Motors pick-up trucks with model C, K, R, or V chassis, which, it was subsequently determined, may have had a design defect in their location of the fuel tank. Objectors claim that the side-saddle- tanks rendered the trucks especially vulnerable to fuel fires in side collisions. Many of the class members are individual owners (i.e., own a single truck), while others are “fleet owners,” who own a number of trucks. Many of the fleet owners are governmental agencies. As will become apparent, the negotiated-settlement treats fleet owners quite differently from individual owners, a fact with serious implications for the fairness of the settlement and the adequacy of representation of the class.
While all the issues we have mentioned are significant (except for the discovery issue), the threshold and most important issue concerns the propriety and prerequisites of settlement classes. The settlement class device is not mentioned in the class action rule, Federal Rule of Civil Procedure 23.1 Rather *778it is a judicially crafted procedure. Usually, the request for a settlement class is presented to the court by both plaintiffs) and defendants); having provisionally settled the case before seeking certification, the parties move for simultaneous class certification and settlement approval. Because this process is removed from the normal, adversarial, litigation mode, the class is certified for settlement purposes only, not for litigation. Sometimes, as here, the parties reach a settlement while the case is in litigation posture, only then moving the court, with the defendants’ stipulation as to the class’s compliance with the Rule 23 requisites, for class certification and settlement approval. In any event, the court disseminates notice of the proposed settlement and fairness hearing at the same time it notifies class members of the pendency of class action determination. Only when the settlement is about to be finally approved does the court formally certify the class, thus binding the interests of its members by the settlement.
The first Manual for Complex Litigation [hereinafter MCL] strongly disapproved of settlement classes. Nevertheless, courts have increasingly used the device in recent years, and subsequent manuals (MCL 2d and MCL 3d (in draft)) have relented, endorsing settlement classes under carefully controlled circumstances, but continuing to warn of the potential for abuse. This increased use of settlement classes has proven extremely valuable for disposing of major and complex national and international class actions in a variety of substantive areas ranging from toxic torts (Agent Orange) and medical devices (Daikon Shield, breast implant), to antitrust cases (the beef or cardboard container industries). But their use has not been problem free, provoking a barrage of criticism that the device is a vehicle for collusive settlements that primarily serve the interests of defendants — by granting expansive protection from law suits — and of plaintiffs’ counsel — by generating large fees gladly paid by defendants as a quid pro quo for finally disposing of many troublesome claims.
After reflection upon these concerns, we conclude that Rule 23 permits courts to achieve the significant benefits created by settlement classes so long as these courts abide by all of the fundaments of the Rule. Settlement classes must satisfy the Rule 23(a) requirements of numerosity, commonality, typicality, and adequacy of representation; as well as the relevant 23(b) requirements, usually (as in this case) the (b)(3) superiority and predominance standards. We also hold that settlement class status (on which settlement approval depends) should not be sustained unless the record establishes, by findings of the district judge, that the same requisites of the Rule are satisfied. Additionally, we hold that a finding that the settlement was fair and reasonable does not serve, as a surrogate for the class findings, and also that there is no lower standard for the certification of settlement classes than there is for litigation classes. But so long as the four requirements of 23(a) and the appropriate requirement(s) of 23(b) are met, a court may legitimately certify the class under the Rule.
In this case the district judge made no Rule 23 findings, and significant questions remain as to whether the class could have met the requisites of the rule had the district court applied them. Principally at issue is adequacy of representation. In particular, the objectors contend that there is a conflict between the positions, of individual owners on the one hand and fleet owners on the other hand. The disparity in- settlement benefits enjoyed:by these different groups, objectors argue, creates an intra-class conflict that precludes the finding of adequacy of representation required by the rule. Moreover, they submit, the large number of different defenses available under the laws of the several states involved also creates a potentially serious commonality and typicality problem.
*779We conclude that the objectors’ adequacy of representation claim probably has merit. At all events, the district court did not properly evaluate the differential impact of the settlement on individual fleet owners, and should determine on remand whether the conflicts among class members are so great as to preclude certification (or at least sufficient to require the creation of subclasses). The district court should also focus on the commonality and typicality problems, to determine whether the national scope of the class litigation and the plethora of defenses available in different jurisdictions prevent these requirements from being met.
For the reasons that follow at some length, we conclude that, although settlement classes are valid generally, this settlement class was not properly certified. We also conclude that the settlement is not fair and adequate; more precisely, we hold that the district court abused its discretion in determining that it was, primarily because the district court erred in accepting plaintiffs’ unreasonably high estimate of the settlement’s worth, in over-estimating the risk of maintaining class status and of establishing liability and damages, and in misinterpreting the reaction of the class. Finally, although our disposition of the foregoing issues makes it. unnecessary for us to pass on the approval of the attorneys’ fees, we clarify the governing standards for these fee awards to guide the district court on remand. We therefore reverse the challenged order of the district court and remand for further proceedings.
I.Facts, PRoceduRal History, and Standard of Review
A. General Background
Between 1973 and 1987, General Motors sold over 6.3 million C/K pickup trucks -with side-mounted fuel tanks.2 In late October 1992, after the public announcement of previously undisclosed information regarding the safety of the fuel tank placement in GM pickups, consumer class action lawsuits were filed in several jurisdictions. The National Highway Traffic Safety Administration (“NHTSA”) commenced an investigation of the alleged defects relating to side-impact fires on these trucks, and consumer advocacy groups sought a recall.3
On November 5, 1992, plaintiffs in one action sought to enjoin allegedly misleading communications to putative class members and filed an application for expedited discovery. On November 8 and 9, 1992, GM filed notices of removal of this and other state court actions, and a motion with the Judicial Panel on Multidistrict Litigation (“MDL Panel”) to transfer and consolidate all actions' for pretrial purposes under 28 U.S.C. § 1407 (1993). The MDL Panel transferred all related actions to the District Court for the Eastern District of Pennsylvania on February 26, 1993. Ultimately, dozens of actions were filed in various courts throughout the United States on behalf of consumer classes; the federal cases were dismissed, remanded to state court, or transferred to the Eastern District of Pennsylvania.
On March 5, 1993, pursuant to an order of the (transferee) District Court, plaintiffs filed a Consolidated Amended Class Action Complaint seeking equitable relief and damages that consolidated all of the actions under the MDL caption and listed nearly 300 representative plaintiffs, including both individual and fleet owners. The Complaint alleged violations of two federal statutes; the Magnuson-Moss Act and the Lanham Trademark Act; a variety of common law and statutory claims, including negligence, fraud, breach of written and implied warranty; and violations of various state consumer statutes. The complaint sought, inter alia, an order remedying the alleged abnormally high incidence of fuel-fed fires following side-impact collisions by requiring GM to recall the trucks or pay for their- repair. GM answered this complaint, denying all substantive allegations and raising numerous affirmative defenses.
Also on March 5, 1993, plaintiffs filed a consolidated motion for nationwide class certification. The court set July 19, 1993, the hearing date on this motion. On March 30, 1993, GM moved to stay this litigation pend*780ing the outcome of the NHTSA investigation, initiated in December 1992. This motion was denied on June 4, 1993. Pursuant to a scheduling order issued by the court, discovery during the spring of 1993 focused on class certification issues. During this discovery, GM produced more than 100,000 pages of documents from prior C/K pickup product liability lawsuits and GM’s responses to NHTSA information requests. Plaintiffs also had access to the depositions and trial testimony in other eases involving the fuel tank design of C/K pickups, including the jury trial in Moseley v. GM, No. 90-V-6276 (Fulton County, Ga.). Plaintiffs consulted with their own experts to evaluate this information,. In addition, depositions were taken of some GM personnel and certain named plaintiffs. Discovery on the merits of the case had been postponed until autumn 1993. Nothing in the record indicates that, as of the spring of 1993, counsel had identified expert witnesses for trial or deposed GM’s engineering experts.
In the midst of these proceedings, the parties began exploring a possible settlement of the litigation. These discussions intensified in June 1993, at which time face-to-face and telephonic meetings, both between the parties and among plaintiffs’ counsel, took place on virtually a daily basis. On July 19, 1993, the parties reached a settlement in principle, reduced the terms to writing, and informed the district court.4 For purposes of settlement only and without prejudice to GM’s substantial opposition to class certification, the named parties agreed to the certification of a settlement class of C/K pickup owners, described below.
B. The Settlement Agreement
In general terms, the settlement agreement provides for members of the settlement class to receive $1,000 coupons redeemable toward the purchase of any new GMC Truck or Chevrolet light duty truck. Settlement certificates are transferable with the vehicle. They are redeemable by the then current owner of the 1973-86 C/K and 1987-91 R/V light duty pickup trucks or chassis cabs at any authorized Chevrolet or GMC Truck dealer for a fifteen month period. Settlement class members do not have to trade in their current vehicle to use the certificate, and the certificates can be used in conjunction with GM and GMAC incentive programs.
The class members can freely transfer the certificate to an immediate family member who resides with the class member. Class members can also transfer the $1000 certificate to a family member who does not reside with the class member by designating the transferee family member within sixty days, running from the date that GM mailed notice of the proposed settlement. Additionally, the $1000 certificate can be transferred with the title to the settlement class vehicle, that is, to a third party who purchases the class member’s vehicle.
In lieu of a $1,000 certificate, and without transferring title to the settlement class vehicle, a class member may instead request that a nontransferable $500 certificate (counterin-tuitively known as the “transfer certificate”) be issued to any third party except a GMC dealer or its affiliates. This $500 certificate is redeemable with the purchase of a new C or K series GMC or Chevrolet full-size pickup truck or its replacement model. The $500 certificate cannot be used in conjunction with any GMC or GMAC marketing incentive, must be used on the more expensive full size models, and is subject to the same fifteen-month redemption period as the $1,000 certificates. The class member must make a notarized request to GM, and GM will mail the $500 certificate to the transferee within 14 days of its receipt of the request for transfer.
Under the terms of the agreement, the approval of the settlement and corresponding entry of final judgment would have no effect *781upon any accrued or future claims for personal injury or death, nor would it affect the rights of settlement class members to participate in any future remedial action that might be required under the National Traffic and Motor Safety Act of 1966, 15 U.S.C. §§ 1381 et seq. (1995).5
The settlement agreement before us also provides that plaintiffs’ counsel would apply to the district court for an award of reasonable attorneys’ fees and reimbursement of expenses, both to be paid by GM. GM reserved the right to object to any fees or expenses it deemed to be excessive and to appeal any amount awarded by the court over its objection. Plaintiffs’ counsel filed their fee applications on or about September 15,1993; the fee applications remained in the files of the clerk of the district court where class members could theoretically review them, but no information about attorneys’ fees other than the fact that a fee application would be made was included in the class notice. GM did not file any formal objections to the fee applications.
C. Approval of the Settlement and Fees
The district court reviewed the substantive terms of the settlement on July 12, 1993 and made the preliminary determination, in Pretrial Order No. 7, entered July 20,1993, that the proposed settlement appeared reasonable. Also in pretrial order no. 7, the court “provisionally” certified the class of GM truck owners as a settlement class (i.e., for settlement purposes only) pursuant to Rule 23(b)(3); however, the court did not make findings that the requisites of Rule 23(a) or 23(b) were satisfied. The court approved the form of and dissemination to putative class members of the combined notice of the pen-dency of the action and the proposed settlement pursuant to Rules 23(c)(2) and 23(e). The class definition included all persons and entities who purchased in the United States (except for residents of the State of Texas) and were owners as of July 19, 1993 of (1) a 1973-1986 model year General Motors full-size pickup truck or chassis cab of the “C” or “K” series; or (2) a 1987-1991 model year General Motors full-size pickup truck or chassis cab of the “R” or “V” series. On August 20 and 21, 1993, GM mailed the notice to all registered owners of class vehicles (including nearly 5.7 million vehicles), and it published the full text of the notice in USA Today and The Philadelphia Inquirer on August 27, 1993.
In response to the notice, over 5,200 truck owners elected to opt out of the class, and approximately 6,500 truck owners (a number which includes" fleet owners who own as many as 1,000 vehicles each) objected to the settlement. The objectors’ filings contained many overlapping claims. The recurring contentions were that: (1) the settlement does nothing to fix the trucks; (2) even with the $1,000 coupon, many owners would be unable to purchase a new truck given their high cost (with list prices from $11,000 to $33,000); (3) state and local government fleet owners would not be able to redeem all of their certificates (by buying new vehicles) within the short redemption period (fifteen months), and they might be further restricted from using the coupons by competitive bidding procurement rules; and (4) GM and class counsel colluded in a manner that compromised the interests of the class and that would preclude a finding of adequate representation. GM rejoined with voluminous material emphasizing the substantial risks plaintiffs faced not only in maintaining class treatment but also in establishing liability and damages.
A settlement fairness hearing was held on October 26, 1993 during which the objectors who submitted written briefs were permitted to speak. The district court approved the settlement in a Memorandum and Order dated December 16, 1993. In that order, the court confirmed its Pretrial Order No. 7, which had provisionally certified the settlement class. Although the court still made no *782findings that the requisites of Rules 23(a) and (b) were met, it did set forth findings of fact and conclusions of law to justify its approval of the settlement as fair, reasonable and adequate based on the nine-factor test established in Girsh v. Jepson, 521 F.2d 153 (3d Cir.1975).
The court found that the total economic value of the settlement was “between $1.98 billion and $2.18 billion.” Against the prospect of settlement, the court weighed each of the nine Girsh factors. It concluded that “the complexity, expense and likely duration of the litigation would be mammoth.” Although the settlement was reached at an early stage of the litigation, just four months after the consolidated complaint was filed, the court found that this did not weigh against the settlement because the court believed that the parties, had access to “extensive discovery on the same issues of product defect that was previously conducted in the various personal injury actions that have been litigated throughout the country.” The district court also found that the reaction of class members to the proposed settlement supported approval citing “the infinitesimal number of truck owners who have either objected to or sought exclusion from the settlement.”
Noting the divided results of the personal-injury jury trials and the numerous defenses GM could raise, the court found that “a substantial risk in establishing liability” weighed in favor of approval. Similarly, the court found that “[pjerhaps the greatest weakness in the plaintiffs’ case is the lack of proof of economic damages.” The court also addressed the objection that the settlement did not provide for a recall or a “fix,” explaining that “no objector that complains that the settlement fails to retrofit the alleged defect has been able to come forth with a practical and safe modification for the trucks that has been designed, evaluated and tested.”
On December 20, 1993, four days after approving the settlement, the district court also approved the class counsel’s request for attorneys’ fees in the amount of $9.5 million. Although the court did not believe at that time that it needed to review that fee award, to which GM had agreed, it subsequently, on February 2,1994, issued an “amplified order” evaluating the award in greater detail. The court determined that the fee request was reasonable under both a lodestar analysis and the percentage-of-recovery method (see Part VII infra).
D.The NHTSA Investigation
While this case was under submission to this court, the NHTSA investigation continued. Over the objections of some of NHTSA’s engineers who had determined that the trucks complied with relevant safety standards, on October 17, 1994, Secretary of Transportation Federico Pena announced the agency’s finding that the trucks contained a safety defect creating an increased and unreasonable risk of side-impact fires. The determination was based on the allegedly enhanced risk of side-impact fires relative to Ford pickups that resulted from GM’s placement of the fuel tanks outside the'frame rails. GM challenged the propriety of the public meeting NHTSA planned to hold and NHTSA’s authority to order a recall of vehicles that met all relevant safety standards. On December 2, 1994, Secretary Pena announced the settlement of the C/K pickup investigation wherein GM contributed over $51 million for a variety of safety programs unrelated to the pickups, and admitted no liability.6
E.Standard of Review
Each of the issues presented here is reviewable for abuse of discretion. See Bryan v. Pittsburgh Plate Glass Co., 494 F.2d 799 (3d Cir.), cert. denied, 419 U.S. 900, 95 S.Ct. 184, 42 L.Ed.2d 146 (1974) (approval of proposed class action settlement); In re School Asbestos Litig., 921 F.2d 1338, 1341 (3d Cir.1990), cert. denied, 499 U.S. 976, 111 S.Ct. 1623, 113 L.Ed.2d 720 (1991) (class certification); Lindy Bros. Builders, Inc. v. American Radiator & Standard Sanitary Corp., 540 F.2d 102, 115 (3d Cir.1976) (award of reasonable attorney’s fees); Marroquin-Manriquez v. INS, 699 F.2d 129, 134 (3d Cir.1983), cert. denied, 467 U.S. 1259, 104 *783S.Ct. 3553, 82 L.Ed.2d 855 (1984) (scope of discovery). An appellate court may find an abuse of discretion where the “district court’s decision rests upon a clearly erroneous'finding of fact, an errant conclusion of law or an improper application of law to fact.” International Union, UAW v. Mack Trucks, Inc., 820 F.2d 91, 95 (3d Cir.1987), cert. denied, 499 U.S. 921, 111 S.Ct. 1313, 113 L.Ed.2d 246 (1991). A finding of fact is clearly erroneous when, although there is evidence to support it, the reviewing court, based on the. entire evidence, concludes with firm conviction that a mistake has been made. Oberti v. Board. of Ed. of Borough of Clementon Sch. Dist., 995 F.2d 1204, 1220 (3d Cir.1993).
II. Anatomy of the Class Claims
The consolidated class complaint filed on behalf of the nationwide class of GM truck owners (except those from Texas) alleged violations of the Magnuson-Moss Warranty Act, 15 U.S.C.A § 2310(d)(1) (1995); and the Lanham Act, 15 U.S.C.A. § 1125(a) (1995); and a variety of state common law and statutory claims, including strict liability in tort for selling a dangerously defective product; negligent design; negligent misrepresentation; fraud (based on defendants’ alleged course of conduct in the advertising, promotion, and sale of the GM pickups intentionally concealing material facts about a dangerous latent defect); breach of warranty, including written (from vehicle warranties), express (from public representations by GM), implied (warranties of merchantability) and statutory warranties; and finally violations of various state consumer protection statutes. The case did not involve any pickup trucks that had actually experienced fuel tank fires caused by side-impact collisions. Moreover, personal injury or death claims were expressly omitted from the complaint as well as from the settlement — class members remain free to pursue such claims if any should accrue.
The aggregated treatment of these claims was potentially complicated by the differences in underlying facts. The trucks at issue had nineteen different fuel tank systems; proof might thus be required for each design on relevant issues. Furthermore, unlike the federal securities laws where there is a presumption of reliance on a material misrepresentation, see Basic v. Levinson, 485 U.S. 224, 108 S.Ct. 978, 99 L.Ed.2d 194 (1988), plaintiffs would likely have had to prove individual reliance on the allegedly misleading materials under the various state laws applicable to most of these claims. More fundamentally, the complaint itself invoked state laws that, implicated different legal standards on, for example, the warranty claims (the laws contain various privity requirements or the need for an allegedly defective product to fail in service before a warranty claim can be sustained), negligent misrepresentation, negligence, and strict products liability. The state laws implicated by the filing of the nationwide class action also differed on such issues as statutes of limitations; whether pickup trucks are “consumer products;” the application of durational limits on implied warranties; the requirement of reliance to recover for fraud, misrepresentation, and warranty claims; whether intent is a required element of negligent misrepresentation claims; whether comparative fault is a defense; and the relevant test for plaintiffs’ design defect claims.
III. Rule 23 — Relevant Fundamental PRINCIPLES
Before turning to the precise questions at issue on this appeal, it is important that we consider the several basic purposes served by class actions in our contemporary, complex litigation-laden legal system. One of the paramount values in this system is efficiency. Class certification enables courts to treat common claims together, obviating the need for repeated adjudications of the same issues. See 1 Herbert Newberg & Alba Conte, Newberg on Class Actions § 1.06 (3d Ed.1992); General Tel. Co. v. Falcon, 457 U.S. 147, 149, 102 S.Ct. 2364, 2366, 72 L.Ed.2d 740 (1982).
