DEBORAH G. HANKINSON, Justice.

This appeal from a trial court's decision to certify a class presents two issues: first, whether the class in a class-action lawsuit may be defined by the ultimate liability issue; and second, whether an appellate court may redefine an erroneously defined class. Plaintiffs allege that Intratex Gas Company did not take natural gas in ratable proportions from the wells of more than 900 producers. The trial court defined the plaintiff class as consisting of producers of natural gas whose gas was taken by Intratex between 1978 and 1988 in less than ratable proportions, and certified it under rule 42 of the Texas Rules of Civil Procedure. The court of appeals held that the trial court properly certified the class. 960 S.W.2d 389. We conclude that because the class definition is improperly based on the ultimate issue of alleged liability in the case, the trial court abused its discretion when it certified the class. Because we cannot modify the class definition in this case without interfering with the trial court's discretion and oversight of the class action, we decline to redefine the class on appeal. Accordingly, we reverse the judgment of the court of appeals and remand the case to the trial court for further proceedings consistent with this opinion.

Intratex is an intrastate pipeline company that purchases gas as the first purchaser from several thousand wells in west Texas along its Oasis pipeline. Plaintiffs, who are producers of natural gas, sued on behalf of a "class of persons who, at any given time from 1978 to the present, have held overriding royalty interests or working interests in natural gas producing properties on which the production was dedicated to or purchased by" Intratex. They allege that from 1978 to 1988, Intratex did not take gas ratably from them.

Plaintiffs' ratable-take claim is a discrimination claim. Texas law requires an intrastate pipeline like Intratex to buy gas without discrimination (i.e., "ratably") from the producers selling to it at the wellhead; the law also requires the producers to produce ratably. The source of the ratability doctrine is the Texas Common Purchasers Act. See Tex. Nat. Res. Code 111.081-.097; see also 16 Tex. Admin. Code 3.34. Plaintiffs predicate their claim on section 111.086 of the Act, which provides that a purchaser cannot discriminate in favor of one producer against another producer in the same field, and cannot discriminate "unjustly" or "unreasonably" between fields. Tex. Nat. Res. Code 111.086.

The Texas Railroad Commission has adopted ratable-take rules to implement section 111.086. See Tex. Admin. Code 3.34. The rules impose ratability obligations on producers and purchasers to prevent waste, protect correlative rights, prevent discrimination, and conserve the state's natural resources. See id. 3.34(b). The Commission's regulations require that a first purchaser nominate gas ratably, and in accordance with the purchaser's market demand. See id. To guard against discrimination, the regulations prohibit a purchaser from considering the contractual price, agreed upon by the purchaser and the producer, in nominating or taking natural gas. See id. 3.34(f)(1), (g)(1). A purchaser's nominations are one of the factors the Commission considers when setting a gas well's allowables, which are used in determining the maximum amount a well is permitted to produce.

Intratex purchased gas along the Oasis pipeline pursuant to contracts with the Plaintiffs; under these gas contracts, prices paid for production from some wells were lower than those paid for production from other wells. In 1978 and 1984, Intratex executed contracts with Dow Chemical Company that allowed Intratex, as the contractual first purchaser, to opt not to take delivery of gas at certain wells in certain production fields. Plaintiffs contend that Intratex selected wells to be part of the Dow-waiver "program" based on those wells' ability to produce gas in large quantities, and because the contractual prices were far below the spot-market price and system-wide weighted average cost of gas. Under these "Dow-waiver agreements," Intratex waived to Dow its proportion of gas under specific gas-purchase contracts so that Dow received the gas and made payments to Intratex at a price calculated according to an agreed-upon formula. The Dow-waiver agreements allegedly enabled Intratex to nominate waiver fields with certain low-cost waiver wells higher than nonwaiver fields, which in turn allowed Intratex to take high volumes of gas from the waiver fields below the market price or system-wide price. Plaintiffs further contend that, rather than spreading the gas takes ratably system-wide, the waiver wells were produced in high, nonratable quantities under the Dow-waiver agreements.

