OPINION OF THE COURT

JORDAN, Circuit Judge.

In this mortgage loan discrimination case, a putative class of minority borrowers seeks permission under Rule 23(f) of the Federal Rules of Civil Procedure to appeal the denial of final approval by the United States District Court for the Eastern District of Pennsylvania of the parties' proposed settlement and certification of the settlement class. We will grant the petition for permission to appeal and, for the reasons that follow, will affirm the order of the District Court.

I. Background

Named plaintiffs John Rodriguez, Jennifer Worthington, Bobby Crouther, Jesus Conchas, and Rosa Maria Conchas (collectively, "Plaintiffs") are African-American and Hispanic borrowers who obtained mortgage loans from Defendant National City Bank in 2006 or 2007. On May 1, 2008, they filed a class action complaint against National City Bank and its parent company, National City Corporation (collectively, "National City"),1 alleging that National City had an established pattern or practice of racial discrimination in the financing of residential home purchases, in violation of the Fair Housing Act, 42 U.S.C. § 3605, and the Equal Credit Opportunity Act, 15 U.S.C. § 1691. Specifically, Plaintiffs asserted that National City issued them loans pursuant to a "Discretionary Pricing Policy" that allowed individual brokers and loan officers to add a subjective surcharge of additional points, fees, and credit costs to an otherwise objective, risk-based financing rate. According to Plaintiffs, as a result of that policy, minority applicants for home mortgage loans were "charged a disproportionately greater amount in non-risk-related charges than similarly-situated Caucasian persons." (J.A. at 117.) In other words, the policy allegedly produced a discriminatory disparate impact.

After the District Court denied National City's motion to dismiss,2 the parties engaged in extensive discovery. National City provided Plaintiffs with data on each of the more than two million loans it issued from 2001 to 2008. That data included, among other things, the annual percentage rate, the term of the loan, the interest rate, the prepayment terms, the origination fee, and the amortization type, as well as information about the borrower, including income, ethnicity, race, and debt-to-income ratio. While discovery was still proceeding, the parties met to explore the possibility of a negotiated settlement. Plaintiffs presented National City with preliminary statistical analyses of the loan data they had received. Although those analyses were shared confidentially and are thus not in the record, the parties agree that they included regression analyses of National City's loan data.3 Plaintiffs say that those regression analyses revealed that, overall, "Blacks and Hispanics paid more for their loans than similarly situated Caucasians (a `disparate impact') that amounted to damages ... of at least $350 and up to $1,100 per loan." (Petitioners' Opening Br. at 5.) Plaintiffs further contend that, because they controlled for "all objective credit and risk factors impacting loan pricing" (Id. at 12), those analyses prove that National City's Discretionary Pricing Policy produced the disparate impact.

After participating in two days of mediation, the parties arrived at a proposed settlement agreement. Under its terms, the class would include "[a]ll African-American and Hispanic persons who obtained a Mortgage Loan" from National City, its affiliates, or its successor-in-interest, PNC, from January 1, 2004, through the date of the settlement's preliminary approval. (J.A. at 250.) National City did not concede any wrongdoing, but it agreed to pay $7,000,000 for the benefit of the settlement class in exchange for a release of claims. Specifically, the agreement provided a service award of $7,500 to each of the named plaintiffs, $200 to each class payee, $75,000 to two organizations that would provide counseling and other services to the settlement class, and $2,100,000 in attorneys' fees. The agreement also included a provision barring either party from attempting to void the agreement, except in the event of an appeal.

On July 21, 2010, the District Court granted preliminary approval of the settlement and preliminarily certified the proposed class under Federal Rule of Civil Procedure 23(b)(3). Notice was then sent to the more than 153,000 members of the putative class. In response to that notice, six people objected to the proposed agreement, 66 opted out of the settlement, and 24,631 sought to take part in it by submitting claim forms. On December 9, 2010, Plaintiffs filed an unopposed motion requesting final approval of the settlement agreement, final certification of the settlement class, and attorneys' fees. In January 2011, after holding an initial fairness hearing, the District Court ordered additional briefing regarding certain aspects of the settlement agreement. Before the Court reached a final determination in light of that briefing, the Supreme Court issued its now well-known opinion in Wal-Mart Stores, Inc. v. Dukes, ___ U.S. ___, 131 S.Ct. 2541, 180 L.Ed.2d 374 (2011). The District Court ordered another round of supplemental briefing, this time asking the parties to discuss the impact of Dukes on class certification. In that briefing, both parties continued to support class certification, as they had promised in their settlement agreement.