The Supreme Court has articulated other important objectives served by class actions. Class actions achieve “the protection of the defendant from inconsistent obligations, the protection of the interests of absentees, the provision of a convenient and economical means for disposing of similar lawsuits, and *784the facilitation of the spreading of litigation costs among numerous litigants with similar claims.” United States Parole Comm’n v. Geraghty, 445 U.S. 388, 402-03, 100 S.Ct. 1202, 1211-12, 63 L.Ed.2d 479 (1980). The Court has explained the significance of the last goal as
an evolutionary response to the existence of injuries unremedied by the regulatory action of government. Where it is not economically feasible to obtain relief within the traditional framework of a multiplicity of small individual suits for damages, aggrieved persons may be without any effective redress unless they may employ the class-action device.
Deposit Guaranty National Bank v. Roper, 445 U.S. 326, 339, 100 S.Ct. 1166, 1174, 63 L.Ed.2d 427 (1980); see also 1 NewbeRG & Conte § 1.06 at 1-19. Cost spreading can also enhance the means for private attorney general enforcement and the resulting deterrence of wrongdoing. Id. § 1.06 at 1-18 to 1-20.
The law favors settlement, particularly in class actions and other complex cases where substantial judicial resources can be conserved by avoiding formal litigation. See Newbeeg & Conte § 11.41 at 11-85 (citing cases); Cotton v. Hinton, 559 F.2d 1326, 1331 (5th Cir.1977); Van Bronkhorst v. Safeco Corp., 529 F.2d 943, 950 (9th Cir.1976). The parties may also gain significantly from avoiding the costs and risks of a lengthy and complex trial. See First Com. Corp. of Boston Customer Accts Litig., 119 F.R.D. 301, 306-07 (D.Mass.1987). These economic gains multiply when settlement also avoids the costs of litigating class status — often a complex litigation within itself. Furthermore, a settlement may represent the best method of distributing damage awards to injured plaintiffs, especially where litigation would delay and consume the available resources and where piecemeal settlement could result, in the Rule 23(b)(1)(B) limited fund context, in a sub-optimal distribution of the damage awards. See, e.g., In re Dennis Greenman Secur. Litig., 829 F.2d 1539, 1542 (11th Cir.1987).
Thus, courts should favor the use of devices that tend to foster negotiated solutions to these actions. Prima facie, this would include settlement classes. True, it was once thought that mass tort actions were ordinarily not appropriate for class treatment, see Fed.R.Civ.P. 23 Advisory Committee’s note, subdivision (b)(3), 39 F.R.D. 69, 103 (1966). It has also been argued that mass tort eases strain the boundaries of Rule 23. See Bruce H. Nielson, Was the 1966 Advisory Committee Right?: Suggested Revisions of Rule 23 to Allow More Frequent Use of Class Actions in Mass Tort Litigation, 25 Habv.J.Legis. 461 (1988) (suggesting necessity of rule revisions to accommodate class action treatment of mass torts). However, the applicability of Rule 23 to mass tort cases has become commonplace, and the use of the class action device, specifically the (b)(3) class, has created some of the largest and most innovative settlements in these contexts. Prominent examples include the recent $4.2 billion settlement of the breast implant litigation. See In re Silicone Gel Breast Implant Prods. Liab. Litig., 1994 WL 578353 (N.D.Ala.1994).
Despite the potential benefits of class actions, there remains an overarching concern — that absentees’ interests are being resolved and quite possibly bound by the operation of res judicata even though most of the plaintiffs are not the real parties to the suit. The protection of the absentees’ due process rights depends in part on the extent the named plaintiffs are adequately interested to monitor the attorneys (who are, of course, presumed motivated to achieve maximum results by the prospect of substantial fees), and also on the extent that the class representatives have interests that are sufficiently aligned with the absentees to assure that the monitoring serves the interests of the class as a whole. In addition, the court plays the important role of protector of the absentees’ interests, in a sort of fiduciary capacity, by approving appropriate representative plaintiffs and class counsel.
Another problem is that class actions create the opportunity for a kind of legalized blackmail: a greedy and unscrupulous plaintiff might use the threat of a large class action, which can be costly to the defendant, to extract a settlement far in excess of the *785individual claims’ actual worth. Because absentees are not parties to the action in any real sense, and probably would not have brought their claims individually, see Mars Steel v. Continental Illinois Nat’l Bank & Trust, 834 F.2d 677, 678 (7th Cir.1987), attorneys or plaintiffs can abuse the suit nominally brought in the absentees’ names. As one court has noted, “[t]his fundamental departure from the traditional pattern in Anglo-American litigation generates a host of problems .... ” Id.
The drafters designed the procedural requirements of Rule 23, especially the requisites of subsection (a), so that the court can assure, to the greatest extent possible, that the actions are prosecuted on behalf of the actual class members in a way that makes it fair to bind their interests. The rule thus represents a measured response to the issues of how the due process rights of absentee interests can be protected and how absentees’ represented status can be reconciled with a litigation system premised on traditional bipolar litigation. Moreover, the requirement in Rule 23(c) that the court decide certification motions “as soon as practicable,” see note 1 supra, aims to reduce even further the possibility that a party could use the ill-founded threat of a class action to control negotiations or the possibility that absentees’ interests could be unfairly bound. Hence, the procedural formalities of certification are important even if the case appears to be headed for settlement rather than litigation.
This expanded role of the court in class actions (relative to conventional bipolar litigation) continues even after certification. While the parties in a normal suit do not ordinarily require a judge’s approval to settle the action, class action parties do. Rule 23(e) provides: “A class action shall not be dismissed or compromised without the approval of the court, and notice of the proposed dismissal or compromise shall be given to all members of the class in such manner as the court directs.” Fed.R.CxvP. 23(e). Courts and commentators have interpreted this rule to require courts to “independently and objectively analyze the evidence and circumstances before it in order to determine whether the settlement is in the best interest of those whose claims will be extinguished.” 2 NewbeRG & Conte § 11.41 at 11-88 to 11-89. “Under Rule 23(e) the district court acts as a fiduciary who must serve as a guardian 'of the rights of absent class members.... [T]he court cannot accept a settlement that the proponents have not shown to be fair, reasonable and adequate.” Grunin v. International House of Pancakes, 513 F.2d 114, 123 (8th Cir.), cert. denied, 423 U.S. 864, 96 S.Ct. 124, 46 L.Ed.2d 93 (1975); Malchman v. Davis, 706 F.2d 426, 433 (2d Cir.1983); Sala v. National RR Passenger Corp., 721 F.Supp. 80 (E.D.Pa.1989); see also Piambino v. Bailey, 610 F.2d 1306 (5th Cir.), cert. denied, 449 U.S. 1011, 101 S.Ct. 568, 66 L.Ed.2d 469 (1980).
Before sending notice of the settlement to the class, the court will usually approve the settlement preliminarily. This preliminary determination establishes an initial presumption of fairness when the court finds that: (1) the negotiations occurred at arm’s length; (2) there was sufficient discovery; (3) the proponents of the settlement are experienced in similar litigation; and (4) only a small fraction of the class objected. See 2 Newbeeg & Conte § 11.41 at 11-91.
As noted above, this court has adopted a nine-factor test to help district courts structure their final decisions to approve settlements as fair, reasonable, and adequate as required by Rule 23(e). See Girsh v. Jepson, 521 F.2d 153, 157 (3d Cir.1975). Those factors are: (1) the complexity and duration of the litigation; (2) the reaction of the class to the settlement; (3) the stage of the proceedings; (4) the risks of establishing liability; (5) the risks of establishing damages; (6) the risks of maintaining a class action; (7) the ability of the defendants to withstand a greater judgment; (8) the range of reasonableness of the settlement in light of the best recovery; and (9) the range of reasonableness of the settlement in light of all the attendant risks of litigation. Id. The proponents of the settlement bear the burden of proving that these factors weigh in favor of approval. See GM Interchange, 594 F.2d 1106, 1126 n. 30 (7th Cir. *7861979); Holden v. Burlington Northern, Inc., 665 F.Supp. 1398, 1407 (D.Minn.1987); MCL 2d § 30.44. The findings required by the Girsh test are factual, see Malchman v. Davis, 706 F.2d at 434; Plummer v. Chemical Bank, 668 F.2d 654, 659 (2d Cir.1982), which will be upheld unless they are clearly erroneous, Weinberger v. Kendrick, 698 F.2d 61, 73 (2d Cir.1982), cert. denied, 464 U.S. 818, 104 S.Ct. 77, 78 L.Ed.2d 89 (1983); In re Corrugated Container Antitrust Litig., 643 F.2d 195, 207 (5th Cir.1981).
IV. SETTLEMENT CLASSES
This appeal challenges (among other things) the district court’s class certification order. Before we may address the propriety of the court’s order we must first decide whether it is ever proper to certify a class for settlement purposes only. We therefore begin our analysis with a closer look at how settlement classes operate.
A. Nature of the Device
As we have explained above, a settlement class is a device whereby the court postpones the formal certification procedure until the parties have successfully negotiated a settlement, thus allowing a defendant to explore settlement without conceding any of its arguments against certification. Despite the directive of Rule 23(c) that courts certify actions as soon as practicable, when a class action has been filed before the settlement has been arrived at courts will often delay the certification determination during the pendency of settlement discussions. If the settlement negotiations succeed, courts will certify the class for settlement purposes only and send a combined notice of class pendency and settlements to the class members. Thus, by the time the court considers certification, the defendant has essentially stipulated to the existence of the class requirements since it now has an interest in binding an entire class with its proffered settlement.
By specifying certification for settlement purposes only, however, the court preserves the defendant’s ability to contest certification should the settlement fall apart. Because the court indulges the assumption of the class’s existence only until a settlement is reached or the parties abandon the negotiations, settlement classes are also sometimes referred to as temporary or provisional classes. Sometimes the specification may also be seen as assuming that the class may only meet the requirements of Rule 23 if the action is settled, and that certification may in fact be inappropriate if the action will actually be litigated. In any event, notwithstanding that there is an absence of clear textual authorization for settlement classes, many courts' have indulged the stipulations of parties by establishing temporary classes for settlement purposes only. See, e.g., Mars Steel v. Continental Illinois Nat’l Bk. & Trust, 834 F.2d 677 (7th Cir.1987); Weinberger v. Kendrick, 698 F.2d 61 (2d Cir.1982), cert. denied, 464 U.S. 818, 104 S.Ct. 77, 78 L.Ed.2d 89 (1983); In re A.H. Robins Co., 880 F.2d 709, 738-39 (4th Cir.1989); In re Dennis Greenman Sec. Litig., 829 F.2d 1539, 1543 (11th Cir.1978); Plummer v. Chemical Bank, 668 F.2d 654 (2d Cir.1982); In re Beef Industry Antitrust Litig., 607 F.2d 167, 173 (5th Cir.1979); Malchman v. Davis, 706 F.2d 426, 433-34 (2d Cir.1983); In re Taxable Mun. Bond Sec. Litig., 1994 WL 643142 (E.D.La. Nov. 15, 1994); In re Silicone Gel Breast Implant Prod. Liab. Litig., 1994 WL 578353 (N.D.Ala.1994); In re First Commodity Corp. of Boston, 119 F.R.D. 301, 306-08 (D.Mass.1987); In re Bendectin, 102 F.R.D. 239, 240 (S.D.Oh.1984), rev’d on other grounds, 749 F.2d 300 (6th Cir.1984); In re Mid-Atlantic Toyota Antitrust Litig., 564 F.Supp. 1379, 1388-90 (D.Md.1983); In re Chicken Antitrust Litig., 560 F.Supp. 957, 960 (N.D.Ga.1980).
There has been a great deal of commentary, both critical7 and laudatory,8 of the use of these “settlement classes.” And some courts have criticized these accommodations *787of the negotiating parties and expressed their ambivalence while continuing nonetheless to use them. See, e.g., Mars Steel, 834 F.2d 677 (7th Cir.1987) (describing considerable dangers of settlement classes but ultimately upholding the settlement). Before we interpret the dictates of Rule 28 with respect to settlement classes, it will be useful to survey both the criticism and the praise.
B. Perceived Problems of Settlement Classes
A number of commentators, particularly the authors of the first edition of the Manual for Complex Litigation, have voiced serious concerns about settlement classes. These criticisms have focused on the fact that Rule 23, a carefully constructed scheme intended to protect the rights of absentees that necessarily relies on active judicial participation to protect those interests, does not authorize a separate category of class certification that would permit a dilution of or dispense with the subsection (a) criteria. § 1.46; see also Mars Steel v. Continental Ill. Nat’l Bank & Trust, 834 F.2d 677, 680 (7th Cir.1987); In re Baldwin United, 105 F.R.D. 475 (S.D.N.Y.1984). Other criticisms focus on the potential prejudice to the parties and the institutional threat posed to the court. See, e.g., Coffee, supra note 7.
Rule 23 does not in terms authorize the deferral of class certification pending settlement discussions. Indeed, Rule 23(c) provides: “As soon as practicable after the commencement of an action brought as a class action, the court shall determine by order whether it is to be so maintained.” Fed. R.Civ.P. 23(a) (emphasis supplied). Deliberately delaying a class certification determination so that settlement discussions can proceed clearly does not represent an effort to resolve the issue “as soon as practicable.” As Judge Posner has noted, “[i]t is hard to see why the propriety of maintaining the'suit as a class action could not ‘practicably1 have been determined much earlier. And, common though the practice of deferring class certification while settlement negotiations are going on is, it not only jostles uneasily with the language of Rule 23(c)(1) but also creates practical problems.” Mars Steel, 834 F.2d at 680.
The danger here is that the court cannot properly discharge its duty to protect the interests of the absentees during the disposition of the action. Because the class has not yet been defined, the court lacks the information necessary to determine the identity of the absentees and the likely extent of liability, damages, and expenses of preparing for trial. See MCL 2d § 30.45 at 243 (“No one may know how many members are in the class, how large their potential claims are, what the strengths and weaknesses of the parties’ positions are, or how much the class will benefit under the settlement.”); In re Baldwin United, 105 F.R.D. 475, 481 (S.D.N.Y.1984). Moreover, the court performs its role as supervisor/protector without the benefit of a full adversarial briefing on the certification issues. With less information about the class, the judge cannot as effectively "monitor for collusion, individual settlements, buy-offs (where some individuals use the class action device to benefit themselves at the expense of absentees), and other abuses. See In re Beef Indus. Antitrust Litig., 607 F.2d at 174. For example, if the court fails to define the class before settlement negotiations commence, then during the settlement approval phase the judge will have greater difficulty detecting if the parties improperly manipulated the scope of the class in order to buy the defendant’s acquiescence.
Settlement classes also make it more difficult for a court to evaluate the settlement by depriving the judge of the customary structural devices of Rule 23 and the presumptions of propriety that they generate. Ordinarily, a court relies on class status, particularly the adequacy of representation required to maintain it, to infer that the settlement was the product of arm’s length negotiations. Cf. Weinberger v. Kendrick, 698 F.2d 61, 74 (2d Cir.1983) (noting protracted nature of negotiations in approving settlement); City of Detroit v. Grinnell, 495 F.2d 448, 463 (2d Cir.1974) (same); In re Baldwin-United, 105 F.R.D. 475, 482 (S.D.N.Y.1984) (same). Where the court has not yet certified a class *788or named its representative or counsel, this assumption is questionable.
In effect, settlement classes can, depending how they are used, evade the processes intended to protect the rights of absentees. Indeed, the draft of the MCL (Third), although considerably more receptive to settlement classes than the earlier editions of the Manual, explains that “[t]he problem presented by these requests is not the lack of sufficient information and scrutiny, but rather the possibility that fiduciary responsibilities of class counsel or class representatives may have been compromised.” MCL (Third) (draft) at 193. Even some courts successfully using these devices to achieve settlements apparently recognize these dangers since they certify these actions more cautiously than ordinary classes. See, e.g., Ace Heating & Plumbing Co. v. Crane Co., 453 F.2d 30, 33 (3d Cir.1971) (court must be doubly careful where negotiation occurs before certification and designation of a class counsel); In re Beef Antitrust Litig., 607 F.2d 167, 176-77 (5th Cir.1979) (examining though ultimately rejecting the charge that collusion precluded the certification of the settlement class); Simer v. Rios, 661 F.2d 655, 664-66 (7th Cir.1981) (requiring a higher showing of fairness where settlement negotiated prior to certification); Weinberger v. Kendrick, 698 F.2d 61, 69 (2d Cir.1982) (judge made findings about discovery and counsel).
In particular, settlement classes create especially lucrative opportunities for putative class attorneys to generate fees for themselves without any effective monitoring by class members who have not yet been apprised of the pendency of the action. Moreover, because the court does not appoint a class counsel until the case is certified, attorneys jockeying for position might attempt to cut a deal with the defendants by underselling the plaintiffs’ claims relative to other attorneys.9 Unauthorized settlement negotiations occurring before the certification determination thus “create the possibility of negotiation from a position of weakness by the attorney who purports to represent the class.” GM Interchange Litigation, 594 F.2d 1106, 1125 (7th Cir.1979). Pre-certification negotiations also hamper a court’s ability to review the true value of the settlement or the legal services after the fact. See supra at 31. In addition, unauthorized negotiations also result in denying other plaintiffs’ counsel information that is necessary for them to make an effective evaluation of the fairness of any settlement that results. See id. at 1125.
Framed as an issue of Rule 23(a) requisites, these considerations implicate adequacy of representation concerns: “[arguments in opposition to settlement classes have merit when they are addressed to the problem of inadequate representation or possible collusion among the named plaintiffs and some or all defendants.” In re Baldwin-United Corp., 105 F.R.D. 475, 480 (S.D.N.Y.1984). Another court has warned that the “danger of a premature, even a collusive, settlement [is] increased when as in this case the status of the action as a class action is not determined until a settlement has been negotiated, with all the momentum that a settlement agreement generates....”; Mars Steel, 834 F.2d at 680; see also Malchman, 706 F.2d at 433 (recognizing special potential for collusion or undue pressure by defendants in settlement negotiations); Weinberger, 698 F.2d at 73 (requiring a higher showing of fairness to accommodate greater potential for improper settlement). Settlement classes, which constitute ad hoe adjustments to the carefully designed class action framework constructed by Rule 23, lack the regulatory mechanisms that ordinarily check this improper behavior: “There is in fact little or no individual client consultation and no judicial oversight of a hidden process of wheeling and dealing to maximize overall recovery and fees for hundreds and thousands of massed cases.” In re Joint Eastern & Southern District Asbestos Litig., 129 B.R. 710, 802 (E. & S.D.N.Y.1991) (discussing the ramifications of class treatment of mass torts).
In addition to these procedural problems (and the problems created for a judge trying to evaluate both class status and the adequa*789cy of a class settlement simultaneously) the earlier achievement of settlement through the use of a settlement class also can lead to a settlement that may provide inadequate consideration in exchange for the release of the class’s claims. With early settlement, both parties have less information on the merits. That is, they have less information on the membership of the class, on the size of potential claims, on whether the settlement purports to resolve class or individual claims, on the strengths and weaknesses of the case, and on how class members will benefit from the settlement. See MCL 2d § 30.45 at 243-44; 2 Newberg & Conte § 11.09 at 11-13. Without the benefit of more extensive discovery, both sides may underestimate the strength of the plaintiffs’ claims.
Turning to the question of due process rights, we note that class members may, as a result of these information deficiencies, not be in a fair position at this early stage to evaluate whether or not the settlement represents a superior alternative to litigating. Perhaps more troubling in light of the reality that absentees tend to lack a real understanding of the actions supposedly pursued in their names is that, “where notice of the class action is ... sent simultaneously with the notice of the settlement itself, [the settlement class paradigm], the class members are presented with what looks like a fait accompli.” Mars Steel, 834 F.2d at 680-81. Thus, even if they have enough information to conclude the settlement is insufficient and unsatisfactory, see In re Beef Antitrust Litig., 607 F.2d 167, 173 n. 4 (5th Cir.1979), cert. denied, 452 U.S. 905, 101 S.Ct. 3029, 69 L.Ed.2d 405 (1981), the mere presentation of the settlement notice with the class notice may pressure even skeptical class members to accept the settlement out of the belief that, unless they are willing to litigate .their claims individually — often economically infeasible — they really have no choice.