During the three-day certification hearing, Plaintiffs' expert, Don Rhodes, testified that Intratex had taken disproportionate shares of natural gas from approximately 970 wells between 1978 and 1988. Rhodes conducted a study of Intratex's takes over the relevant period. He claimed to have identified, based on separate gas takings, those wells that were overtaken and those that were undertaken. According to Rhodes, between 1978 and 1988, Intratex took large quantities of low-priced gas while taking less high-priced gas. In response, Intratex presented its own expert, Don George, who challenged Rhodes's methodology. Based on George's testimony, Intratex argued that a class definition derived from a suspect study could not meet the rule 42 requirements.

After hearing the certification evidence, the trial court granted Plaintiffs' motion to certify the class, but only on the issue of whether Intratex took ratably from the class members. And, instead of certifying the class originally requested by Plaintiffs, the trial court defined the class as:

All persons who were producers of natural gas sold to the defendant between January 1, 1978 and December 31, 1988 whose natural gas was taken by the defendant in quantities less than their ratable proportions.

The record suggests that the court based its class definition and its certification decision on the Rhodes study. The trial court denied certification of all other issues and claims, including allegations of statutory and regulatory violations, breach of contract, intentional or negligent misrepresentation, and fraud.

Affirming the trial court's certification order, the court of appeals concluded that the class was properly defined, and the rule 42(a) and (b)(4) requirements of numerosity, commonality, typicality, adequacy of representation, and predominance were satisfied. 960 S.W.2d at 395-99. The court of appeals held that Intratex's criticisms of the trial court's class definition were merits based and thus needed to be decided at trial. Id. at 394-95. It further concluded that while problems in the class definition might eventually require the trial court to modify the definition or decertify the class, merits-based issues should not defeat certification at such an early stage of the litigation. Id. One justice dissented on the ground that the trial court's definition requires an "inappropriate determination of the merits of the cause of action itself," because the concept of ratability "subsumes the notion of discrimination." Id. at 399.

On petition for review, Intratex argues that the class-certification order must be reversed because not only have Plaintiffs failed to fulfill the rule 42(a) and (b)(4) prerequisites, but the class definition is fundamentally defective. Specifically, Intratex argues that Plaintiffs did not satisfy the burdens imposed by rule 42(a) for class certification because: (1) ratability is an inherently individual determination, hence there is no common question of law or fact to be decided at trial; (2) the class representatives' claims are not typical of the claims of all class members; and (3) the proposed class representatives cannot adequately protect the interests of the class when intra-class antagonisms and conflicts of interest abound.

Intratex also contends that the certification order is improper because the trial court has not sufficiently defined the class: the definition is not based on objective criteria, and class membership is not presently ascertainable. Intratex argues that the trial court's class definition turns on the central question of alleged liability — whether Intratex took gas from Plaintiffs in less than ratable shares. Therefore, the trial court created a "fail-safe" class, that is, a class bound only by a judgment favorable to Plaintiffs, but not by an adverse judgment. According to Intratex, a fail-safe class, by definition, is not objective, and membership in the class cannot be determined until final judgment. Additionally, Intratex asserts that redefining the class to focus on the producers excluded from the Dow-waiver agreements, as Plaintiffs suggest, would not eliminate the problems associated with maintaining the suit as a class action.

Plaintiffs respond that the class, as defined by the trial court, takes both the Rhodes study and the Dow-waiver agreements into account, and was properly certified; not only does the class satisfy rule 42(a)'s numerosity, commonality, typicality, and adequacy-of-representation prerequisites, but it fits squarely into rule 42(b)(4) because common questions concerning Intratex's conduct predominate over any individual issues. Furthermore, they argue that the trial court did not abuse its discretion in defining the class as it did because the definition allows the court to determine the general outlines of class membership before trial.