In a different vein, a number of cases have also criticized settlement classes on the grounds that they create an opportunity for “one-way intervention,” allowing putative class members to wait to see whether they think the settlement is favorable before deciding whether they want to be bound by it. See McDonald v. Chicago Milw. Corp., 565 F.2d 416, 420 (7th Cir.1977); Watkins v. Blinzinger, 789 F.2d 474, 475 n. 3 (7th Cir.1986) (“A deferred ruling [on certification] converts the class action to an opportunity for one-way intervention, which Rule 23 is designed to avoid-”); Premier Electrical Constr. Co. v. National Elec. Contractors Ass’n, Inc., 814 F.2d 358, 363 (7th Cir.1987) (criticizing delay of certification). Because class members have the opportunity to wait until the outcome is known (i.e., the settlement’s terms are determined) to decide whether they want to be bound by the result, courts and- defendants are exposed to the same potential for multiple lawsuits that class actions are designed to avoid, and the supposed advantages of settlement classes are largely eroded.
Perhaps more troubling, the possibility of pre-certification negotiation and settlement may facilitate the filing of strike suits. Since settlement classes can involve a settlement achieved either before or after the filing of class claims, recognition of the settlement class device allows plaintiffs to file as class actions cases that counsel never intended to have certified, but instead only to settle the claims individually. Mars Steel, 834 F.2d at 681 (“[P]laintiffs will be tempted to add class claims in order to intimidate the defendant, then delete them by way of compromise.”). Knowing that they would not face judicial scrutiny if they settle before certification, plaintiffs’ lawyers face no deterrent from attempting to extract larger settlements by threatening class litigation than they could with the cases filed individually.
In many respects then, the failings of settlement classes are a function of the dearth of information available to judges attempting to scrutinize the settlements in accordance with their Rule 23(e) duties. Because the. issue of certification is never actively contested, the judge never receives the benefit of the adversarial process that provides the information needed to review propriety of the class and the adequacy of settlement. This problem is exacerbated where the parties agree on a settlement of the case before the class action is filed, since a motion for certifi*790cation and settlement are presented simultaneously.
Last, but by no means least, the use of settlement classes also risks transforming the courts into mediation forums. See Coffee, supra note 7, at A15. Cases could be filed without any expectation or intention of litigation, with the foreknowledge that the natural hydraulic pressure for settlement may in fact lead to a class settlement, especially given the incentive a defendant has to bind as many potential claimants as possible with an approved class settlement. Courts may approve these class settlements even if the case is highly inappropriate for class treatment, since judges confronting the reality of already over-taxed judicial resources, see Proposed Long Range Plan for the Federal Courts (March 1995) at 9-12, may feel constrained to dispose of such onerous litigation through the settlement class device. The losers in this type of scenario are not only inadequately represented class members but also the federal courts as an institution, because their resources are further sapped by entertaining cases that arguably do not belong there.10 This increased burden will be especially problematic if the standards for certification are relaxed for settlement classes; as this appeal demonstrates, proceedings attendant to settlement class certification can consume considerable federal judicial time.
C. Arguments Favoring Settlement Classes
Although settlement classes are vulnerable to potent criticisms, some important dynamics militate in favor of a judge’s delaying or even substantially avoiding class certification determinations. Because certification so dramatically increases the potential value of the suit to the plaintiffs and their attorneys as well as the potential liability of the defendant, the parties will frequently contest certification vigorously. As a result, a defendant considering a settlement may resist agreeing to class certification because, if the settlement negotiations should fail, it would be left exposed to major litigation. See In re Beef Indus. Antitrust Litig., 607 F.2d at 177-78 (“[A blanket rule against settlement classes] may render it virtually impossible for the parties to compromise class issues and reach a proposed class settlement before a class certification_”); In re Baldwin-United, 105 F.R.D. 475 (S.D.N.Y.1984).
In mass tort cases, in particular, use of a settlement class can help overcome certain elements of these actions that otherwise can considerably complicate efforts to settle. These hurdles include “the large number of individual plaintiffs and lawyers; ... the existence of unfiled claims by putative plaintiffs; and ... the inability of any single plaintiff to offer the settling defendant reliable indemnity protection....” Transgrud, 70 CORNELL L.Rev. at 835. By using the courts to overcome some of the collective action problems particularly acute in mass tort cases, the settlement class device can make settlement feasible. The use of settlement classes can thus enable both parties to realize substantial savings in litigation expenses by compromising the action before formal certification. See 2 Newberg & Conte § 11.09 at 11-13. Through settlement class certification, courts have fostered settlement of some very large, complex cases that might otherwise never have yielded deserving plaintiffs any substantial remuneration.
Settlement classes also increase the number of actions that are amenable to settlement by increasing the rewards of a negotiated solution, in at least four ways. First, the prospect of class certification increases a defendant’s incentive to settle because the settlement would then bind the class members and prevent further suits against the defendant. Second, settlement classes may reduce litigation costs by allowing defendants to stipulate to class certification without forfeiting any of their legal arguments against certification should the negotiations fail. Third, because the payment of settlement proceeds, even relatively small amounts, may palliate class members, settlement can reduce differences among class members, and *791thus make class certification more likely, increasing the value of settlement to the defendant, since a larger number of potential claims can thus be resolved.
Fourth, the use of settlement classes reduces the probability of a successful subsequent challenge to the class-wide settlement. By treating the'class as valid pending settlement, a temporary class facilitates notice to those persons whom the court might consider part of the class. The expanded notice afforded by access to the customary class action notification process protects both the absentees and the defendants by eliminating negotiations between the defendants and the named plaintiffs with respect to the class, definition that could leave the defendant vulnerable to additional suits by absentees whose interests, a court later determines, were not adequately served or protected. 2 Newberg & Conte § 11.27 at 11-10 (citing Midland Mut. Life Ins. Co. v. Sellers, 101 B.R. 921 (Bankr.S.D.Ohio 1989)). Increasing the certainty that the settlement will be upheld augments the value of settling to the defendant and consequently the amount defendants will be willing to pay. Thus, delaying certification, in contravention of a strict reading of Rule 23, encourages settlement, an important judicial policy, by increasing the prospective gains to the defendant , (and thus potentially to the plaintiffs as well) from exploring a negotiated solution.
Moreover, critics of settlement classes may underestimate the safeguards that still inhere. Although courts are often certifying settlement classes with sub-optimal amounts of information, and without the full benefit of the processes meant to protect the absentees’ interests, the provisional certification of a settlement class does not finally determine the absentees’ rights. When the simultaneous notice of the class and the settlement is distributed to the proposed class, objecting class members can still challenge the class on commonality, typicality, adequacy of representation, superiority, and predominance grounds — they are not limited to objections based strictly on the settlement’s terms. 2 Newbekg & Conte ■§ 11.27 at 11-10 (citing Midland Mut. Life Ins. Co. v. Sellers, 101 B.R. at 921).
Furthermore, the view that, in settlement class cases, the court lacks the information necessary to fulfill its role as protector of the absentees, may reflect an assumption that the court’s approval always comes early in the case. See 2 Newberg & Conte § 11.27 at 11-43 to 11-44. While it often does, the certification decision is sometimes made later in the case, when the parties have presumably developed the merits more fully (in discovery or in thé course of wrangling over the settlement terms) and when prior governmental procedures or investigations might have also yielded helpful information. Id. Whatever the timing of the certification ruling, the judge has the duty of passing on the fairness and adequacy of the settlement under Rule 23(e) and also of determining whether the class meets the Rule’s requisites under 23(a).11 Whether or not the court certifies the class before settlement discussions, these duties are the same. 2 Newberg & Conte § 11.27, at 11-46.
Although a judge cannot presume that the putative class counsel actively represented the absentees’ interests, the court can still monitor the negotiation process itself to assure that both counsel and the settlement adequately vindicate the absentees’ interests. Thus, there is no reason to inflexibly limit the use of settlement classes to any specified categories of cases (for example, those' cases with few objectors, those which do not involve partial settlements,12 or those which do not involve an expanded class). Even apparently troublesome litigation activity, such as expanding the class just before settlement approval at the defendant’s request, is no more free from judicial scrutiny in a settlement class context than it would be otherwise. The court still must give notice to the *792now-expanded class and satisfy itself that the requisites of class certification are met.. Id. at 11 — 49. Since the party advocating certification bears the burden of proving appropriateness of class treatment, Davis v. Romney, 490 F.2d 1360 (3d Cir.1974), where the procedural posture is such that the court lacks adequate information to make those determinations, it can and should withhold the relevant approvals. 2 Newberg & Conte § 11.27 at 11-46.
But even if the use of settlement classes did reduce a judge’s capacity to safeguard the class’s interests, it does not necessarily impair the ability of absentees to protect their own interests. Individual class members retain the right to opt out of the class and settlement, preserving the right to pursue their own litigation. See Premier Elec. Const. Co. v. N.E.C.A., Inc., 814 F.2d 358 (7th Cir.1987) (criticizing settlement classes because they ereaté opportunities for one-way intervention). In fact, the use of the settlement class in some sense enhances plaintiffs’ right to opt out. Since the plaintiff is offered the opportunity to opt out of the class simultaneously with the opportunity to accept or reject the settlement offer, which is supposed to be accompanied by all information on settlement, the plaintiff knows exactly what result he or she sacrifices when opting out. See 2 NewbeRG & Conte § 11.27 at 11-51. See In re Beef Indust. Antitrust Litig., 607 F.2d at 174.
In sum, settlement classes clearly offer substantial benefits. However, the very flexibility required to achieve these gains strains the bounds of Rule 23 and comes at the expense of some of the protections the rule-writers intended to construct. As Judge Schwarzer has explained:
one way to see [the settlement class] is as a commendable example of the law’s adaptability to meet the needs of the time — in the best tradition of the Anglo-American common law: But another interpretation might be that it is an unprincipled subversion of the Federal Rules of Civil Procedure. True, if it is a subversion, it is done with good intentions to help courts cope with burgeoning dockets, to enable claimants at the end of the line of litigants to recover compensation, and to allow defendants to manage the staggering liabilities many face. But as experience seems to show, good intentions are not always enough to ensure that all relevant private and public interests are protected. The siren song of Rule 23 can lead lawyers, parties and courts into rough waters where their ethical compass offers only uncertain guidance.
William W. Schwarzer, Settlement of Mass Tort Class Actions: Order Out of Chaos, Cornell L.Rev. (forthcoming).
D. Are Settlement Classes Cognizable Under Rule 23?
Although not specifically authorized by Rule 23, settlement classes are not specifically precluded by it either; indeed, Judge Brieant has read subsection (d), giving the court power to manage the class action, as authorizing the creation of “tentative”, “provisional”, or “conditional” classes through its grant of power to modify or decertify classes as necessary. See, e.g., In re Baldwin-United Corp., 105 F.R.D. 475, 478-79 (S.D.N.Y.1984). And because of the broad grant of authority in Rule 23(d), at least one commentator has noted that the validity of temporary settlement classes is usually not questioned. 2 Newberg & Conte § 11.22 at 11-31. Courts apparently share this confidence. Indeed, one court believed that “[i]t is clear that the Court may provisionally certify the Class for settlement purposes.” South Carolina Nat’l Bank v. Stone, 749 F.Supp. 1419 (D.S.C.1990).
We believe that the “provisional” 13 or “conditional”14 conception of the settlement class device finds at least a colorable textual basis in the Rule. Rule 23(d) enables *793a court to certify a class, if it complies with its duty to assure that the class meets the rule’s requisites by making appropriate Rule 23 findings (see Part IV(E) infra). Some courts appear to have concluded that the built-in flexibility of the Rule, which enables the court to revisit the requisites and modify or decertify the class should its nature change dramatically during the negotiation process, renders it acceptable to determine class status after settlement and thus avoid scrutinizing and adjudicating class status at an earlier stage when the outcome is unknown. See, e.g., In re Baldwin-United, 105 F.R.D. at 483; In re Beef Antitrust Litig., 607 F.2d at 177 (“[T]he Court finds that a conditional class should be certified for the purpose of considering the proposed settlements.”)
Alternatively, some courts have conceived of settlement classes as a “temporary assumption” by the court to facilitate settlement. See Mars Steel, 834 F.2d at 680; In re Beef Indust. Antitrust Litig., 607 F.2d at 177; 2 Newberg & Conte § 11.27 at 11-50. The arguments of the late Herbert Newberg, one of the leading advocates of settlement classes, reflect an assumption that the Rule 23 determinations are merely postponed, not eliminated:
On analysis, however, it would appear that this argument [that courts using settlement classes circumvent the need to test the propriety of the class action according to the specific criteria of Rule 23] may be rebutted by perceiving the temporary settlement class as nothing more than a tentative assumption indulged in by the court_ The actual class ruling is deferred in these circumstances until after hearing on the settlement approval.... At that time, the court in fact applies the class action requirements to determine whether the action should be maintained as a class action....
2 Newbérg & Conte § 11.27 at ■ 11-50.15 Newberg posits, therefore, that the temporary assumption conception of the settlement needs no special authorization since the court eventually follows the ordinary certification process, only deferring it until the settlement approval stage.
Courts have also relied on the more general policies of Rule 23 — promoting justice and realizing judicial efficiencies — to justify this arguable departure from the rule.
[T]he hallmark of Rule 23 is flexibility.... Temporary settlement classes have proved to be quite useful in resolving major class action disputes. While their use may still be controversial, most Courts have recognized their utility and have authorized the parties to compromise their differences, including class action issues through this means.
Weinberger, 698 F.2d at 72-73. One commentator found implicit authorization for settlement classes under a settlement-oriented interpretation of Rule 23:
[Rule 23] provides that a court may certify a common question class action when it will prove “superior to other, available methods for the fair and efficient adjudication of the controversy.” A judicially supervised and approved class action settlement, like a judicially supervised trial, is a means of hearing and determining judicially, in other words “adjudicating,” the value of claims arising from a mass tort. As a result, if conditional certification ■ of the case as a common question class action for settlement purposes would enhance the prospects for a group settlement, then Rule 23 authorizes certification.
Roger H. Transgrud, Joinder Alternatives in Mass Tort Litig., 70 Cornell L.Rev. 779, 835 (1985) (footnotes omitted).
It is noteworthy that resistance to more flexible applications of Rule 23 has diminished over time. See In re Taxable Mun. Bond Secur. Litig., 1994 WL 643142, *4 (E.D.La. 1994) (commenting upon this trend). The evolution of the reception accorded settlement classes has manifested itself in the *794successive versions of the Manual for Complex Litigation. The first edition of the Manual criticized the initiation of settlement negotiations before certification, and discouraged all such negotiations. See MCL 1st § 1.46. The second edition recognizes the potential benefits of settlement classes but still cautioned that “the court should be wary of presenting the settlement to the class.” MCL § 30.45 at 243. The (draft) third version acknowledges that “Settlement classes offer a commonly used vehicle for the settlement of complex litigation” and aims only to supervise rather than discourage their use. See MCL § 30.45 at 192.
A survey of the caselaw confirms the impression that resistance to settlement classes has diminished: few cases since the late 1970’s and early 1980’s even bother to squarely address the propriety of settlement classes. Moreover, no court of appeals that has had the opportunity to comment on the propriety of settlement classes has held that they constitute a per se violation of Rule 23. See, e.g., Ace Heating & Plumbing Co. v. Crane Co., 453 F.2d 30, 33 (3d Cir.1971) (finding no prohibition but granting absentees standing to appeal settlement approval); Marshall v. Holiday Magic, Inc., 550 F.2d 1173, 1176 (9th Cir.1977) (describing how court approved combined notice of the pen-dency of the class and the terms of the proposed settlement); In re Beef Antitrust Litig., 607 F.2d at 167; Corrugated Container Antitrust Litig., 643 F.2d 195, 223 (5th Cir.1981) (upholding settlement despite pre-certification negotiations with some defendants); Weinberger v. Kendrick, 698 F.2d 61 (2d Cir.1982); Mars Steel, 834 F.2d at 681 (criticizing settlement classes but ultimately approving settlement). But some courts recognize that this practice represents a significant departure from the usual Rule 23 scenario and thereby counsel that courts should scrutinize these settlements even more closely.
We acknowledge that settlement classes, conceived of either as provisional or conditional certifications, represent a practical construction of the class action rule. Such construction affords considerable economies to both the litigants and the judiciary and is also fully consistent with the flexibility integral to Rule 23. A number of other jurisdictions have already accepted settlement classes as a reasonable interpretation of Rule 23 and thereby achieved these substantial benefits. Although we appreciate the concerns raised about the device, we are confident that they can be addressed by the rigorous applications of the Rule 23 requisites by the courts at the approval stages, as we discuss at greater length herein. For these reasons, we hold that settlement classes are cognizable under Rule 23.
E. Are the Rule 23(a) and (b) Findings Required for Settlement Classes ? Does Finding the Settlement to Be Fair and Reasonable Serve as a Surrogate for the Findings'?
There is no explicit requirement in Rule 23 that the district judge make a formal finding that the requisites of the rule have been met in order to certify a class. However, most district judges have routinely done so, assuming that it was required, and in published opinions, a number of courts have endorsed or at least acknowledged the compelling policy reasons for doing so. See, e.g., Eisenberg v. Gagnon, 766 F.2d 770, 785 (3d Cir.1985); Plummer, 668 F.2d at 659; Interpace Corp. v. Philadelphia, 438 F.2d 401, 404 (3d Cir.1971); MCL2d § 30.13 (“The judge should enter findings and conclusions after the hearing, addressing each of the applicable requirements of Rule 23(a) and (b).”). For example, where there has been some dispute over certification, a court should give the litigants, particularly the absentees, some statement of the reasons for its decision. Eisenberg, 766 F.2d at 785. Articulated findings also simplify the review of complex cases generally. Id. With respect to settlement classes, we hold that courts must make the findings because the legitimacy of settlement classes depends upon fidelity to the fundaments of Rule 23.16
*795Inasmuch as collusion, inadequate prosecution and attorney inexperience are the paramount concerns in precertification settlements, see Malchman, 706 F.2d at 483; Beef, 607 F.2d at 174, the need for the adequacy of representation finding is particularly acute in settlement class situations, given the inquiry’s purpose of detecting eases where there is a “likelihood that the litigants are involved in a collusive suit....” Eisen v. Carlisle & Jacquelin, 391 F.2d 555, 562 (2d Cir.1968).
There appears to be no authority contra this practice. Indeed, the courts and commentators that have endorsed settlement classes have seemed to assume that the approving court made the requisite class determinations at some point. For example, New-berg’s argument rebutting the charge that the “tentative assumption” of class status by the court to foster settlement evades the Rule’s strictures continues:
The actual class ruling is deferred in these circumstances until after [the] hearing on the settlement approval, following notice to the class. At that time, the court in fact applies the class action requirements to determine whether the action should be maintained as a class action....
2 NewbeRG & Conte § 11.27 11-50. See also Whitford v. First Nationwide Bk., 147 F.R.D. 135, 142 (WD Ky.1992) (disregarding even the possibility that these classes would not have to meet all of the normal certification requisites). Even the cases where the courts did not recognize a need to make the determinations demonstrate a heightened concern for fairness and a more cautious approach to settlement approval. See Ace Heating & Plumbing Co. v. Crane Co., 453 F.2d 30, 33 (3d Cir.1971) (court must be doubly careful where negotiation occurs before certification and designation of a class counsel); Mars Steel, 834 F.2d at 681 (applying a higher standard of fairness); Simer v. Rios, 661 F.2d 655, 664-66 (7th Cir.1981) (requiring a higher demonstration of fairness); Weinberger v. Kendrick, 698 F.2d 61, 69 (2d Cir.1982) (emphasizing the extensive discovery and ability and experience of counsel).
Some courts have certified settlement classes “without articulating or consciously applying Rule 23 tests.” ' 2 Newberg & Conte- § 11.27 at 11-52. See, e.g., Mars Steel, 834 F.2d at 681 (suggesting that the certification procedure may not be necessary to combat the potential for abuse created by the use of settlement classes since that potential is “held in check by the requirement that the judge determine the fairness of the settlement ... ”); Weinberger v. Kendrick, 698 F.2d at 73 (determination that proposed settlement is fair, reasonable and adequate substitutes for Rule 23 findings); In re Beef Antitrust Litig., 607 F.2d 167, 177 (5th Cir.1979); City of Detroit v. Grinnell, 495 F.2d 448, 464-65 (2d Cir.1974) (rejecting contention that the court erred when it approved a settlement and acquiesced in the settlement’s assumption of the existence of a proper class). Some courts neglecting the findings have taken the view that the notice of proposed settlement, which must be preliminarily approved by the court, “carries the necessary implication that the action complies with Rule 23.” Beef, 607 F.2d at 177.
We disagree both with this suggestion and with the conclusion that a- fairness determination is a surrogate for Rule 23 findings.17 Even if we set aside the problem of the court’s inadequate information, the inquiry into the settlement’s fairness cannot conceptually replace the inquiry into the propriety of class certification. Normally, a court makes the required commonality and typicality determinations by referencing the original class complaints in order to assure that the claims alleged by the named plaintiffs are common to the class (although the class need not share every claim in common, Hassine v. Jeffes, 846 F.2d 169, 177-78 (3d Cir.1988)), and that the claims alleged,by the named *796plaintiff occupy approximately the same position of centrality to the named plaintiffs as they do to the rest of the class. Weiss v. York Hosp., 745 F.2d 786, 810 (3d Cir.1984), cert. denied, 470 U.S. 1060, 105 S.Ct. 1777, 84 L.Ed.2d 836 (1985). Neither the existence of a settlement nor the terms of settlement affect the nature of this important inquiry.
The Rule 23(a) class inquiries (numerosity, commonality, typicality, and adequacy of representation) constitute a multipart attempt to safeguard the due process rights of absentees. Thus, the ultimate focus falls on the appropriateness of the class device to assert and vindicate class interests. Conversely, however, the process of negotiation does not reveal anything about commonality and typicality. One might argue that these requisites are merely means to the end of vindicated rights, and that observing the process of negotiation could demonstrate adequate vindication — the true aim of the Rule. In our view, a court cannot infer that the rights of the entire class were vindicated without having assured that commonality and typicality were satisfied.
The 23(b)(3) determination is also important in the regulatory scheme. To be certified as a (b)(3) class, the judge must determine that “questions of law or fact common to the members of the class predominate over any questions affecting only individual members and that a class action is superior to other available methods for the fair and efficient adjudication of the controversy.” Fed.R.Civ.P. 23(b)(3).18 But the settlement approval inquiry is far different from the certification inquiry. In settlement situations, the superiority requirement arguably translates into the question whether the settlement is a more desirable outcome for the class than individualized litigation, and may assure’ that the settlement has not grossly undervalued plaintiffs’ interests. But even if this is so, a point we neither concede nor decide, there remains the concern about conflicts between those appointed to represent class interests — the lawyers and named plaintiffs — and the rest of the class. These concerns, particularly acute with settlement classes, concentrate the focus of the certification inquiries on the representational elements.
Certainly, evaluating the settlement can yield some information relevant to the adequacy of representation determination under 23(a)(4). The settlement evaluation involves two types of evidence: a substantive inquiry into the terms of the settlement relative to the likely rewards of litigation, see Weinberger, 698 F.2d at 73; Protective Comm. for Indep. Stockholders v. Anderson, 390 U.S. 414, 424, 88 S.Ct. 1157, 1163, 20 L.Ed.2d 1 (1968), and a procedural inquiry into the negotiation process. The focus on the negotiation process results from the realization that a judge cannot really make a substantive judgment on the issues in the case without conducting some sort of trial on the merits, exactly what the settlement is intended to avoid. See Malchman v. Davis, 706 F.2d at 433. Instead, the court determines whether negotiations were conducted at arms’ length by experienced counsel after adequate discovery, in which case there is a presumption that the results of the process adequately vindicate the interests of the absentees. Weinberger, 698 F.2d at 74; City of Detroit v. Grinnell, 495 F.2d at 463; Baldwin-United, 105 F.R.D. at 482 (“In order to supplement judicial examination of the substance of a compromise agreement, and because a court cannot conduct a trial in order to avoid a trial, attention must be paid to the process by which a settlement has been reached.”).
Although the procedural focus on the fairness determination yields information pertinent to the adequacy of representation inquiry, it cannot fully satisfy the inquiry. That is because reliance on the negotiation process used to approve the settlement to satisfy the class certification requirements puts excessive pressure on the settlement approved determinations, and, more fundamentally, such a reliance may be circular. Cf. 2 New-beRG & Conte § 11.28 at 11-54 (suggesting a greater need for a court to carefully articu*797late if reasons for settlement approval where the class was not separately certified).
Courts approving settlements have examined the negotiating process in light of the “experience of counsel, the vigor with which the case was prosecuted, and the coercion or collusion that may have marred the negotiations themselves.” Malchman v. Davis, 706 F.2d 426, 433 (2d Cir.1983) (citing Weinberger, 698 F.2d at 73; Grinnell, 495 F.2d at 465.). Some of these courts have suggested that the fact that vigorous, arm’s length negotiations occurred should allay concerns about adequacy of representation. But these inferences depend on the implicit assumption that the lawyers actually negotiating really were doing so on behalf of the entire class, see 2 NbwbeRG & Conte § 11.28 at 11-59, assumptions which are clearly unjustified in a, context where the potential for intra-class conflict further emperils the class’s representation. Far too much turns on the adequacy of representation to accept it on blind faith.
Without determining that the class actually was adequately represented, the district judge has no real basis for assuming that the negotiations satisfactorily vindicated the interests of all the absentees. The focus on the negotiation process also cannot address the part of the adequacy of representation inquiry intended to detect situations where the named plaintiffs are unsuitable representatives of the absentees’ claims. To state that class members were united in the interest of maximizing over-all recovery begs the question. Although that observation might allay some concern about a conflict between the attorney and the class, a judge must focus on the settlement’s distribution terms (or those sought) to detect situations where some class members’ interests diverge from those of others in the class. For example, a settlement that offers considerably more value to one class of plaintiffs than to another may be trading the claims of the latter group away in order to enrich the former group.
In short, the prophylactic devices used by judges to approve these pre-certification settlements without ever formally certifying the class fail to satisfy the requirements of Rule 23. Without determining that the class claims are common and typical of the entire putative class and that the class representatives and their counsel are adequate representatives, we have no assurance that the district court fully appreciated the scope and nature of the interests at stake.19 Finally, we note that courts adopting the view that the formal class determinations are not necessary for settlement classes may be contravening not only the language of the rule but also the Supreme Court’s requirement in General Telephone Co. of Southwest v. Falcon, 457 U.S. 147, 160, 102 S.Ct. 2364, 2372, 72 L.Ed.2d 740 (1982) (disapproving the trial court’s insufficient scrutiny of the named plaintiffs capacity to adequately represent the class), that “[ajctual, not presumed, conformance with Rule 23(a) remains, however, indispensable.” Thus, while we approve the provisional certification of a settlement class to facilitate settlement discussions, final settlement approval depends on the finding that the class met all the requisites of Rule 23.
F. Can There be a Valid Settlement Class That Would Not Serve as a Valid Litigation Class?
As we have previously explained, courts using the settlement class device must at some point definitively certify the class and satisfy themselves that the requisites of Rule 23 have been satisfied. To avoid that pro*798cess entirely would dismantle the rule’s carefully constructed mechanism that serves to protect absentees’ due process rights. Moreover, despite some courts’ suggestions that the standards are less rigorous for settlement classes, we do not believe that Rule 23 authorizes separate, liberalized criteria for settlement classes.
At the outset we note that, while some other courts have nominally complied with the rule, they appear to have assumed that lower standards apply in settlement class cases. See Officers for Justice v. Civil Serv. Comm’n of San Francisco, 688 F.2d 615, 633 (9th Cir.1982), cert. denied, 459 U.S. 1217, 103 S.Ct. 1219, 75 L.Ed.2d 456 (1983) (“[Certification issues raised by class action litigation that is resolved short of a decision on the merits must be viewed in a different light.”); Fisher Bros. v. Phelps Dodge Indus. Inc., 604 F.Supp. 446, 450 (E.D.Pa.1985); In re Dennis Greenman Secur. Litig., 829 F.2d 1539, 1543 (11th Cir.1987) (“In reviewing settlement certifications, a special standard has been employed.”); A.H. Robins, 880 F.2d at 740 (in deciding whether to certify a class, settlement is at least an important factor in favor and might even be a per se ground for certification); Manual.2d at § 30.45. Other courts have stated that settlement reduces the potential conflicts among the class and thus enhances the likelihood of meeting the criteria, presumably the same criteria a litigation class must satisfy. See, e.g., Bowling v. Pfizer, Inc., 143 F.R.D. 141, 159 (S.D.Oh.1992). Newberg is of this view. See 2 New-berg & Conte § 11.28, at 11-58.
According to Newberg, though settlement does not impact the numerosity requirement it may indeed increase the likelihood of meeting the commonality and typicality inquiries. “Typicality of claims in a settlement class context requires proof that the interests of the class representative and the class are commonly held for the purposes of receiving similar or overlapping benefits from a settlement.” 2 NewbeRG & Conte § 11.28 at 11-58. On this theory, because the court has delayed the findings until the outcome of the litigation (i.e., the settlement agreement) is known, the judge conducts the inquiry based on the relative rewards to the class members rather than based on the various legal claims of class members. So long as all plaintiffs get similar benefits from the settlement, irrespective of the different strengths of their initial claims, the commonality and typicality inquiries are viewed as likely to be satisfied.
Under this approach, the adequate representation inquiry is also simplified in the settlement class context by a result-oriented approach toward the class requirement findings. Rather than asking whether the lawyers have sufficient resources and skills to prosecute the action (as would be the case with customary class certification procedures), courts, it is said, need only determine, in hindsight, whether the settlement was negotiated at arms’ length, and whether the negotiations were long, thorough and deliberative. See In re Corrugated Container Antitrust Litig., 643 F.2d 195, 212 (5th Cir.1981) (adequacy judged by sufficiency of settlement); In re Domestic Air Transp. Antitrust Litig., 148 F.R.D. 297, 341 (N.D.Ga.1993) (inequitable distribution). Courts adopting this approach require proof only that named plaintiffs’ and class interests are not antagonistic. See, e.g., Goodman v. Lukens Steel Co., 777 F.2d 113, 123 (3d Cir.1985) (relying on absence of conflict to find adequate representation); Lewis v. Curtis, 671 F.2d 779, 788 (3d Cir.1982) (finding named plaintiff an adequate representative despite small stake in litigation and ignorance of facts and claims); Steiner v. Equimark Corp., 96 F.R.D. 603, 610 (W.D.Pa.1983) (“The key question [for the adequacy of representation inquiry] is whether their interests are antagonistic.”). In these cases, courts have effectively relied on the settlement’s terms — the outcome of the action — to find the required absence of antagonism.20
*799We disagree with this approach, championed primarily by Newberg. There is no language in the rule that can be read to authorize separate, liberalized criteria for settlement classes.21 Although we acknowledge the need for flexible interpretation of Rule 23 to enable it to achieve its broader purposes of vindicating difficult individual claims and conserving judicial resources, see Beef, 607 F.2d at 177-78 (discussing the policy needs for flexibility); Ace Heating, 453 F.2d at 33 (recognizing need to give small claimants who did not opt out the right to appeal a settlement approval), we emphasize that Rule 23 is designed to assure that courts will identify the common interests of class members and evaluate the named plaintiffs and counsel’s ability to fairly and adequately protect class interests. See Katz v. Carte Blanche Corp., 496 F.2d 747, 757 (3d Cir.1974). Thus, actions certified as settlement classes must meet the same requirements under Rule 23 as litigation classes. To allow lower standards for the requisites of the rule in the face of the hydraulic pressures confronted by courts adjudicating very large and complex actions would erode the protection afforded by the rule almost entirely.
Judge Posner has explained the animating concern behind this strict application. “The danger of a premature, even a collusive, settlement is increased when as in this case the status of the action as a class action is not determined until a settlement has been negotiated, with all the momentum that a settlement agreement generates....” Mars, 834 F.2d at 680.
The foregoing discussion has focused on adequacy of representation,.but the presence of commonality and typicality are equally important'to the class action regime. Certifying a class without the existence of questions common to the class (or where the class representatives’ claims are not typical) perverts the class action process and converts a federal court into a mediation forum for cases that belong elsewhere, usually in state court. On the other hand, the cases that make the settlement class device appear most useful are eases presenting the most unwieldy substantive and procedural issues, i.e., those diversity cases in which plaintiffs from many states are confronted with differing defenses, differing statutes of limitations, etc. — precisely those cases that stretch the Rule to its outer-most limits.
This is a troublesome issue — and a close one. Many mass tort actions have this problem. The School Asbestos cases and the Breast Implant eases had it, and this case does, as well. It may initially seem difficult to envision an actual trial of these cases because of the differing defenses certain to be raised under the various bodies of governing law. While the problem may be overstated,22 settlement classes still serve the useful purpose of ridding the courts — state and federal — of this albatross even though the case may never have been triable in class form. But if that were- the primary function of the settlement class, the federal courts would have become a mediation forum, a result inconsistent with their mission and limited resources. In sum, “a class is a class is a class,” and a settlement class, if it is to qualify under Rule 23, must meet all of its *800requirements. The district court should keep these matters in mind on remand.
V. Is the Settlement Class PropeR Here?
A. Were There Adequate Findings Under Rule 23(a)?
Certain of the objectors in this ease contend that the district court committed plain error by never actually certifying the class as required by Rule 23. See Brief of French Objectors, at 18. This, of course, would be a serious error, since without certification there is no class action, and “[i]n a settlement entered without class certification the judgment will not have res judicata effect on the claims of absent class members.” Simer v. Rios, 661 F.2d 655, 664 (7th Cir.1981).
The district court certified the class provisionally in a pretrial order. See Pretrial Order No. 7. We have already noted that provisional certification constitutes an acceptable means of facilitating settlement-negotiations. See 2 Newberg & Conte § 11.27, at 55-56. It appears that the court believed that it certified the class by “confirming” the provisional certification in its order approving the settlement. However, the court did not make the findings we today hold that Rule 23 requires, not even upon approving the settlement. Because we hold that courts employing settlement classes must still make findings that the class complies with Rule 23(a) and the appropriate parts of Rule 23(b), the court’s failure to comply with the rule in this respect is a plain error of law, and hence an abuse of discretion, requiring that the certification be set aside.
Our conclusion that the settlement class was not properly certified does not mean that the class could not be certified on remand. Accordingly, we must consider whether the existing record is adequate to support class certification, or whether further record development is required.
B. Could the Class Requisites Have Been Met on the Current Record?
1. Numerosity, Commonality, and Typicality
As we have explained, a class action — whether certified for settlement or litigation purposes — must meet the class requisites enunciated in Rule 23. The district court did not make findings on these issues. The numerosity requirement of Rule 23(a) is plainly satisfied in this action encompassing nearly six million truck owners. The commonality and typicality inquiries of 23(a), however, raise substantial concerns about the sufficiency of this class. The record currently lacks the facts needed to establish these requisites, and the defendants also ardently maintain that the applicability of different defenses to different groups of plaintiffs would prevent the class from satisfying the commonality and typicality requirements. At this juncture, we leave open the possibility that, on remand, the district court may indeed find facts sufficient to support these elements.
2. Adequacy of Representation
a. The Situation of the Fleet Owners
This settlement class appears to fail to meet Rule 23(a)’s adequacy of representation test. The adequacy of representation inquiry has two components intended to assure that the absentees’ interests are fully pursued: it considers whether the named plaintiffs’ interests are sufficiently aligned with the absentees, and it tests the qualifications of the counsel to represent the class. See Weiss v. York Hospital, 745 F.2d 786, 811 (3d Cir.1984); 2 Newberg & Conte § 11.28, at 11-58. On the first prong, we are not satisfied that the interests of various class members were sufficiently aligned; indeed the settlement appears to create antagonism within the class. While some courts have been satisfied that there is no intraclass conflict where “all class members are united in asserting a common right, such as achieving the maximum possible recovery for the class,” In re Corrugated Container Antitrust Litig., 1980-1 Trade Cas. (CCH) ¶ 63,-163 at 77,788 n. 10 (S.D.Tex.1979), aff'd, 643 F.2d 195 (5th Cir.1981), we disapprove such a myopic focus on the settlement terms.
In this ease in particular, the conclusion that the settlement — which (supposedly) maximized class recovery — satisfied the requirement that class members’ interests not *801be antagonistic ignores the conspicuous evidence of such an intra-class conflict in the very terms of this settlement. The substantial impediments to fleet owners’ using these certificates creates a conflict between their interests in this settlement and those of individual owners. (The named plaintiffs are all individual owners.) Moreover, the dubious value of the transfer option, see Part VLA.1.C infra, one of the principal responses to the fleet owners’ objection, does little to reduce the disparity in the prospective value to the different sections of the class.
This is not a case where some plaintiffs share the prospect of a future claim with other class members who currently have such a claim. The fleet owners will never enjoy the benefits of the settlement terms, such as the intra-household transfer option, intended specifically for the benefit of individual owners. Thus, we must be concerned that the individual owners had no incentive to maximize the recovery of the government entities; they could skew the terms of the settlement to their own benefit. Not surprisingly, the settlement leaves fleet owners with significantly less value than individual owners. At the very least, the class should have been divided into sub-classes so that a court examining the settlement could consider settlement impacts that would be uniform at least within the sub-classes.
b. Did Counsel Adequately Represent ' the Interests of the Entire Class?
The other aspect of the adequacy of representation test, whether counsel are qualified and serve the interests of the entire class, also gives us reason to pause. Courts examining settlement classes have emphasized the special need to assure that class counsel: (1) possessed adequate experience; (2) vigorously prosecuted the action; and (3) acted at arm’s length from the defendant. See, e.g., Malchman, 706 F.2d at 483; Alvarado Partners, 723 F.Supp. at 546. The first criterion is no problem, for these counsel clearly possess the experience and skills to qualify them to pursue these sorts of actions. But the second and third points require attention in view of the lack of significant discovery and the the extremely expedited settlement of questionable value accompanied by an enormous legal fee.
Before addressing the latter points, it. is necessary to begin with some legal theory discussing the structural nature of fee arrangements in class actions of this type, having in mind that even honorable counsel— like class counsel here — may be compromised by the possibility of a large fee.
(1) Class Action Attorneys’ Fees Theory and Structure
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. See 2 NewbeRG & Conte § 11.65 at 11-183; Greenfield v. Villager Indus., Inc., 483 F.2d 824, 832 (3d.Cir.1973). The large fees garnered by some class lawyers can create the impression of an ethical violation since it may appear that the lawyer has an economic stake in their clients’ case. But class actions cannot be analyzed in the same framework as conventional bipolar litigation. Because of the collective action problems associated with cases where individual claims are relatively small, see 7A Chaeles A. WRIght, ARThur R. Miller & Mary Kay Kane, Federal Practice and Procedure § 1754, at 49, and the social desirability of many class suits (the private enforcement model), id. at 51; Sprogis v. United Air Lines, Inc. 444 F.2d 1194 (7th Cir.1971), large attorneys’ fees serve to motivate capable counsel to undertake these actions. Thus, large fee awards standing alone do not suffice to show that the representation was inadequate or unethical. These allowances generally reflect the realization that the lawyer represents numerous individuals with somewhat varying interests, not acceptance of a situation where the lawyer’s personal interests trump the interests of the entire class.
Some commentators blame the system of compensating class action lawyers in a manner that fails to confront fully the differences between class action litigation and classical bipolar litigation for creating incentives that diverge markedly and predictably from their clients’ interests. The leading critic is Professor Coffee. See John C. Coffee, Jr., Un*802derstanding the Plaintiffs Attorney: The Implications of Economic Theory For Private Enforcement of Law Through Class and Derivative Actions, 86 ColumL.Rev. 669, 671-72 (1986) (noting that critics “have argued that the legal rules governing the private attorney general have created misincen-tives that unnecessarily frustrate the utility of private enforcement. These critics have focused chiefly on the conflicts that arise between the interests of these attorneys and their clients in class and derivative actions ....”)([hereinafter] Understanding the Plaintiff’s Attorney); id. at 677 (“Ultimately, the most persuasive account of why class actions frequently produce unsatisfactory results is the hypothesis that such actions are uniquely vulnerable to collusive settlements that benefit plaintiffs attorneys rather than their clients.”); John C. Coffee, Rescuing the Private Attorney General: Why the Model of the Lawyer As Bounty Hunter Is Not Working, 42 MdL.Rev. 215 (1983); John C. Coffee, The Unfaithful Champion: The Plaintiff as Monitor in Shareholder Litigation 48 Law & Contemp. PRObs., 5 (1985); Kevin M. Clermont & John D. Currivan, Improving on the Contingent Fee, 63 Cornell L.Rev. 529 (1978); Murray L. Schwartz & Daniel J.B. Mitchell, An Economic Analysis of the Contingency Fee in Personal-Injury Litigation, 22 Stan-L.Rbv. 1125 (1970).