The parties present this Court with the question of whether the trial court abused its discretion in certifying the class under Texas Rule of Civil Procedure 42. See General Motors Corp. v. Bloyed, 916 S.W.2d 949, 955 (Tex. 1996). But before a reviewing court can consider a putative class's compliance with the rule 42(a) and (b) class-certification requirements, it must first consider the appropriateness of the certified class definition.

The threshold inquiry in determining class certification must be into the parameters of the proposed class. See Ad Hoc Comm. v. City of St. Louis, 143 F.R.D. 216, 219 (E.D.Mo. 1992). Rule 42 implicitly requires the representative plaintiffs to demonstrate not only that an identifiable class exists, but that it is susceptible to precise definition. See DeBremaecker v. Short, 433 F.2d 733, 734 (5th Cir. 1970) (holding that it is "elementary that in order to maintain a class action, the class sought to be represented must be adequately defined and clearly ascertainable"); see also 5 Moore, Moore's Federal Practice 23.20-.21 (3d ed. 1999). In fact, absent a cognizable class, evaluating whether the putative class representatives satisfy the rule 42(a) and (b) requirements would be impossible. See Metcalf v. Edelman, 64 F.R.D. 407, 409 (N.D.Ill. 1974); Hettinger v. Glass Specialty Co., 59 F.R.D. 286, 296 (N.D.Ill. 1973); see also Davoll v. Webb, 160 F.R.D. 142, 146 (D.Colo. 1995) (holding that determining whether putative class meets the certification requirements is unnecessary when class definition is untenable). Hence, in evaluating whether an alleged class satisfies the certification requirements, the trial court must determine whether the putative class is sufficiently defined. See Davoll, 160 F.R.D. at 144; Ladd v. Dairyland County Mut. Ins. Co. of Tex., 96 F.R.D. 335, 338 (N.D.Tex. 1982).

A properly defined class is imperative for a suit to proceed as a class action because the class definition facilitates identifying, at the outset, the individuals affected by the litigation, and protects their interests. First, the definition determines who is entitled to notice, and, for class actions maintained under rule 42(b)(4), provides an opportunity to opt out of the class. Failure to define a class precisely creates a substantial risk that putative class members cannot adequately exercise their right to opt out of or remain in the suit before they are bound by a class judgment. Cf. General Tel. Co. of the Southwest v. Falcon, 457 U.S. 147, 160-61 (1982) (noting the unfairness to class members bound by a judgment when the class is overly broad). Second, the class definition determines the nature of the relief that can be awarded and who is entitled to that relief. See 5 Moore, supra, 23.21(3). Finally, clearly defining the class identifies the plaintiffs who will be bound by the judgment if they lose, and insures that those actually harmed by the defendant's wrongful conduct will receive the relief ultimately awarded. See Simer v. Rios, 661 F.2d 655, 670 (7th Cir. 1981).

For a class to be sufficiently defined, it must be precise: the class members must be presently ascertainable by reference to objective criteria. See DeBremaecker, 433 F.2d at 734; Newton v. Southern Wood Piedmont Co., 163 F.R.D. 625, 632 (S.D.Ga. 1995), aff'd, 95 F.3d 59 (11th Cir. 1996); Gomez v. Illinois State Bd. of Educ., 117 F.R.D. 394, 397 (N.D.Ill. 1987); Joseph v. General Motors Corp., 109 F.R.D. 635, 639 (D.Colo. 1986); see also Federal Judicial Center, Manual for Complex Litigation 30.14 (1995). This means that the class should not be defined by criteria that are subjective or that require an analysis of the merits of the case. See DeBremaecker, 433 F.2d at 734 (affirming dismissal of class action because definitional language was subject to individual interpretation); Simer, 661 F.2d at 669 (affirming denial of class certification because identifying class members based on each individual's state of mind would be a "Sisyphean task"); Forman v. Data Transfer, Inc., 164 F.R.D. 400, 403 (E.D.Pa. 1995) (finding proposed definition untenable because determining class membership would require mini-hearings on the merits of each case). An imprecise class definition, which does not give rise to presently ascertainable class members, undermines judicial economy and efficiency, thereby interfering with one of the primary purposes of class-action suits. See 5 Moore, supra, 23.02. Thus, the class definition must meet a minimum standard of definiteness for the trial court to proceed with certification. We note, however, that a class definition will not fail merely because every potential class member cannot be identified at the suit's commencement. See Carpenter v. Davis, 424 F.2d 257, 260 (5th Cir. 1970).