Economic models have shown how conventional methods of calculating class action fee awards give class counsel incentives to act earlier than their clients would deem optimal. See Coffee, Understanding the Plaintiffs Attorney, 86 ColumL.Rev. at 688. Because, under a percentage of recovery award mechanism, the attorney will only enjoy a relatively small portion of whatever incremental award he can extract from the defendant, the defendant can pressure the plaintiffs’ attorney into early settlement by threatening to expend large sums on dilatory tactics that would run the expenses up beyond what plaintiffs’ attorneys can expect to profit. Id. at 690. Rather than presenting a possible solution, the lodestar method seemingly exacerbates the problem of cheap settlement by divorcing the fee award from the settlement’s size, since plaintiffs’ attorneys have no incentive to take the risk on a trial for a potentially-larger award to the class where their own fees will not necessarily reflect the greater risk taken at trial. See also id. at 718 (discussing how lodestar method may create structural collusion).
Coffee also blames the principal-agent problem endemic to class actions for creating a situation where the defendants and plaintiffs can collusively settle litigation in a manner that is adverse to the class’s interest: “At its worst, the settlement process may amount to a covert exchange of a cheap settlement for a high award of attorney’s fees. Although courts have long recognized this danger and have developed some procedural safeguards intended to prevent collusive settlements, these reforms are far from adequate to the task.” Id. at 714 n. 121 (citing cases). A number of commentators have identified settlements that afford only nonpecuniary relief to the class as prime suspects of these cheap settlements. See Coffee, Understanding The Plaintiffs Attorney, 86 ColumL.Rev. at 716 n. 129; Jonathan R. Macey & Geoffrey P. Miller, The Plaintiffs’ Attorneys Role in Class Action and Derivative Litigation: Economic Analysis and Recommendations for Reform, 58 U.Chi.L.Rev. 1, 45 n. 10 (1991); Nancy Morawetz, Bargaining, Class Representar tion, and Fairness, 54 Ohio St.L.J. 1, 5 n. 40 (1993).
While courts may fail to appreciate adequately the distinction between conventional bipolar litigation and class actions in many respects, they may over-emphasize these differences in other respects. To be sure, courts will be willing to award fees in class actions that would appear extraordinary and arguably improper in conventional litigation. Nevertheless, some of the critiques based on ethical or collusive concerns remain instructive. Although subsequent versions seem to avoid a discussion, the Manual for Complex Litigation (First) acknowledged the potential for attorney-class conflict. It condemned fees that are paid separate and apart from the settlement funds paid to the class because amounts “paid by the defendants) are properly part of the settlement funds and should be known and disclosed at the time the fairness of the settlement is considered.” MCL 1st § 1.46.
*803One court has noted that the “effect of such an arrangement [where the counsel fees are not resolved and the details not included in the class notice] may be to cause counsel for the plaintiffs to be more interested in the amount to be paid as fees than in the amount to be paid to the plaintiffs.” In re General Motors Corp. Engine Interchange, 594 F.2d at 1131. Commentators have also noted how, where there is an absence of objectors, courts lack the independently-derived information about the merits to oppose proposed settlements. See Coffee, Understanding the Plaintiff’s Attorney, 86 Colum.L.Rev. at 714. n. 131. Of course, by endorsing a practice where the class is, for practical purposes, deprived of information concerning the fees, courts foster a situation where there will be fewer objectors.23
(2) The Stewardship of Counsel Here
A number of factors militate against the conclusion that the class’s interests were sufficiently pursued here. First, the settlement arguably did not maximize the class members’ interests. Every owner received a coupon whose value could only be realized by purchasing a new truck. Significant obstacles existed to the development of a secondary market in the transfer certificates given that the transfer restrictions and the certificates’ limited lifespan minimize the value of the transfer option. Second, class counsel effected a settlement that would yield very substantial rewards to them after what, in comparison to the $9.5 million fee, was little work.
Third, the fact that the settlement involves only non-cash relief, which is recognized as a prime indicator of suspect settlements, increases our sense that the class’s interests were not adequately vindicated. The separate negotiation of the fee agreement and the failure to disclose the amount of the award in the class notice only enhance this sense that counsel may have pursued a deal with the defendants separate from, and perhaps competing for the defendant’s resources with, the deal negotiated on behalf of the class. And although the degree to which a settlement hurts a defendant is not ordinarily a measure of the settlement’s adequacy, the fact that this settlement might actually benefit GM by motivating current owners to buy new trucks from the company (the settlement may arguably be viewed as a GM sales promotion device) certainly does little to allay the concern that the settlement did not advance the interests of the class as much as it might have.
Fourth, our concern about the vigor of counsels’ prosecution of the class claims, specifically the possibility that counsel did not do right by the class, is buttressed by the legacy of Prandini v. National Tea Co., 557 F.2d 1015, 1021 (3d Cir.1977). In Prandini, this court recognized the potential for attorney class conflicts where the fees, while ostensibly stemming from a separate agreement, were negotiated simultaneously. We characterized simultaneity of fee and settlement negotiations as a “situation ...' having, in practical effect, one fund divided between the attorney and client.” Id. To respond to this danger of collusion between the class counsel and defendant, Prandini and the Third Circuit Task Force Report on court awarded attorneys’ fees disapproved .fee discussions until after the achievement and approval of settlement. See Prandini, 557 F.2d at 1021; Court Awarded Attorney’s Fees, Report of the Third Circuit Task Force, 108 F.R.D. 238, 266 (1985) [hereinafter Task Force ].24
*804In this case, there were strong indications that such simultaneous negotiations in fact transpired. Indeed, there was evidence in a letter from class counsel that at least some portion of the fees and expenses had to have been negotiated simultaneously with the settlement. The court justified its dismissal of the allegation of simultaneous negotiation by citing (1) a statement in the letter that the “attorneys’ fees were negotiated separately, after we agreed on everything else,” and (2) GM’s reservation of the right to contest any award of fees that it deemed unreasonable. Even though we assume that these are factual findings, thus ordinarily deserving deference, we think these findings were made by reference to an erroneous legal standard. Indeed, neither of these bases is persuasive, especially in view of GM’s aequiecence in a patently baseless ground for augmenting the counsel fee, see Part VII infra.
In considering the adequacy of representation, we are loath to place such dispositive weight on the parties’ self-serving remarks. And even if counsel did not discuss fees until after they reached a settlement agreement, the statement would not allay our concern since the Task Force recommended that fee negotiations be postponed until the settlement was judicially approved, not merely until the date the parties allege to have reached an agreement. We recognize that Evans v. Jeff D., 475 U.S. 717, 734-38, 106 S.Ct. 1531, 1541-43, 89 L.Ed.2d 747 (1986), overruled Prandini’s strict rule prohibiting simultaneous negotiations. However, many of the concerns that motivated the Prandini rule remain, and we see no reason why Jeff D. or its underlying policy of avoiding rules that impede settlement preclude us from considering the timing of fee negotiations as a factor in our review of the adequacy of the class’s representation. Consequently, the likelihood that the parties did negotiate the fees concurrently with the settlement in this case increases our concern about the adequacy of representation.25
Nor would GM’s reservation of the right to appeal the fee award establish that the fee was negotiated separately since the likelihood that GM would want to contest an award based on a fee petition to which it agreed is quite small. The fact is confirmed by GM’s “lay down” position with respect to the fee application. Although the Supreme Court clearly invalidated the use of multipliers in lodestar awards in 1992, see City of Burlington v. Dague, — U.S. -, 112 S.Ct. 2638, 120 L.Ed.2d 449 (1992), GM did not apprise the district court of this fact when it was approving the fee award, or complain when the district court used a multiplier in the calculations. This posture of GM suggests that its reservation of the right to appeal the fee award should not be given great weight in determining whether the settlement and attorneys’ fee were negotiated separately. But we hasten to add that we have not resolved these factors. We only hold today that the court did not make the necessary findings, and we remand to the district court so that it can make the necessary Rule 23 findings.
The thrust of the foregoing discussion is that the circumstances under which the settlement evolved, made possible by the settlement class device, may have compromised class counsel in a manner raising doubts as to adequacy of representation: The district court will examine this aspect of the matter on remand. Perhaps, on a more developed record, the adequacy of representation will be established. These concerns underscore the importance of having the district court make Rule 23 findings. Although we do not believe that the class would meet the requirements for certification on the current record, we do not preclude the possibility that certification could be properly supported on a more developed record. Thus, we will remand this action to the district court so that it can re-examine the class certification and the settlement and, if appropriate, certify the class by making the findings required by Rules 23(a) and (b).
VI. Is the Settlement FaiR, Reasonable, and Adequate?
Invoking the correct standard of review under Girsh v. Jepson, 521 F.2d 153, *805157 (3d Cir.1975), the objectors also argue that the district court abused its discretion when it approved the settlement as fair, reasonable, and adequate. Because wé leave open the possibility that the district court may on remand properly certify the class pursuant to Part V of this opinion, we address the district court’s approval of the settlement. Rule 23(e) imposes on the trial judge the duty of protecting absentees, which is executed by the court’s assuring that the settlement represents adequate compensation for the release of the class claims. See 2 Newberg & Conte § 11.46, at 11-105 to 11-106. Some courts have described their duty under Rule 23(e) as the “fiduciary responsibility” of ensuring that the settlement is fair and not a product of collusion. In re Warner Commun. Secur. Litig., 798 F.2d 35, 37 (2d Cir.1986); see also Plummer v. Chemical Bank, 668 F.2d 654, 658 (2d Cir.1982); Grunin v. International House of Pancakes, 513 F.2d 114, 123 (8th Cir.), cert. denied, 423 U.S. 864, 96 S.Ct. 124, 46 L.Ed.2d 93 (1975); Alvarado Partners L.P. v. Mehta, 723 F.Supp. 540, 546 (D.Colo.1989). At all events, where the court fails' to comply with this duty, absentees have an action to enjoin the settlement. 2 NewbeRG & Conte § 11.23.
In order for the determination that the settlement is fair, reasonable, and adequate “to survive appellate review, the district court must show it has explored comprehensively all relevant factors.” Malchman, 706 F.2d at 434 (citing Protective Committee, 390 U.S. at 434, 88 S.Ct. at 1168; Plummer, 668 F.2d at 659). A number of courts have recognized the need for a special focus on precluding the existence of collusion. See Malchman, 706 F.2d at 433 (advocating a focus on the negotiation process to uncover possible collusion); General Motors Interchange, 594 F.2d at 1125 (finding a need for heightened scrutiny of the settlement stemming from the potential for collusive settlement).
The topic of class action settlement has received much attention, which is understandable given the growing frequency of the settlement of increasingly large claims through the class action device. See In re AH. Robins Co., 880 F.2d 709, 739-40 (4th Cir.1989) (discussing the use of the device to settle various mass tort cases); In re Taxable Municipal Bond Secur. Litig., 1994 WL 643142 at *5 (noting the dramatic change in attitudes of courts and commentators toward the settlement class). The drive to settle class actions has also grown, notwithstanding the potential for collusive settlements to compromise absentee interests. Courts undertaking the special role of supervising class action settlements are apparently heeding the public policy in favor of settlement, see 2 Newberg & Conte § 11.41, at 11-85, and acknowledging the urgency of this policy in complex actions that consume substantial judicial resources and present unusually large risks for the litigants.
We have already noted the special difficulties the court encounters with its duties under Rule 23(e) in approving settlements where negotiations occur before the court has certified the class. Because of such difficulties, many courts have required the parties to make a higher showing of fairness to sustain these settlements. See, e.g., Ace Heating & Plumbing Co. v. Crane Co., 453 F.2d 30, 33 (3d Cir.1971) (“[Wjhen the settlement is not négotiated by a court designated class representative the court must be doubly careful in evaluating the fairness of the settlement to the plaintiffs class.”); General Motors Interchange, 594 F.2d at 1125 (attributing a need for heightened scrutiny of the settlement to the potential for collusive settlement); Weinberger v. Kendrick, 698 F.2d 61, 73 (2d Cir.1982) (higher showing of fairness required in pre-certification settlements, and special focus on assuring adequate representation and the absence of collusion); Malchman v. Davis, 706 F.2d 426, 434 (2d Cir.1983); Mars Steel v. Continental Ill. Nat’l Bank & Trust, 834 F.2d 677, 681 (7th Cir.1987); County of Suffolk v. Long Island Lighting Co., 907 F.2d 1295, 1323 (2d Cir.1990); 2 Newberg & Conte § 11.23; MCL 2d § 30.42 (citing the informational deficiencies faced by the court and counsel in pre-certifi-cation settlements). We affirm the need for courts to be even more scrupulous than usual in approving settlements where no class has yet been formally certified.
Settlements that have survived this heightened standard have involved much stronger *806indications of sustained advocacy by the de facto class counsel than we observe in this case. See Weinberger, 698 F.2d 61 (settlement discussions did not commence until after four years of discovery supplemented by another investigation by a trustee and after plaintiffs rejected first settlement offer); In re Beef Indus. Antitrust Litig., 607 F.2d at 177-78 (settlement discussions began after six months of discovery; action pending for three years, court fully briefed); City of Detroit v. Grinnell, 495 F.2d 448, 464 (2d Cir.1974) (approving settlement after several counsel vied for position for four years and voiced strenuous objections, explaining that Manual’s concerns about settlement classes articulated by the Manual for Complex Litigation only pertained to settlement in the early stages of litigation); cf Plummer, 668 F.2d 654 (rejecting settlement where plaintiffs’ counsel relied on information voluntarily furnished by defendants).
There are certain basic questions that-courts can ask to detect those cases settled in the absence of sustained effort by class representatives sufficient to protect the interests of the absentees. See MCL 2d § 30.41. For instance: Is the relief afforded by the settlement significantly less than what appears appropriate in light of the preliminary discovery? Have major causes of action or types of relief sought in the complaint been omitted by the settlement? Did the parties achieve the settlement after little or no discovery? Does it appear that the parties negotiated simultaneously on attorneys’ fees and class relief? Even acknowledging the possibility of some overpleading, these- questions raise a red flag in this case.
With the courts’ heightened duty to scrutinize this precertification settlement and some of these rudimentary indicators in mind, we now apply our nine-factor Girsh test, see Part III supra, and conclude from the balance of these factors that the district court’s conclusion that the settlement was fair and reasonable constitutes an abuse of discretion. Coincidentally, this result tracks the conclusions of a Texas appeals court that, based on an analysis similar to that of Girsh, set aside an order approving a substantially identical settlement of similar claims brought by residents of Texas. See Bloyed v. General Motors, 881 S.W.2d at 422.
A. Adequacy of Settlement— General Principles
This inquiry measures the value of the settlement itself to determine whether the decision to settle represents a good value for a relatively weak case or a sell-out of an otherwise strong case. The Girsh test calls upon courts to make this evaluation from two slightly different vantage points. According to Girsh, courts approving settlements should determine a range of reasonable settlements in light of the best possible recovery (the eighth Girsh factor) and a range in light of all the attendant risks of litigation (the ninth factor). See Girsh v. Jepson, 521 F.2d at 157; see also Malchman v. Davis, 706 F.2d 426, 433 (2d Cir.1983) (identifying a similar test); City of Detroit v. Grinnell Corp., 495 F.2d 448, 463 (2d Cir.1974) (same).
In formulaic terms we agree that “in cases primarily seeking monetary relief, the present value of the damages plaintiffs would likely recover if successful, appropriately discounted for the risk of not prevailing, should be compared with the amount of the proposed settlement.” MCL 2d § 30.44, at 252. This figure should generate a range of reasonableness (based on size of the proposed award and the uncertainty inherent in these estimates) within which a district court approving (or rejecting) a settlement will not be set aside. See Newman v. Stein, 464 F.2d 689, 693 (2d Cir.1972). The evaluating court must, of course, guard against demanding too large a settlement based on its view of the merits of the litigation; after all, settlement is a compromise, a yielding of the highest hopes in exchange for certainty and resolution. See Cotton v. Hinton, 559 F.2d 1326, 1330 (5th Cir.1977). The primary touchstone of this inquiry is the economic valuation of the proposed settlement.
We turn to this analysis. As will appear, the district court’s conclusion that the settlement was within the range of reasonableness rests heavily on the proposition that the class had never proven any diminution in value of the trucks. It ignored the fact that the coupons provided no cash value *807and made no provision for repairing the allegedly life-threatening defect. For the reasons that follow, we believe that the district court did not sufficiently scrutinize the valuations of the settlement, and that, on this record, the settlement appears to be inadequate. Consequently, we will conclude that the district court erred when it found that the settlement fell within the range of reasonableness.
1. Valuation of the Settlement— Introduction
The value of the $1,000 certificates is sharply disputed. GM argues that the certificates are worth close to their face value since they can be redeemed for a broad array of GM trucks and can be used in combination with dealer incentives. For those unable or unwilling to purchase another GM truck, GM argues, cash can be realized from transferring the certificate within the household for full value or selling the certificate for $500. Plaintiffs presented an expert, Dr. Itamar Simonson, who placed the value of the certificates between $1.98 and $2.18 billion, based on an estimate that 34% to 38% of the class would redeem the certificate in purchasing a new truck and an additional 11% of the class would sell their certificates for $500. Objectors contest these estimates and many of the assumptions used to generate them.
We therefore analyze several of the foundations for the district court’s evaluation. First, we inquire about the reliability of plaintiffs’ witness’s valuation. Second, we explore the adequacy of the district court’s consideration of the possibility that some class members would not be able to use the coupons at all. Third, we inquire as to whether the quite significant restrictions on transfer of the certificates present obstacles to the development of a market so as to render the estimates of their worth unreasonably inflated. Finally, we consider whether the size of the attorneys’ fees agreement suggests that GM attached a greater value to the class claims than proponents of the settlement would have us believe. These factors lead ineluctably to the conclusion that the district court overvalued this settlement, which in turn gives credence to the contention of the objectors that the proffered settlement was, in reality, a sophisticated GM marketing program.
a. Plaintiffs’ Witness Dr. Itmar Simonsen
Dr. Simonsen’s methods and assumptions raise serious doubts about the reliability of the valuations they generated. Although Si-monsen's conclusion was based on his estimate that between 34% and 38% of the class members would use the certificate, his own telephone survey revealed that only 14% of the class reported that they would “definitely” or “probably” buy a new truck. Apparently Simonsen only excluded those who responded that they would “definitely not buy” or “probably not buy” a new truck, a methodological choice which is questionable. Furthermore, Simonsen discounted the statistics by seemingly arbitrary factors in an effort to be “conservative,” but without some basis or explanation for the derivation of those factors, we have no way of judging whether they were conservative or aggressive.
Even more importantly, the raw survey data probably overstate the prospects that the certificates will be used since there are substantial obstacles to obtaining and transferring the certificates, none of which Simon-sen deals with. Finally, Simonsen supposed that a higher percentage of fleet owners than individual owners would redeem the certificates, but this seems to disregard the statutory and regulatory constraints that often restrict fleet buyers’ purchase decisions. Indubitably all of these concerns reduce the value of the settlement, yet Simonsen appears simply to have multiplied his estimated number of users by the coupon amount or transfer value.
On the other hand, although various objectors have made a good argument that the net value of the certificates will also be eroded by rising truck prices (which would allegedly be influenced both by the huge number of certificates that would need to be redeemed within a relatively brief time and by the fact that dealers may take advantage of customers they know to be somewhat tied to the purchase of a GM truck by their desire to realize value from the coupon), we will, to be conservative, not take this factor into ac*808count. Even so, Simonsen’s methodology undermines his conclusion to the extent that his valuation cannot support the settling parties’ case.
b. Inability of Class Members to Use Certificates
The district court also erred by not adequately accounting for the different abilities (not inclinations) of class members to use the settlement. One sign that a settlement may not be fair is that some segments of the class are treated differently from others. See Piambino v. Bailey, 610 F.2d at 1329; In re GM Corp. Engine Interchange Litig., 594 F.2d at 1128; MCL 2d § 30.41, at 236. Consequently, the fact that the coupon settlement benefits certain groups of the class more than others suggests that the district court did not adequately discharge its duties to safeguard the interests of the absentees. See In re Fine Paper Antitrust Litig., 617 F.2d 22 (3d Cir.1980) (ongoing duty of the judge to protect absentees); Piambino v. Bailey, 610 F.2d at 1329 (duty to assure the settlement is fair, reasonable, and adequate with respect to each category of the class).
People of lesser financial means will be unable to benefit comparably from the settlement. GM cites a number of other judicially approved class action settlements that awarded coupons and argues that, since this coupon provides far more consideration, it necessarily merits approval. See, e.g., New York v. Nintendo of Am. Inc., 775 F.Supp. 676, 679 (S.D.N.Y.1991) ($5 discount coupon for video game purchase of approximately $200); In re Cuisinart Food Processor Antitrust Litig., 1983-2 Trade Cas. (CCH) ¶ 65, at 680, 1983 WL 153 (D.Conn.1983) (discount coupons with maximum value of $100 for machines costing approximately $100 to $300); In re Domestic Air Transp. Antitrust Litig., 148 F.R.D. 297, 331 (N.D.Ga.1993) (certificates worth between $10 and $200 for flights costing between $50 and $1500).
These cases, however, differ dramatically in the amount of money required to purchase the good — i.e. to realize the certificate’s value — and in the frequency with which a typical consumer might expect to purchase the good. Whether a new truck costs between $20,000 and $33,000 as some objectors claim or some amount “far less” than that, as GM claims, this purchase is not comparable to buying a new food processor or even an airline ticket. As the district court acknowledged, “a substantial number of class members” would not be able to afford a new truck within the fifteen month coupon period. Both the high cost of the trucks and the infrequency of a consumer’s purchase of a new truck (relative to the fifteen month redemption period) make using these certificates significantly more difficult than those in the other coupon settlements, for all class members but particularly for the poorer ones.
Even where class members do manage to use the certificates, we are concerned about their real value. It may not be the case that the certificates saved those class members $1,000 on something they would have otherwise bought; those class members may only have purchased new GM trucks because they felt beholden to use the certificates. Thus, rather than providing substantial value to the class, the certificate settlement might be little more than a sales promotion for GM, in just the way that the Bloyed court characterized the settlement as a “tremendous sales bonanza” for GM. Bloyed v. General Motors Corp., 881 S.W.2d at 431.
We turn then to the fleet buyers, who constitute a readily identifiable category of plaintiffs arguably disadvantaged by the settlement. Budgetary constraints prevent some of them from replacing their entire fleets within the fifteen month redemption period.26 Competitive bidding requirements also apparently impede many of these entities from being able to use the certificates. Because there is no assurance that GM will be the lowest bidder, the government entities bound by these requirements may not be able to use the certificates. See, e.g. The *809Louisiana Public Bid Law, LaJR.Stat. § 38:2212(A)(l)(a). The district court dismissed these objections saying it was “confident that ingenious counsel will be able to structure bidding requirements so that the governmental entities can take full advantage of the certificates.” The district court’s observation, while perhaps partially accurate, represents far too cavalier a dismissal of a potentially serious intra-class conflict and inequity.
The named plaintiffs argue that, if certain fleet buyers and individuals were dissatisfied with the settlement’s terms, they could simply opt out of the class and pursue their own relief individually. While such an argument might theoretically be true, it ignores the realities of pursuing small claims. It would cost considerably more to litigate individual claims than the litigant could recover, using either a retrofit or a warranty theory to measure damages. And the district court apparently did not consider the possibility of a subclass of fleet owners, though that might alter the anatomy of the settlement. At all events, the right of parties to opt out does not relieve the court of its duty to safeguard the interests of the class and to withhold approval from any settlement that creates conflicts among the class. In sum, the relative inability of class members to use the certificates militates against settlement approval.
c. Value of the Transfer Option
In order to support its conclusion that the settlement was reasonable and fair, the district court cited the ability of fleet buyers and those consumers with budget constraints to realize value from the certificates by transferring them. We believe that the value of the transfer option is dubious, and consequently that the settlement was unfair to substantial portions of the class.
Simonsen’s valuation of the settlement includes $157 million attributable to transferred certificates. Simonsen calculated that holders of the certificates could realize $250 from the sale of the transferred certificates (with a $500 face value). He gave no explanation for his assumption of a $250 market value. To the extent that this methodology is also dubious, it compounds the skewing of the valuation wrought by his usage estimates, see Part VI(A)(l)(a) supra.
The value of this option depends on the development of a secondary market for these certificates. But there is no assurance that a market will develop; indeed, the restrictions on transfer, which GM claims are necessary to prevent fraud, pose significant barriers to the creation of such a market. The requirement that holders send in their $1,000 or original certificate to exchange for the $500 transfer certificate imposes very significant transaction costs since the parties must agree on a price before the original holder initiates the transfer process (which could easily last several weeks). During that process, there is substantial market risk, for the price of the transfer certificate could well move dramatically and induce a breach in the purchase agreement by one of the parties. Breaches would pose a real problem in this case because the transfer certificate cannot be reissued in another’s name and thus cannot be resold. Because of these risks, individuals will be quite reluctant to contract for these transfer certificates. Even worse, the one-time transfer restriction also precludes the development of a market-making clearing house mechanism. In our view, therefore, it-is quite possible that holders will be unable to realize any significant value from the' transfer option.
Aside from the effect of the transfer restrictions, we also question Simonsen’s valuation' on the basis that it did not account for the inability to use the transfer certificates in conjunction with other incentive plans. For example, the incremental value of the $500 transfer certificate to class members would be completely eroded if GM offers a $1,000 dealer rebate program, since the class member would be forced to choose between the plans and would therefore be no better off than the general public.
The district court did not take cognizance of these factors. It erred when it presumed development of a liquid market for these transfer certificates with very little support in the record for it, and when it relied on a putative value of the transfer option arbitrarily ascribed by plaintiffs’ expert to find *810that the settlement was fair and reasonable. Although objectors might have made out an even stronger case by proffering their own expert on this valuation, the court has an independent duty to scrutinize the settlement’s value and any evidence offered to support it. Accordingly, we find that evidence pertaining to the incremental value created by the transfer option does not support the valuation of the settlement.
d. GM’s Implicit Valuation of the Claim
Our concerns about the adequacy of the settlement are complicated by the generous attorneys’ fees GM agreed to pay in this case. Although originally GM vigorously contested the viability of the class claims and the class, the company, in view of its willingness to pay attorneys’ fees of $9.5 million, may, at the time of settlement, have valued the claims at some substantial multiple of the fee award.27 This $9.5 million attorney’s fee award seems unusually large in light of the fact that the settlement itself offered no cash outlay to the class. GM’s apparent willingness to pay plaintiffs’ counsel close to $9.5 million indicates that the party in perhaps the best position to evaluate the claim may have thought the action, which both plaintiffs’ counsel and the defense contend was not worth much, posed a significant enough threat to cause GM to strike a lucrative deal with plaintiffs counsel.
On the other hand, perhaps GM’s valuation results only from the class counsel’s decision to settle the action at an early stage and GM’s desire to encourage that decision. Of course, a decision to settle that occurs at too incipient a stage of the proceedings also weighs against settlement approval. In short, while the settlement certainly presented difficult valuation issues, we believe that the district court erred when it uncritically accepted such high estimates of the settlement’s value.
2. Valuing This Settlement Relative to the Relief Requested
The ninth Girsh factor also undermines the district court’s decision. In the class action context, “the relief sought in the complaint” serves as a useful benchmark in deciding the reasonableness of a settlement. See Cotton v. Hinton, 559 F.2d 1326, 1330 (5th Cir.1977). Here the adequacy of the certificate settlement is particularly dubious in light of the claims alleged and the relief requested in the original complaint. The coupons offered by GM simply do not address the safety defect that formed the central basis of the amended complaint filed barely four months before the settlement.28 The district court gave two justifications for its conclusion that, notwithstanding this discrepancy, the settlement was fair. First, the court explained, “no objector that complains that the settlement fails to retrofit the alleged defect has been able to come forth with a practical and safe modification for the trucks that has been designed, evaluated and tested.” Second, the court also relied on the fact that “[t]he proposed class settlement does not affect the rights of settlement class members to participate in any recall that NHTSA orders.”
Having considered the validity of these arguments, we conclude that they do not alleviate the substantial concerns created by the dramatic divergence of the settlement terms from the relief originally sought. This factor, therefore, strengthens our conviction that the settlement was not fair, reasonable, or adequate.
a. The Retrofit Issue
It is true that there does not appear to be a consensus retrofit. For each of the sug*811gested retrofits — relocation of the gas tank to the spare tire location, installation of a tank with a rubber bladder, or installation of a metal cage around the gas tank — there was evidence that the retrofit either was ineffective or caused other performance problems for at least some model years. On the other hand, there was also evidence supporting the efficacy of various retrofits. For instance, GM’s own documents considered all three options and found that all would enhance the safety of the fuel systems. In addition, there was also potentially damaging testimony by Ronald E. Elwell, an engineering analyst at GM for fifteen of his twenty-eight years with the company and its chief expert in defending the fuel tank location and design on the full-size trucks in a number of significant product liability cases, see, e.g., Bowman v. General Motors, 427 F.Supp. 234, 236 (E.D.Pa.1977).
In his deposition, Elwell testified that GM designed a retrofit using a steel cage which prevented the gas tanks from rupturing in side impact testing. He further testified that GM abandoned the retrofit (knowing, because of its own secret crash tests, of the increased fire danger) only because GM feared that it would give the public the wrong impression. GM attempted to impeach Elwell by characterizing him as a “disgruntled,” (former) employee. By way of rehabilitation, objectors explain Ellwell’s reduced duties as a result of health problems. Whether or not Ellwell’s testimony could itself establish that the steel cage enhances safety, his testimony might have been important if the case had proceeded to trial. As a consequence, the district court abused its discretion when it summarily dismissed Elwell’s testimony.
b. Availability of Other Remedies
The district court also relied on the existence of the NHTSA recall mechanism and the class members’ unencumbered right to bring personal injury suits to justify its approval of a settlement that did not secure any of the equitable relief originally requested. While individual tort suits are not barred, the court’s approval of this settlement (which does nothing to redress the alleged danger) foregoes the opportunity presented by the pleadings29 to prevent injuries that tort suits can at best address only retrospectively. More importantly, the NHTSA remedy may be extremely limited in that it can only require a manufacturer to repair a vehicle first purchased within eight calendar years of the investigation. The court’s observation that “all the plaintiffs may have statute of limitations problems in this action that may be equally as severe or worse than the eight year NHTSA limitation” does nothing to increase the value of the theoretical access to a NHTSA recall remedy to the owner or others who may be injured by the trucks at some future point. Hence, the potential existence of a partial recall under NHTSA does not dispel our doubts about the settlement terms’ diverging so far from the original complaint. In so concluding, we do not rely on the subsequent resolution of the NHTSA investigation, which did not include any recall.
In sum, we agree with the district court that the evidence of the existence of an effective retrofit to be contradictory; nevertheless, we think that the very murkiness of this evidence and the fact that certain key evidence was wrongly excluded, especially in light of the magnitude of the alleged safety defect, militate against approving a settlement attained at such an early stage of the litigation which does nothing to repair the vehicles, not even creating a fund to finance retrofits.30
*812 B. Complexity of the Suit
This factor is intended to capture “the probable costs, in both time and money, of continued litigation.” Bryan v. Pittsburgh Plate Glass Co., 494 F.2d 799, 801 (3d Cir.), cert. denied, 419 U.S. 900, 95 S.Ct. 184, 42 L.Ed.2d 146 (1974). By measuring the costs of continuing on the adversarial path, a court can gauge the benefit of settling the claim amicably. The district court here concluded that the litigation “would be mammoth” and would have resulted in a “substantial delay in ... recovery.”
While it is true, as the Youngs objectors argue, that the district court’s conclusion in part depended on the ambitious definition of the class in terms of both geography and truck models included, the action would still involve a complex web of state and federal warranty, tort, and consumer protection claims even if the class had been subdivided and some of the legal issues simplified. Had the ease not been settled, both plaintiffs and GM would have had to conduct discovery into the background of the six million vehicles owned by class members, including any representations allegedly made to plaintiffs. Each side would also have needed to hire or produce a retinue of experts to testify on a variety of complex issues. Undoubtedly, GM would have ardently contested the action at every step, leading to a plethora of pretrial motions. In contrast, this settlement made its remedies immediately available and avoided the substantial delay and expense that would have accompanied the pursuit of this litigation. The district court thus correctly concluded that the complexity factor weighed in favor of approving the settlement.
C. Reaction of the Class
In an effort to measure the class’s own reaction to the settlement’s terms directly, courts look to the number and vociferousness of the objectors. Courts have generally assumed that “silence constitutes tacit consent to the agreement.” Bell Atlantic Corp. v. Bolger, 2 F.3d 1304, 1313 n. 15 (3d Cir.1993). However, a combination of observations about the practical realities of class actions has led a number of courts to be considerably more cautious about inferring support from a small number of objectors to a sophisticated settlement. See, e.g., In re Corrugated Container Antitrust Litig., 643 F.2d at 217-18; GM Interchange Litig., 594 F.2d at 1137.
In a class action case involving securities litigation, this court has recognized the possibility that the assumption that silence constitutes tacit consent “understates potential objectors since many shareholders have small holdings or diversified portfolios, ... and thus have an insufficient incentive to contest an unpalatable settlement agreement because the cost of contesting exceeds the objector’s pro rata benefit.” Bell Atlantic Corp. v. Bolger, 2 F.3d at 1313 n. 15. Although this is not a securities class action and the amounts at stake could be significant, the absentees may not fully appreciate the size of their potential claims since, by excluding those owners whose trucks have already experienced some mishap related to the fuel tank design, the class may include only those who have no reason (outside of media coverage) to know of the latent defect or the claim based on the alleged existence of that defect.
Even where there are no incentives or informational barriers to class opposition, the inference of approval drawn from silence may be unwarranted. As we noted earlier, Judge Posner has explained that “where notice of the class action is ... sent simulta*813neously with the notice of the settlement itself, the class members are presented with what looks like a fait accompli.” Mars Steel, 834 F.2d at 681. In this case especially, the combined notice largely defeats the potential for objection since the notice did not inform the class that the original complaint had sought a retrofit.31 Without information about the original complaint, absentees lacked any basis for comparing the settlement offered to them to the original prayer. It is instructive that many of the better-informed absentees, the fleet owners, did object.
The fact that a poll conducted by class counsel’s marketing expert reported that a minimum of sixty-three percent of the class would probably or definitely not use the coupon to purchase a new truck also suggests that the class could not possibly have so wholeheartedly endorsed the settlement. Moreover, one cannot infer approval of the settlement from requests for the transfer of the certificates, as the district court did. Those requests only signify that certain class members attempted to maximize the value they could realize from the settlement with which they were presented and thus might illustrate how futile class members thought objecting would be.
Although the absolute number of objectors was relatively low,32 there are other indications that the class reaction to the suit was quite negative: The seemingly low number of objectors includes some fleet owners who each own as many as 1,000 trucks, and those who did object did so quite vociferously. In conjunction with the already-noted problems associated with assuming that the class members possessed adequate interest and information to voice objections, the appeals of those who actually objected demonstrate that the reaction of the class was actually negative, and not supported by the “vast majority of the class members” as the district court concluded. The class reaction factor plainly does not, contrary to the district court’s conclusion, weigh in favor of approving the settlement.
D. Stage of Proceedings
The stage-of-proceedings facet of the Girsh test captures the degree of case development that class counsel have accomplished prior to settlement. Through this lens, courts can determine whether counsel had an adequate appreciation of the merits of the case before negotiating. The district court found that this factor favored settlement approval, relying on the fact that settlement was presented for approval less than six months prior to the scheduled trial date.
Given the purpose of this inquiry, however, it is more appropriate to measure the stage by reference to the commencement of proceedings either in the class action at issue or in some related proceeding. See In re Beef Antitrust, 607 F.2d at 180 (court referred to discovery in companion cases); City of Detroit v. Grinnell Corp., 356 F.Supp. 1380, 1386 (S.D.N.Y.1972), rev’d on other grounds, 495 F.2d 448 (2d Cir.1974) (noting the extensive discovery in that and parallel cases); In re Baldwin-United Corp., 105 F.R.D. at 483 (access to expert testimony and other evidence from parallel state court proceedings as well as to relevant public documents led court to believe counsel “availed themselves of all of these sources of information and conducted full adversarial negotiations ... ”); 2 NewbeRG & Conte § 11.45, at 11-102 n. 247.
The relevant period of time this case was in litigation was quite brief; approximately four months elapsed from the filing of the consolidated complaint to the reaching of the settlement agreement. To be sure, we cannot measure the extent of counsel’s effort from the time of the litigation alone; class counsel in this case are known to be quite industrious, and the district court properly considered class counsel’s review of the ma*814terials from prior product liability proceedings and from the Moseley personal injury case, Mosley v. General Motors Corp., No. 90-V-6276 (Fulton County, Ga. Feb. 4, 1993). However, mere access to the materials from other proceedings does not establish that counsel developed the merits, particularly where the other cases were premised on different theories of recovery. While we have no doubt that class counsel diligently reviewed those materials during the relevant period, nothing in the record demonstrates that they had conducted significant independent discovery or investigations to develop the merits of their case (as opposed to supporting the value of the settlement), that they had retained their own experts, or that they had deposed a significant number of the individuals implicated in the materials from these other proceedings. It is particularly noteworthy that the plaintiffs did not depose Ronald Elwell, although he could potentially have offered evidence that would have substantially bolstered the plaintiffs’ case. See Part VI.A.2.a supra.
At all events, the inchoate stage of case development reduces our confidence that the proceedings had advanced to the point that counsel could fairly, safely, and appropriately decide to settle the action. While the district court may have, laudably, been attempting to minimize the funds expended on discovery in order to maximize the funds available to the class, we think that the district court erred by not assuring that adequate discovery had been taken.
Beyond the incipient stage of the ease and the modest indications of substantive development, there is little basis for presuming vigorous prosecution of the case from the fact that settlement negotiations occurred. In ordinary class action settlements (i.e., where the court certifies the class before settlement negotiations commence) courts can presume that the negotiations occurred at arm’s length because they have already determined that the counsel negotiating on behalf of the class adequately represents the class’s interests. See Part IV.E supra, (discussion of adequacy of representation). In cases such as this one, however, where there has been no determination by the court that a proper class exists, the mere fact that negotiations transpired does not tend to prove that the class’s interests were pursued. Id. In short, the incipient stage of the proceedings poses an even larger obstacle to settlement approval in settlement class situations than it would in normal class action settlements since courts have no other basis on which to conclude that counsel adequately developed the claims before deciding to settle.
Furthermore, to the extent that this stage-of-proceedings factor also aims to assure that courts have enough exposure to the merits of the case to enable them to make these evaluations, it cannot support settlement approval here. With little adversarial briefing on either class status or the substantive legal claims, the district court had virtually nothing to aid its evaluation of the settlement terms. We therefore conclude that the district court clearly erred in finding that this factor weighed in favor of settlement approval.
E. Risks of Establishing Liability
By evaluating the risks of establishing liability, the district court can examine what the potential rewards (or downside) of litigation might have been had class counsel elected to litigate the claims rather than settle them. See In re Baldwin-United Corp., 105 F.R.D. 475, 482 (S.D.N.Y.1985) (comparing advantages of an immediate cash payment with risks involved in long and uncertain litigation). The district court here concluded that this factor also weighed in favor of approving the settlement since “there appear[ed] to be a substantial risk in establishing liability because of the complexity and size of the case along with the legal and factual problems raised by GM.” While we agree with the district court that, on balance, the prospective difficulty faced by a nationwide class of establishing liability favored settlement, we believe the question is much closer than it thought, and thus the factor does not weigh heavily in favor of settlement as the district court believed.