For a class definition to be objective and its members presently ascertainable, the definition cannot require a determination of the merits. Deciding the merits of the suit in order to determine the scope of the class or its maintainability as a class action is not appropriate. See Eisen v. Carlisle Jacquelin, 417 U.S. 156, 177 (1974); Sirota v. Solitron Devices, Inc., 673 F.2d 566, 572 (2d Cir. 1982); Irving Trust Co. v. Nationwide Leisure Corp., 95 F.R.D. 51, 62 (S.D.N.Y. 1982). By evaluating the merits at the certification stage, the trial court would allow a representative plaintiff to secure the benefits of proceeding with a class-action suit without first satisfying the requirements for maintaining one. See Eisen, 417 U.S. at 177. However, in determining whether the class-certification requirements have been satisfied, the trial court may look beyond the pleadings. See Falcon, 457 U.S. at 160; Coopers Lybrand v. Livesay, 437 U.S. 463, 469 (1978); Castano v. American Tobacco Co., 84 F.3d 734, 744 (5th Cir. 1996). Because class determinations generally involve considerations that are "enmeshed in the factual and legal issues comprising the plaintiff's cause of action," the trial court must be able to make a reasoned determination of the certification issues. Livesay, 437 U.S. at 469 (quoting Mercantile Nat'l Bank v. Langdeau, 371 U.S. 555, 558 (1963)); see Love v. Turlington, 733 F.2d 1562, 1564 (11th Cir. 1984) (holding that Eisen should not be construed to limit artificially the trial court's class-certification determination).

A proposed class definition that rests on the paramount liability question cannot be objective, nor can the class members be presently ascertained; when the class definition is framed as a legal conclusion, the trial court has no way of ascertaining whether a given person is a member of the class until a determination of ultimate liability as to that person is made. A fail-safe class that is based on resolving the ultimate liability issue is bound only by a judgment favorable to plaintiffs but not by a judgment favorable to defendants. See Dafforn v. Rousseau Assocs., Inc., 1976-2 Trade Cases 61, at 219 (N.D.Ind. 1976) (holding that because class would consist of only those homeowners who paid illegal fees, a jury determination that defendants did not charge illegal fees would mean there was no class); see also Adashunas v. Negley, 626 F.2d 600, 604 (7th Cir. 1980) (holding that class definition of children with learning disabilities who were not receiving adequate special education was deficient).

Certifying a fail-safe class inevitably creates one-sided results. If the defendant is found liable, class membership is then ascertainable and the litigation comes to an end. A determination that the defendant is not liable, however, obviates the class, thereby precluding the proposed class members from being bound by the judgment. We do not support such a result when "[r]ule [42] was never meant to be an exception to the rules of res judicata or to provide a risk-free method of litigation." Dafforn, 1976-2 Trade Cases 61, at 219. Moreover, a finding of no liability would logically mean that the class had been certified improperly because the class did not in fact exist. See Dunn v. Midwest Buslines, Inc., 94 F.R.D. 170, 172 (E.D.Ark. 1982).