We do not gainsay that the plaintiff class faced considerable obstacles in establishing liability. First, it is not clear that the plain*815tiffs could maintain the federal causes of action (the Lanham Act and Magnuson-Moss Act claims) without some proof that the trucks suffered some decline in value, which the class was unable to demonstrate by published Kelley Blue Book figures. Second, the trucks complied with the applicable federal safety standards during the relevant times. This would undoubtedly be strong, though not necessarily conclusive, evidence that the trucks were not (legally) defective. Statistics offered by GM also suggest that the trucks presented no greater risk than other trucks or vehicles. Moreover, data from actual accidents, as opposed to crash tests, failed to reveal any statistically significant difference in post-collision fires, injuries or death relative to Ford or Dodge full-size pickups. Finally, to the extent that state law requires proof of individualized reliance for the misrepresentation claims, that would seem to pose a substantial barrier to proving class-wide liability (though, as noted below, that issue can be the subject of separate proceedings).
On the other hand, we are not impressed by some of the factors relied upon by the district court to support its finding of substantial risk in proving liability. The court cited the legal obstacles faced by the class, such as statutes of limitations varying in different states, the lack of vertical privity for the warranty claims (required in some states), the varying expiration of warranty durational limits, and the bar under some state laws to recovery for economic losses on tort claims. In response to these concerns, we point out that variations in the state procedural rules applicable to the class members have not prevented courts, including this one, from adjudicating class claims. See Hoxworth v. Blinder, Robinson & Co., 980 F.2d 912, 924 (3d Cir.1992) (affirming finding of (b)(3) predominance despite differences in state law). Moreover, even if these variations precluded the successful prosecution of the class claims, qua class claims, in this case they would not necessarily doom the action to failure.
Many of the difficulties posed by these variations could have been surmounted (or were more likely to be surmounted) if the action were not treated as a national class. Hence, the fact that the only other national automotive product defect class action ended in a defense verdict does not weigh heavily in favor of settlement. Indeed, to the extent that state-by-state variations in procedural laws created legal obstacles, the district court should have considered dividing the action into geographic sub-classes instead of considering the entire nationwide class to be hobbled. Additionally, the court should have considered making the inquiry we made in In re School Asbestos Litig., 789 F.2d at 1011, as to whether the case in terms of claims and defenses might fall into three or four patterns so that, with the use of special verdict forms, the case might have been manageable.
We also note that, in other cases, courts have certified nationwide mass tort class actions, which also include myriad individual factual and legal issues, relying on the capacity for a court to decertify or redefine the class subsequently if the ease should become unmanageable. See, e.g., In re School Asbestos Litig., 789 F.2d at 1011 (3d Cir.1986). See also Bruce H. Nielson, Was the Advisory Committee Right?: Suggested Revisions of Rule 23 to Allow More Frequent Use of Class Actions in Mass Tort Litigation, 25 HaRV.J.Legis. at 469 (“Some federal district court judges have suggested that this problem could be overcome and that multistate and nationwide classes could be certified by ... using Rule 23(c)(4) subclasses to account for variances in state law....”); see infra discussion on Risks of Maintaining Class Status. In any event, the failure of the district court to analyze the applicability of these various defenses to the different groups of plaintiffs may itself constitute an abuse of discretion. See Piambino v. Bailey, 610 F.2d 1306, 1329 (5th Cir.1980) (“[Vacatur] is demanded by the failure to assess the interests of the categories of plaintiffs and whether the settlement was fair, adequate and reasonable as to each.” (emphasis in original)); see also Piambino v. Bailey, 757 F.2d 1112, 1140 (11th Cir.1985).
In addition, a plethora of other evidence buttressed the class claims. First, the depositions and affidavits from GM engineers, including Elwell, characterizing the design as *816indefensible would have strongly supported the class claims notwithstanding the fact that Elwell was arguably vulnerable to impeachment on the basis of his own employment history. Second, the evidence from Zelenuk v. GM, No. 96-131262 (Tex.1992), that GM concealed crash tests might have been admitted in this proceeding. Third, the fact that GM has prevailed in three of the eight C/K pickup product liability trials does not support the settlement by confirming the weakness of the underlying claims, for at least two other plausible interpretations could explain this statistic. The fact that plaintiffs prevailed in five of the eight actions suggests that the claim alleged here was not so weak after all, at least if alleged by suitable plaintiffs. Moreover, such a statistic understates plaintiffs recoveries for these types of claims by not accounting for the individual settlements that have been reached.
While we recognize that establishing liability would by no means have been easy or certain for the plaintiffs, the district court over-emphasized the importance of defenses applicable to only some class members under certain state laws and incorrectly discounted a significant body of evidence pertinent to proving liability. Therefore, it is not clear that plaintiffs faced the grave prospects on the merits that the district court apparently believed existed when it approved this settlement. In any event, the district court’s failure to distinguish between groups of plaintiffs that did and those that did not confront difficult state law defenses constitutes an abuse of discretion. Piambino, 610 F.2d at 1329.
F. Risks of Establishing Damages
Like the previous factor, this inquiry attempts to measure the expected value of litigating the action rather than settling it at the current time. The district court relied heavily on this factor in approving the settlement: “[B]ecause the plaintiffs cannot adequately prove diminished value [of the pickups], the court concludes that risks of proving damages weigh strongly in favor of approval of the proposed class settlement.” We do not share the district court’s confidence, and conclude that this factor does not weigh strongly in favor of settlement.
GM argues that the class’s warranty claim amounts to a claim for diminished resale value. Some United States Courts of Appeals and some state courts have rejected such claims either on the grounds that a warranty of merchantability does not include any guarantee about the product’s resale value, see, e.g., Carlson v. General Motors Corp., 883 F.2d 287, 298 (4th Cir.1989), cert. denied, 495 U.S. 904, 110 S.Ct. 1923, 109 L.Ed.2d 287 (1990), or on the basis that the tort law of many states precludes tort claims for purely economic loss. In assessing this Girsh factor, the district court relied on its belief that the class could not demonstrate any diminution of the trucks’ value relative to Ford and Dodge trucks by referring to the Kelley Blue Book.
We do not, however, believe that this is the only permissible approach to measuring the value of the defect. According to the Uniform Commercial Code, “the measure of damages for breach of warranty is the difference at the timé and place of acceptance between the value of the goods accepted and the value they would have had if they had been as warranted, unless special circumstances show proximate damages of a different amount.” UCC § 2-714(2). Although diminished resale value might represent one method of measuring the damage suffered by owners from the publicity about the fuel tanks, it does not fully measure the difference between the value the defect-free truck would have had at delivery and the actual value of the truck as delivered. Measuring damages with a focus on resale value confounds the effects of varying rates of depreciation with the effect of the defect on the market value. The comparisons to the trucks of other manufacturers are similarly deficient measures since they fail to gauge the effect of the defect on the value of the trucks at delivery.
The cost of a retrofit, which effectively puts the truck in the condition in which it allegedly should have been delivered, may constitute an alternative measure of the damages arising from the breach of warranty. It *817has the advantage of avoiding the speculative exercise of ascertaining the hypothetical value of defect-free trucks. See, e.g., McGrady v. Chrysler Motors Corp., 46 Ill.App.3d 136, 4 Ill.Dec. 705, 360 N.E.2d 818, 821-22 (1977) (affirming an award of actual repair expenses where measuring value of vehicles as warranted upon delivery would be speculative); Nelson v. Logan Motor Sales, Inc., 179 W.Va. 539, 370 S.E.2d 734, 737 (1988) (reversing a ruling that repair costs were not evidence relevant to the value of the goods as accepted). Nothing in the UCC precludes such a measure; in fact, § 2-714(1) of the Uniform Commercial Code provides:
Where the buyer has accepted goods and given notification he may recover as damages for any nonconformity the loss resulting in the ordinary course of events from the seller’s breach as determined in any manner which is reasonable.
(emphasis added) (citation omitted).33
Because the district court based its appraisal of this factor on its exclusive reference to the Kelley Blue Book and refused to consider alternative measures that appear to provide concrete (and substantial) damage figures, we believe that the court erred in finding that the risks of proving damages were so great that they strongly favored settlement approval.
G. Risks of Maintaining Class Status
The value of a class action depends largely on the certification of the class because, not only does the aggregation of the claims enlarge the value of the suit, but often the combination of the individual cases also pools litigation resources and may facilitate proof on the merits. Thus, the prospects for obtaining certification have a great impact on the range of recovery one can expect to reap from the action.
The district court found that this factor favored settlement, although it did not place great weight on it. The court cited the “myriad factual and legal issues”34 and the vigorous contest waged by GM prior to settlement negotiations as the basis for this finding. Id. Two observations, which the district court appeared to ignore, weaken the basis for its finding that the risk involved in maintaining class status favored settlement.
First, Rule 23(a) does not require that class members share every factual and legal predicate to meet the commonality and typicality standards. Baby Neal v. Casey, 43 F.3d 48, 56 (3d Cir.1994); Hassine v. Jeffes, 846 F.2d 169 (3d Cir.1988). Indeed, a number of mass tort class actions have been certified notwithstanding individual issues of causation, reliance, and damages. See, e.g., In re School Asbestos Litig., 789 F.2d at 1009. Because separate proceedings can, if necessary, be held on individualized issues such as damages or reliance, such individual questions do not ordinarily preclude the use of the class action device. See, e.g., Eisenberg v. Gagnon, 766 F.2d 770, 786 (3d Cir. 1985).
For example, in School Asbestos, the court certified a nationwide (b)(3) class after counsel demonstrated to the court how the laws of the 50 states could be reduced to four general patterns, providing the framework for subclasses if the nationwide action had proven unmanageable.35 School Asbestos, *818789 F.2d at 1011. Although there was no such demonstration in this case, we have no reason to doubt that such a demonstration would have been possible, for we cannot conceive that each of the forty-nine states (excluding Texas) represented here has a truly unique statutory scheme, or that all of the model years possessed distinct fuel tank designs. Damage issues, moreover, are not as individualized as the district court seemed to assume: the cost of repair could have served as the measure, and that cost would not vary much among class members. Hence, it is quite possible that a nationwide class could have been properly certified here.
Second, even if the action could not be certified as it was originally filed, the district court disregarded the possibility that there were other ways to aggregate the litigation and/or adjudication of these claims. The court might have considered dividing the class into geographic or model-year subclasses or allowing the case to continue as a multi-district litigation for the remainder of pre-trial discovery. Each of those alternatives could have surmounted some of the individual issues while retaining some of the substantive advantages of the class action as framed here. Thus, the court’s conclusion that this factor favored settlement may have reflected its mistaken all-or-nothing approach to certifying this national class.
Additionally, some of the district court’s bases for finding such a significant risk in the ability to maintain class status undermine our confidence in the appropriateness of the district court’s certification of the settlement class. For instance, if the district court correctly concluded that there were insurmountable barriers to class treatment, it could not certify the class for settlement purposes. See Part IV.F supra. It is true that settlement can reduce the differences among class members. But as we have explained, the standard for certification is the same for settlement classes as for conventional classes.
Moreover, if the class members’ claims differed so much as to preclude certification even of geographic subclasses, a settlement that treats all class members alike cannot be adequate and fair to all of them. For reasons stated above, this settlement does not even appear to treat all members of the class equitably. See Parts Y.B and VI.A.l.b supra. Indeed, the settlement arguably affords the least relief to those class members with the most valuable claims, i.e., the fleet owners. See Part VLA.l.b supra. The district court’s concern, therefore, that the class could not maintain its class status, is somewhat inconsistent with its certification of the class for settlement purposes.36
We must agree that this class, even if appropriately crafted, confronted significant difficulties in maintaining its status in light of the claims alleged. Nevertheless, we are once again left with the impression that the district court too hastily approved a settlement because of its perhaps exaggerated concern that the quite ambitious initial nationwide definition of the class made it too difficult to form a class or group of classes capable of litigating these claims.
H. Ability to Withstand Greater Judgment
We find no error in the district court’s resolution of this final Girsh factor — whether the defendant has the ability to withstand a greater judgment. The district court determined that GM “could withstand a judgment greater than the proposed settlement,” although it did not attribute any significance to this finding “under these facts.”
I. Summary
Assuming arguendo that the district court had validly certified the settlement class (i.e., properly determined that it met the requisites of Rule 23), we hold that the settlement is not fair, reasonable, or adequate under the nine factor Girsh test of this circuit. The case was simply settled too quickly with too little development on the merits for certificates that may well be worth significantly less than the $1.98 to $2.18 billion estimate accepted by the district court. We conclude that the district court erred by accepting plaintiffs’ witness’s estimated valuations *819when those so clearly lacked a sound methodological basis and when there were so many other indications — including the inability of fleet owners and less wealthy class members to use the certificates, the dubious value of the transfer option, and GM’s own apparent valuation of the claim — that the settlement was inadequate and unreasonable, and may even have been a marketing boon to GM.
Additionally, the failure of this settlement to abate the lingering safety problem, despite the voeiferousness of the arguments for some recall or retrofit in the initial complaint, enhances our conviction that this settlement is inadequate. Beyond its dubious valuation of the settlement, the district court also overestimated the risks of proving liability and damages and of maintaining class-status, and under-estimated the true degree of opposition to the settlement. The district court, however, correctly applied the complexity-of-suit and defendant’s-capacity-for-greater-judgment factors.
Although we are not bound in any way by the proceedings in the separate Texas action, our decision today shares many of the concerns expressed by the Texas appellate court which set aside an approval of a very similar coupon settlement. See Bloyed, 881 S.W.2d at 422. Rule 42 of the Texas Rules of Civil Procedure, which governs class actions, is patterned after Federal Rule of Civil Procedure 28. The Bloyed court also was concerned about a settlement that provided absolutely nothing to those unwilling or unable to purchase another GM truck and that did nothing about the allegedly dangerous vehicles left on the road. The Texas court objected as well to the $9.5 million in attorneys’ fees negotiated between that class’s counsel and GM.
Balancing the Girsh factors, on the current record, this settlement clearly fails to meet the standards required for judicial approval. We leave open the possibility, however, that the district court on remand might develop the record more fully, properly approve the settlement, in either its original or a renegotiated form, and, following the guidance offered by this opinion, certify the settlement class.
VII. APPROVAL OF THE ATTORNEYS’ Fee Award
The French and Young objectors also contest the district court’s award of attorneys’ fees. (Order dated Dec. 20, 1993, 1993 WL 533155 and Feb. 2, 1994, 1994 WL 30301.) The court initially awarded fees without an independent review of the agreement, explaining its refusal to review the award: “[The fee agreement] is a matter of contract between the parties, rather than a statutory fee case, ... and payment of the fees will have no impact on the class members....” Subsequently, on February 2, 1994, the court issued an “amplification” of its prior ruling, which justified the award under both the lodestar37 and the percentage of recovery38 methods. Class counsel maintain that the objectors lack standing to contest the agreement made between GM and themselves, and that the objectors waived their right to appeal the award by not raising their objections below. Although our disposition of the certification and settlement approval issues obviates the need for a review of the fee award at this stage (and moots the waiver question), we highlight some of the primary issues in analyzing the appropriateness of a particular fee agreement for the district court on remand (in the event that the record is expanded, the class certified, and the settlement approved).
At the outset, we note that a thorough judicial review of fee applications is required in all class action settlements. The district court did not accommodate practical realities here when, rationalizing its initial refusal to review the fee, it stated that the fee award was “to be paid by General Motors Corporation and will in no way reduce the recovery to any of the settlement class members.” Indeed, this court has recognized that “a defendant is interested only in dispos*820ing 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.” Prandini v. National Tea Co., 557 F.2d 1015, 1020 (3d Cir.1977); 2 Newberg & Conte § 11.09 (purpose of judicial review is to police abuses even where defendant pays plaintiffs fees). In light of these realities, class counsel’s argument that objectors have no standing to contest the fee arrangement is patently meritless: the fee agreement clearly does impact their interests, as it is, for practical purposes, a constructive common fund.
Moreover, as discussed at length in the adequacy of representation section, see Part V.B.3 supra, the divergence in financial incentives present here creates the “danger ... that the lawyers might urge a class settlement at a low figure or on a.less-than-optimal basis in exchange for red-carpet treatment for fees,” Weinberger v. Great Northern Nekoosa Corp., 925 F.2d 518, 524 (1st Cir.1991). See also Prandini, 557 F.2d at 1020 (“When the statute provides that a fee is to be paid as a separate item, the conflict between client and attorney may not be as apparent_ It is often present nonetheless.”). This generates an especially acute need for close judicial scrutiny of fee arrangements that implicate this concern. See In re Agent Orange Prod. Liab. Litig., 818 F.2d 216, 224 (2d Cir.1987) (“The test to be applied is whether, at the time a fee sharing agreement is reached, class counsel are placed in a position that might endanger the fair representation of their clients and whether they will be compensated on some basis other than for legal services performed.”); Piambino v. Bailey, 757 F.2d at 1139 (“Because of the potential for a collusive settlement, a sellout of a highly meritorious claim, or a settlement that ignores the interests of minority classes members, the district judge has a heavy duty to ensure that ... the fee awarded plaintiffs’ counsel is entirely appropriate.”). We have previously acknowledged that the potential for conflict between the class and its counsel is not limited to situations meeting the strict definitions of a common fund.39
As we have also explained in this opinion, courts must be especially vigilant in searching for the possibility of collusion in pre-certification settlements. See Part IV.E supra. In addition, the court’s oversight task is considerably complicated by the fact that these attorney-class conflicts are often difficult to discern in the class action context, “where full disclosure and consent are many times difficult and frequently impractical to obtain.” Agent Orange, 818 F.2d at 224 (citations omitted). Finally, we emphasize that the court’s oversight function serves not only to detect instances of “the actual abuse [that potential attorney-class conflicts] may cause, but also [the] potential public misunderstandings they may cultivate in regard to the interests of class counsel.” Agent Orange, 818 F.2d at 225 (citing Susman v. Lincoln American Corp., 561 F.2d 86, 95 (7th Cir.1977), and Prandini v. National Tea Co., 557 F.2d 1015, 1017 (3d Cir.1977)); see also Grinnell I, 495 F.2d at 469; Model Code of PROFESSIONAL Responsibility Canon 9 (1975). On remand, therefore, the district court must be alert to the presence in the fee agreement of any actual abuse or appearance of abuse capable of creating a public misunderstanding.
Having emphasized the necessity for judicial review of fee awards in all class action settlements, we will briefly clarify some principles of fee approval for the district court to apply on remand if it certifies a class and approves a settlement.
Because the district court purported to use both the lodestar and the percentage-of-re-eovery methods, the actual grounds for its approval of the fee are not entirely clear. Although it is sensible for a court to use a second method of fee approval to cross check its conclusion under the first method, we believe that each method has distinct advantages for certain kinds of actions, which will make one of the methods more appropriate as a primary basis for determining the fee. *821Here, for the reasons that follow, the court should probably use the percentage-of-reeov-ery rather than the lodestar method as the primary determinant, although the ultimate choice of methodology will rest within the district court’s sound discretion.
The lodestar and the percentage of recovery methods each have distinct attributes suiting them to particular types of cases. See Task Force, 108 F.R.D. at 250-53. Ordinarily, a court making or approving a fee award should determine what sort of action the court is adjudicating and then primarily rely on the corresponding method of awarding fees (though there is, as we have noted, an advantage to using the alternative method to double check the fee).40
Courts generally regard the lodestar method, which uses the number of hours reasonably expended as its starting point, as the appropriate method in statutory fee shifting cases. Because the lodestar award is decoupled from the class recovery, the lodestar assures counsel undertaking socially beneficial litigation (as legislatively identified by the statutory fee shifting provision) an adequate fee irrespective of the monetary value of the final relief achieved for the class.
This de-coupling has the added benefit of avoiding subjective evaluations of the monetary worth of the intangible rights often litigated in civil rights actions. Outside the pure statutory fee case, the lodestar rationale has appeal where as here, the nature of the settlement evades the precise evaluation needed for the percentage of recovery method. The lodestar method has the added benefit of resembling modes of fee determination in conventional bipolar litigation. On the other hand, the lodestar method has been criticized as giving class counsel the incentive to delay settlement in order to run up fees while still failing to align the interests of the class and its counsel, and for not rewarding counsel incrementally for undertaking the risk of going to trial. See Coffee, Understanding the Plaintiffs Attorney, 86 Co-lumL.Rev. at 691.
Courts use the percentage of recovery method in common fund cases on the theory that the class would be unjustly enriched if it did not compensate the counsel responsible for generating the valuable fund bestowed on the class. See Task Force, 108 F.R.D. at 250. Because these cases are not presumed to serve the public interest (as evidenced by the lack of a fee statute), there is no social policy reason that demands an adequate fee. Instead, the court apportions the fund between the class and its counsel in a manner that rewards counsel for success and penalizes it for failure. Courts have relied on “common fund” principles and the inherent management powers of the court to award fees to lead counsel in cases that do not actually generate a common fund. See, e.g., In re Air Crash Disaster at Florida Everglades, 549 F.2d 1006 (5th Cir.1977) (using common fund principles in settlement of consolidated cases). The rationale behind the percentage of recovery method also applies in situations where, although the parties claim that the fee and settlement are independent, they actually come from the same source.
We believe that this case presents a situation more closely aligned with the common fund paradigm than the statutory fee paradigm. Although class counsel and GM contend (and the district court believed) that the fee was a separate agreement, thus superficially resembling the separate awards in statutory fee cases, private agreements to structure artificially separate fee and settlement arrangements cannot transform what is in economic reality a common fund situation into a statutory fee shifting case. Certainly, the court may select the lodestar method in some non-statutory fee cases where it can calculate the relevant parameters (hours expended and hourly rate) more easily than it can determine a suitable percentage to award. But the court must vigilantly guard against the lodestar’s potential to exacerbate the misalignment of the attorneys’ and the class’s interests. See Coffee, Understanding *822the Plaintiffs Attorney, 86 Colum.L.Rev. at 717.
In this case, the fee clearly was not made pursuant to a statute; therefore no legislatively endorsed policy favors assuring counsel an adequate fee. And the settlement, though difficult to value, did not award the even more hard-to-value intangible rights that could in some limited circumstances justify using the lodestar method. In sum, although this ease presents a hybrid, we believe that it more closely resembles a common fund case.
At all events, to the extent that the district court relied on the lodestar method, it erred by applying a multiplier. In the lodestar section of its analysis, the district court calculated the multiplier needed to apply to the simple lodestar result, $3,158,182, to obtain the requested amount, $9,500,000. (see Feb. order, 1994 WL 30301 at *2.) After estimating the multiplier to be between 2.5 and 3, the court proceeded with a “contingent nature of the success” analysis of the multiplier’s appropriateness from Lindy. See Lindy Bros. Builders Inc., v. American Radiator & Standard Sanitary Corp., 540 F.2d 102, 116-17 (3d Cir.1976). The Supreme Court, however, has rejected the use of multipliers to enhance the lodestar’s hourly rate amount. See City of Burlington v. Dague, — U.S. at ——, 112 S.Ct. at 2638. Notwithstanding this clear Supreme Court precedent, GM’s counsel failed to apprise the district court about Dague even though its pertinence was patent.
To the extent that the district court construed the fee agreement as a common fund, its analysis also appears to misapprehend key aspects of the percentage of recovery method. In common fund cases, a district judge can award attorneys’ fees as a percentage of the fund recovered. See Blum v. Stenson, 465 U.S. 886, 900 n. 16, 104 S.Ct. 1541 n. 16, 79 L.Ed.2d 891 (1984); In re Smithkline Beckman Corp. Secur. Litig., 751 F.Supp. 525 (E.D.Pa.1990). One court has noted that the fee awards have ranged from nineteen percent to forty-five percent of the settlement fund. Id. at 533. Here, the district court summarily asserted that, although it could not value the settlement precisely, “whatever method is used in computing the ultimate value of the settlement, the attorneys’ fees sought in this action will constitute an extremely small percentage of the total value and will be minute compared to the aforesaid 19-45% range.” (Feb. order, 1994 WL 30301 at *4.)
Given our skepticism of the settlement’s value generally and of Simonsen’s estimates in particular (see supra discussion on settlement fairness), we are much less sanguine that the $9,500,000 fee actually constitutes an acceptable percentage of the class recovery. On the current record, we are constrained to reject that conclusion. At the very least, the district court on remand needs to make some reasonable assessment of the settlement’s value and determine the precise percentage represented by the attorneys’ fees. The problem, however, is not simple, for arguably, any settlement based on the award of certificates would provide too speculative a value on which to base a fee award. (See Task Force, 108 F.R.D. at 250-53 (discussing the preferability of the lodestar method for civil rights actions where the difficulty of valuing injunctive relief complicates the calculation of a fee using the percentage method.))
On remand, the district court might wish to examine the fee primarily under the percentage of recovery scheme. If so, the court will need to determine a precise valuation of the settlement on which to base its award. The court may however, as a check, want to use the lodestar method to assure that the precise percentage awarded does not create an unreasonable hourly fee.
VIII. Other Issues; Conclusion
Objectors also appealed the district court’s denial of discovery into the settlement negotiations and the adequacy of the notice with respect to the attorneys’ fees. In light of our holding on the certification and settlement approval issues and its effect on the need for us to judge the fee award, we need not reach these issues.
For the foregoing reasons, we will vacate the orders certifying the provisional class and approving the settlement and remand for *823further proceedings consistent with this opinion. Parties to bear their own costs.
. Rule 23 provides, in pertinent part:
(a) Prerequisites to a Class Action. One or more members of a class may sue or be sued as representative parties on behalf of all only if
(1)the class is so numerous that joinder of all members is impracticable, (2) there are questions of law or fact common to the class, (3) the claims or defenses of the representative parties are typical of the claims or defenses of the class, and (4) the representative parties will fairly and adequately protect the interests of the class.
(b) Class Actions Maintainable. An action may be maintained as a class action if the prerequisites of subdivision (a) are satisfied, and in addition:
(1) the prosecution of separate actions by or against individual members of the class would create a risk of
(A) inconsistent or varying adjudications with respect to individual members of the class which would establish incompatible standards of conduct for the party opposing the class, or
(B) adjudications with respect to individual members of the class which would as a practical matter be dispositive of the interests of the other members not parties to the adjudications or substantially impair or impede their ability to protect their interests; or
(2) the party opposing the class has acted or refused to act on grounds generally applicable to the class, thereby making appropriate final injunctive relief or corresponding declaratory relief with respect to the class as a whole; or
(3) the court finds that the questions of law or fact common to the members of the class predominate over any questions affecting only individual members, and that a class action is superior to other available methods for the fair and efficient adjudication of the controversy. The matters pertinent to the findings include: (A) the interest of members of the class in individually controlling the prosecution or defense of separate actions;
(B) the extent and nature of any litigation concerning the controversy already commenced by or against members of the class;
(C) the desirability or undesirability of concentrating the litigation of the claims in the particular forum; (D) the difficulties likely to be encountered in the management of a class action.
(c)Determination by Order Whether Class Action to be Maintained; Notice; Judgment; Actions Conducted Partially as Class Actions.
(1) As soon as practicable after the commencement of an action brought as a class *778action, the court shall determine by order whether it is to be so maintained. An order under this subdivision may be conditional, and may be altered or amended before the decision on the merits....
(e) Dismissal or Compromise. A class action shall not be dismissed or compromised without the approval of the court, and notice of the proposed dismissal or compromise shall be given to all members of the class in such manner as the court directs.
.The class includes both mid-and full-size tracks with chassis model types C, K, R, or V.
.See note 5 infra.
. GM reached a substantially identical agreement with counsel representing a class of C/K pickup truck purchasers who are Texas residents in Dollar v. General Motors, No. 92-1089 (71st Judicial District, Marshall, Tex.) (JA 1708, 1746). That settlement was approved in November 1993, but was overturned on appeal on June 22, 1994. See Bloyed v. General Motors Corporation, 881 S.W.2d 422 (Tex.Ct.App.1994), discussed infra at VI(I). The Texas Supreme Court granted GM’s Application for Writ of Error on February 16, 1995 and set the case for oral argument on March 21, 1995.
. After oral argument in this case, United States Transportation Secretary Federico Pena announced that NHTSA had settled the proceeding involving the C/K trucks at issue here without ordering a recall, finding an acceptable retrofit, or giving any compensation to the truck owners. The settlement provided that GM would contribute $51 million to general safety programs unrelated to the trucks' alleged problems. See Statement by Secretary Federico Pena on Dec. 2, 1994, Settlement Regarding DOT Investigation of General Motors C/K Pickup Trucks.
. See note 5 supra.
. See, e.g., John C. Coffee, Jr., The Corruption of the Class Action, Wall St. J. Sept. 7, 1994, at A15.
. 2 Newberg & Conte § 11.27 (First) § 1.46; Roger H. Transgrud, Joinder Alternatives in Mass Tort Litigation, 70 Cornell L.Rev. 779 (1985); Bruce H. Nielson, Was the 1966 Advisory Committee Right?: Suggested Revisions of Rule 23 to Allow More Frequent Use of Class Actions in Mass Tort Litigation, 25 HarvJ.Legis. 461, 480.
. These sorts of dynamics have led some critics to accuse class action attorneys of ethical violations. While we emphasize that counsel here committed no such violations, we do not preclude the possibility that these violations could occur.
. Because the parties do not come before the court until the action has settled, some courts have even expressed concern that such cases do not present a case or controversy for Article III purposes. Cf. Carlough v. Amchem Products, Inc., 834 F.Supp. 1437, 1462-67 (E.D.Pa.1993).
. We are somewhat dubious of the court's ability to discharge its duties completely under these circumstances. See Part IV.E infra.
. MCL2d expressed concern about partial settlements (settlements only as to certain plaintiffs or certain defendants) since "[m]embers of the settlement class will almost certainly find it difficult to understand their position in the litigation.” MCL2d § 30.45. '
. The terms "tentative” and "provisional” appear to be used interchangeably.
. "Conditional” is actually a term that can be properly applied to all class actions, even those that are certified in the normal process. Under Rule 23(c)(1), the court retains the authority to re-define or decertify the class until the entry of final judgment on the merits. This capacity renders all certification orders conditional until the entry of judgment. See MCL2d § 30.18.
. See also In re Mid-Atlantic Toyota Antitrust Litig., 564 F.Supp. 1379, 1388 n. 13 (D.Md.1983) ("Completely ancillary to the proposed settlement, [a temporary settlement class] lasts only as long as the period between the preliminary approval of the settlement and the court’s final determination on the settlement. In effect, a temporary settlement class serves only as a procedural vehicle for providing notice to putative members of a proposed class....”).
. This conclusion is supported by the text of Rule 23(e). That section provides that “class action” may not be compromised without court approval, and arguably a case is not a “class action” in the absence of such findings.
. We note in this regard that other courts have made the determinations of adequacy of representation and homogeneity of the class when evaluating the fairness of the settlement for the express purpose of assuring that they possess enough information to execute their Rule 23(e) duty. See In re Beef Industry Antitrust Litig., 607 F.2d at 173 n. 4 (quoting Arthur R. Miller, An Overview'op Federal Class Actions: Past, Present and Future (Federal Judicial Ctr.1977)); see also In re Baldwin-United Corp., 105 F.R.D. 475, 483 (S.D.N.Y.1984) (making findings in the opinion which preliminarily approved the settlement).
. As the case before us involves a damages class under Rule 23(b)(3), we do not address the application of the (b)(1) and (b)(2) requisites which. without the important right to opt out, involve . different considerations.
. In Malchman v. Davis, 706 F.2d at 433, the court was satisfied by the district court's determination that the settlement class satisfied the adequacy of representation inquiry noting: “There is no doubt that the district court must make an independent evaluation of whether the named plaintiffs were adequate representatives of the class.... A judge has an obligation to consider whether the interest of the class are adequately represented." (citing East Texas Motor Freight Sys., Inc. v. Rodriguez, 431 U.S. 395, 403-06, 97 S.Ct. 1891, 1896-98, 52 L.Ed.2d 453 (1977)); see also Plummer, 668 F.2d at 659 & n. 4. We agree that this is an appropriate focus given the heightened potential for collusion, buy-offs and other abuses in settlement class situations where the negotiations occur before the court appoints class representatives and counsel. We still believe, however, that courts should assure that settlement classes meet all of the requirements of 23(a) and (b). This prescription is consistent with the heightened duty of courts in class action settlements to assure that the absentees' rights are adequately protected.
. For example, in finding adequate representation, one court noted: "[S]o long as all class members are united in asserting a common right, such as achieving the maximum possible recovery for the class, the class interests are not antagonistic for representation purposes.” In re Corrugated Container Antitrust Litig., 1980-1 Trade Cas. (CCH) ¶ 63, 163 at 77, 788 n. 10, 1979 WL 1751 (S.D.Tex.1979), aff'd, 643 F.2d 195 (5th Cir.1981) (citing Wright & Miller, Fed Practice & Procedure Civil § 1768, at n. 7 & 8).
. Indeed, if any difference in standards is warranted, pre-certification settlement may raise the adequacy of representation standard.. Since this inquiry must ascertain “whether there has been any collusion or undue pressure by the defendants on would be class representatives,” see First Comm. Corp. of Boston Consumer Accts. Litig., 119 F.R.D. 301, 308 (D.Mass.1987); Alvarado Partners LP v. Mehta, 723 F.Supp. 540, 546 (D.Colo.1989), it must carry greater weight in the settlement class context where there is an enhanced potential for those evils. Thus, while the other 23(a) findings remain important when the action settles, the need to assure an absence of collusion and an alignment of interests assumes an especially crucial role. Reliance, for the class requisites analysis, on the settlement's terms and process also increases the importance of an independent conclusion of adequate representation (i.e., one not derived solely by reference to the nature of the negotiations).
. In the School Asbestos case, 789 F.2d at 996, the panel asked counsel to analyze all the claims and defenses and write a report reflecting whether the differing claims and defenses evidence a small number of patterns that would be amenable to trial through a series of special verdicts. The plaintiffs came up with a demonstration that the claims and defenses were reducible to four patterns. That, in our view, was sufficient to satisfy the commonality and typicality inquiries. The same might be true in this case.
. The information on fee agreements may prompt potential objectors to oppose not only the awards but, also, to the extent they conclude that arm’s length negotiations were compromised, the adequacy of the settlement and the propriety of the class.
. Other cases and authorities have followed this guide. See, e.g., Ashley v. Atlantic Richfield Co., 794 F.2d 128 (3d Cir.1986); MCL2d § 30.41; 2 Newberg & Conte § 11.29, at 11-62 (recognizing potential for conflict where settlement and fees to be paid by defendant simultaneously negotiated). To implement this prophylactic bar fully, courts would have to require class counsel to disclose all understandings as to fees, not simply concluded, formal agreements. See MCL2d § 34.42, at 237-39. Although it recognized that this prophylactic rule could impede some settlements by making it impossible for the defendant to size up its total liability (i.e. the sum of the settlement amount and any fees the defendant agrees to pay), Task Force, 108 F.R.D. at 267-69, the Task Force concluded that avoiding the conflicts justified this cost.
. While the parties could have sought a waiver pre-approving simultaneous negotiations, Task Force at 269, the parties did not seek one here.
. This is true of, for example, the State of Iowa; State of Indiana; West Virginia Department of Transportation; State of New York; Commonwealth of Pennsylvania Department of Transportation and Department of General Services; County of Los Angeles, California; Jefferson Parish, Louisiana; and the City of New York.
. GM was apparently so eager to have this $9.5 million fee approved that its counsel did not even object when the district court applied a multiplier notwithstanding clear Supreme Court precedent invalidating the use of multipliers. See City of Burlington v. Dague, — U.S. at -, 112 S.Ct. at 2638. In our view, the fact that counsel to this large multinational corporation did not object to this clear error raises a smoking gun signaling GM’s awareness of the questionable settlement it made.
. In the amended consolidated complaint, class counsel described the trucks as "rolling firebombs” and estimated that an additional 200 deaths would occur unless GM took prompt corrective action.
. The pleadings alleged a dominant control theory which, if successful, would have required GM, the manufacturer and distributor of these vehicles, to remedy the allegedly unreasonable safety defects before they could cause or exacerbate the damage and injury resulting from a side impact collision.
. The district court also based its conclusion that the settlement was reasonable relative to the best possible recovery (i.e., relative to the relief requested) on its doubts that a court could or should award the recall/retrofit remedy requested. The court expressed some doubt that it had the power to order a recall by injunction, citing Walsh v. Ford Motor Co., 130 F.R.D. 260 (D.D.C.1990), and National Women's Health Network Inc. v. A.H. Robins Co., 545 F.Supp. 1177 (D.Mass.1982). Neither of these cases, however, conclusively establishes that the district court *812would lack the power to order a recall. The fact that no court has done it before and that there may be some logistical issues to surmount do not themselves support the court's conclusion; other class actions and complex litigation settlements have developed mechanisms for supervising and enforcing compliance with detailed affirmative injunctions. See, e.g., MCL2d § 33.55 ("The court may also decide to appoint a master under Rule 53 to monitor future implementation of injunctive features of the settlement.”). Although we intimate no view on the matter, it does seem to warrant further consideration. At all events, the district court could clearly have awarded relief that would require GM to set up a fund to finance retrofits initiated by the owners individually. See Bloyed, 881 S.W.2d at 433. The district court, therefore, did not lack the power to order a remedy that would have been more responsive to the class's concern about leaving the trucks on the road.
. There may also have been other deficiencies in the notice. The fact that the notice did not disclose the attorneys fees that the class counsel and defendants agreed to, and the fact that the notice suggested that class members could also have a recall remedy from NHTSA (though many of the trucks were so old that NHTSA lacked the power to recall them), may also have helped suppress potential objection.
. Of approximately 5.7 million class members, 6,450 owners objected and 5,203 opted out.
. GM argues that a repair remedy is available only when it is less costly to the defendant than diminution in value. We think that such rigid rules are inappropriate, and that the court should carefully consider all of the proffered measures of damages. In any event, the costs of retrofit, though unsettled by the district court as of this juncture, will be less than the diminution in value (if the settling parties’ valuation of the certificates is any indication of that diminution).
. The legal issues that might vary among class members included the claims of breach of warranty, negligent misrepresentation, and negligence and products liability, which would be based on the various state laws. Potentially variable factual issues included the fact that the disputed trucks did not use a single gas tank design, and the individualized proof of reliance required in some jurisdictions for fraud, negligent misrepresentation, and breach of warranty claims.
.The district court could retain the sub-classes although they might not have properly been brought in that court originally. Cf. In re Agent Orange Prod. Liab. Litig., 100 F.R.D. 718, 724 (E.D.N.Y.1983) (recognizing hypothetical need of the court to apply the laws of different states); In re Agent Orange Prod., 580 F.Supp. 690 (E.D.N.Y.1984) (performing choice of law analysis).
. This anomaly is at least partially attributable to the court’s failure to certify the class in the manner required by Rule 23. But some part of the inconsistency signals that the district court ignored the various ways that the class claims could be manageably litigated.
. The lodestar method calculates fees by multiplying the number of hours expended by some hourly rate appropriate for the region and for the experience of the lawyer.
. The percentage of recovery method resembles a contingent fee in that it awards counsel a variable percentage of the amount recovered for the class.
. The common fund doctrine provides that a private plaintiff, or plaintiff's attorney, whose efforts create, discover, increase, or preserve a fund to which others also have a claim, is enti-tied to recover from the fund the costs of his litigation, including attorneys’ fees. Vincent v. Hughes Air West, Inc., 557 F.2d 759 (9th Cir.1977).
. For example, a court can use the lodestar method to confirm that a percentage of recovery amount does not award counsel an exorbitent hourly rate; similarly, the percentage of recovery method can be used to assure that counsel's fee does not dwarf class recovery.