In this case, the trial court abused its discretion because the class it defined is a fail-safe class. The trial court's class definition — those producers "whose natural gas was taken by the defendant in quantities less than their ratable proportions" — turns on the issue certified by the trial court — whether Intratex took nonratably from the class members. Class membership is not presently ascertainable because the certified issue of whether Intratex took ratably has yet to be decided. Under the trial court's definition, the parties will have to await the outcome of the litigation to determine who, if anyone, is in the class because whether Intratex took nonratably is the central issue to be determined at trial. If judgment is rendered for Plaintiffs, the various producers' ability to exercise their right to opt out of the class action will have been compromised because the judgment will bind those producers who were taken from nonratably. But if Intratex wins at trial, there is no class because its existence is dependent on a finding that Plaintiffs were taken from nonratably. Plaintiffs cannot be bound by such a judgment because without a class, there is no way to ascertain class membership. Certification of the class under these circumstances is clearly impermissible. Furthermore, while the trial court possesses significant discretion to modify the class definition or even decertify the class as the case develops, the trial court must still certify only those classes that are sufficiently defined and meet the rule 42 requirements. Because the class definition in this case is not precise, and its members cannot be ascertained until the alleged ultimate liability issue is decided, the trial court abused its discretion when it certified the class.

Having identified the problems with the trial court's class definition, the next question is whether this Court should rectify the definitional error or remand the case to the trial court for consideration. Noting the lack of federal case law on this question, Intratex argues that we should reverse the class-certification order and remand to the trial court without instructions. Plaintiffs, on the other hand, contend that it is well within this Court's authority to redefine the class because the record reveals that Plaintiffs, both in their pleadings and at the certification hearing, requested a precisely defined class of those producers excluded from the Dow-waiver agreements. Although some appellate courts have undertaken to redefine insufficiently defined classes, we decline to do so in this case. Here, the better course is to remand the action to the trial court for it to determine if the definitional problems can be eliminated.

Remanding for the trial court to consider redefinition is consistent with the trial court's discretion with regard to class certification and its responsibility to manage a class action. Cf. In re Corrugated Container Antitrust Litig., 643 F.2d 195, 225 (5th Cir. 1981) (holding that trial court's inherent power to manage class actions gives it discretion to decide when to review adequacy of settlements); see also 7B Wright, Miller Kane, Federal Practice and Procedure 1785 (2d ed. 1986). First, the class definition is "a matter within the broad discretion of the district court." Battle v. Pennsylvania, 629 F.2d 269, 271 n. 1 (3rd Cir. 1980). The trial court is also accorded discretion in determining whether to grant or deny a class certification. See Tana Oil Gas Corp. v. Bates, 978 S.W.2d 735, 740 (Tex. — Austin 1998, no pet.); American Express Travel Related Servs. Co. v. Walton, 883 S.W.2d 703, 707 (Tex.App.-Dallas 1994, no writ). In making a class-certification decision, the trial court in the first instance must determine whether a class exists, and must ensure that the class is properly defined, based on the available evidence. See 7B Wright, Miller Kane, supra, 1760. Appellate courts should be reluctant to redefine a fundamentally flawed class definition because the parameters of the new definition may not be evident from the record, and unlike trial courts, appellate courts do not have access to any additional relevant evidence that may facilitate evaluating the definition. Indeed, the recasting of the class definition by an appellate court raises concerns similar to those raised when an appellate court certifies a class that the trial court previously determined was not suitable for class status:

Courts of appeals are less likely to direct the certification of a class than to direct a decertification simply because it is easier to notice abusively exercised discretion, or legal error, in a certification already granted, than it is to pick one hypothetical class definition, from a myriad of possible definitions, and confidently conclude that one definition is so clearly the right one that the [trial] court abused its discretion in failing to certify that one class.

Armstrong v. Martin Marietta Corp., 138 F.3d 1374, 1386 n. 35 (11th Cir. 1998), cert. denied, 119 S.Ct. 545 (1998).

Second, under rule 42(c)(1), the trial court may alter, amend, or withdraw class certification at any time before final judgment. See Rio Grande Valley Gas Co. v. City of Pharr, 962 S.W.2d 631, 640 (Tex.App.-Corpus Christi 1997, pet. dism'd w.o.j.); Health Tennis Corp. of Am. v. Jackson, 928 S.W.2d 583, 587 (Tex.App.-San Antonio 1996, writ dism'd w.o.j.). For example, the contours of the case may change after discovery is completed and as the parties prepare for trial, necessitating modification of the class definition. See Richardson v. Byrd, 709 F.2d 1016, 1019 (5th Cir. 1983). Rule 42(c)(1) invests the trial court with the responsibility of managing the class action, and provides it with the tools to respond to changes in the case's development. Prescribing the class definition for the trial court, therefore, interferes with the trial court's discretion to monitor the class. Cf. Marisol A. v. Giuliani, 126 F.3d 372, 379-80 (2d Cir. 1997) (remanding to trial court to identify appropriate subclasses in the first instance with suggestions for possible subclasses). Thus the trial court's discretion to define, modify, subclassify, or decertify in response to the case's development counsels in favor of remanding to the trial court when an appellate court identifies definitional problems. See Richardson, 709 F.2d at 1019 (noting that "complex [class action] cases cannot be run from the tower of the appellate court given its distinct institutional role and that it has before it printed words rather than people").

We note that some appellate courts have redefined a class in those limited instances when correcting the definitional infirmity did not interfere with the trial court's discretion and oversight of the class action. Redefining a class in those circumstances furthers resolution of class-certification issues and conserves judicial and litigant resources by avoiding repeated and unnecessary appeals. However, we are not presented with such an opportunity here.

In this case, we disagree with Plaintiffs that this Court can easily recast the class definition from the record to include those producers excluded from the Dow-waiver agreements. We have already determined that the present class definition must be abandoned because it is fundamentally flawed. Plaintiffs propose a wholly new class. They acknowledge as much by highlighting that they presented two distinct certification theories to the trial court: one based on takings under the Dow-waiver agreements, and the other based on a common course of nonratable takings posited by the Rhodes study. That these theories are distinct counsels against appellate redefinition of the class.

The trial court defined a class based on the Rhodes study's identification of those producers who had been taken from nonratably. While the pleadings and the record of the class-certification proceedings mention the Dow-waiver program, the trial court and the parties focused primarily on the methodology and results of the Rhodes study at the certification hearing. As a result, the parameters of the proposed new class are not easily identified from the record. Thus, if we were to redefine the class, we would be assuming the trial court's discretion to define the class under rule 42.

Furthermore, the trial court on remand will still have to determine whether the newly defined class satisfies the rule 42(a) and (b)(4) requirements of numerosity, commonality, typicality, adequacy of representation, and predominance. That decision will require the trial court to resolve questions of fact, as well as legal issues, from this record or whatever additional evidence is developed in the trial court. For this Court to redefine the class in this case would therefore constrain the trial court by imposing on it a definition it would be foreclosed from changing, even if the proceedings on remand revealed a more appropriate class definition or if later case developments called for modification under rule 42(c)(1). In light of the record and the trial court's considerable authority to monitor this class action, including its discretion to certify, modify, or decertify the class if it becomes necessary, we cannot redefine the class. For these same reasons, we cannot decide in this case, as Intratex urges, whether attaining a precise class definition is futile.

Without a sufficiently defined class to bring this action, Plaintiffs cannot currently meet rule 42's prerequisites. Cf. Metcalf, 64 F.R.D. at 409 (holding that plaintiffs' attempts to define class were futile, therefore, they could not satisfy certification requirements). Therefore, we do not reach the parties' arguments concerning the enumerated requirements of rule 42(a) and (b)(4). Only with a properly defined class can the explicit class-certification provisions be examined appropriately. If, on remand, the trial court finds a suitable class definition, it must also ensure that the newly defined class complies with the requirements of rule 42(a) and (b).

Because the trial court abused its discretion when it certified the class, we reverse the judgment of the court of appeals and remand to the trial court for further proceedings consistent with this opinion.