The District Court, however, read Dukes as preventing certification, and, on September 8, 2011, it issued an order to that effect, denying at the same time Plaintiffs' motion for final settlement approval. In its memorandum opinion, the Court held that the settlement class failed to meet Rule 23(a)'s commonality and typicality requirements.4 It explained that Dukes had clarified the standard for establishing commonality, and that, under that standard, "Plaintiffs would likely have to show the disparate impact and analysis for each loan officer or at a minimum each group of loan officers working for a specific supervisor" in order to demonstrate commonality. Rodriguez v. Nat'l City Bank, 277 F.R.D. 148, 155 (E.D.Pa.2011). The regression analyses' demonstration of an overall race-based disparity was inadequate, the Court said, because, even if the analyses "remove[d] all credit related reasoning, there may be non-credit related reasoning that individual loan officers contemplated that is not based on race." Id. Accordingly, the Court decided that Plaintiffs had "fail[ed] to show that the class could be certified," and it denied their motion. Id. This timely appeal followed.

II. Jurisdiction and Standard of Review

The District Court had jurisdiction under 28 U.S.C. § 1331. The matter is before us on Plaintiffs' petition for leave to appeal, filed in accordance with Rule 23(f) of the Federal Rules of Civil Procedure, which allows us to "permit an appeal from an order granting or denying class-action certification ... if a petition for permission to appeal is filed with the circuit clerk within 14 days after the order is entered."

We have "very broad discretion in deciding whether to grant permission to pursue a Rule 23(f) appeal." Gutierrez v. Johnson & Johnson, 523 F.3d 187, 192 (3d Cir.2008); see also Fed.R.Civ.P. 23(f) advisory committee's note ("Appeal from an order granting or denying class certification is permitted in the sole discretion of the court of appeals."). In Newton v. Merrill Lynch, Pierce, Fenner & Smith, Inc., 259 F.3d 154 (3d Cir.2001), we identified several circumstances in which appellate review is appropriate, including: (1) "when denial of certification effectively terminates the litigation because the value of each plaintiff's claim is outweighed by the costs of stand-alone litigation"; (2) when class certification risks placing "inordinate... pressure on defendants to settle"; (3) "when an appeal implicates novel or unsettled questions of law"; (4) when the district court's class certification determination was erroneous; and (5) when the appeal might "facilitate development of the law on class certification." Id. at 164-65. By contrast, review is discouraged when the natural course of litigation will provide the moving party with an adequate remedy, or when the certification decision was routine and easily resolved. Id.

Permitting this appeal facilitates the development of the law on class certification by allowing us to consider the nature of the commonality inquiry in light of the Supreme Court's important instruction in Dukes. We therefore will grant Plaintiffs' petition and exercise our jurisdiction pursuant to Rule 23(f) and 28 U.S.C. § 1292(e).

We review a district court's decision to approve or reject a class action settlement agreement for abuse of discretion. Newton, 259 F.3d at 165; see also In re Hydrogen Peroxide Antitrust Litig., 552 F.3d 305, 310 (3d Cir.2008) ("The trial court, well-positioned to decide which facts and legal arguments are most important to each Rule 23 requirement, possesses broad discretion to control proceedings and frame issues for consideration under Rule 23.") A district court abuses its discretion if its decision "rests upon a clearly erroneous finding of fact, an errant conclusion of law or an improper application of law to fact." Marcus v. BMW of N. Am., LLC, 687 F.3d 583, 590 (3d Cir.2012) (internal quotation marks omitted). "Whether an incorrect legal standard has been used is an issue of law to be reviewed de novo." Id. (internal quotation marks omitted).

III. Discussion

Plaintiffs argue that the District Court abused its discretion in two ways: it contravened the "limited role" a court should occupy when deciding whether to certify a settlement class, and it based its commonality determination on an erroneous application of Dukes. National City, now free under the terms of the settlement agreement to object to class certification,5 contends that the Court occupied its prescribed role and reached the correct result under Dukes. We address those competing arguments in turn and conclude that the scope of the District Court's inquiry was fully consistent with Dukes, as well as with our own precedent and Rule 23, and that the Court rightly concluded that the putative class lacks commonality.

A. Certification of a Settlement Class

Federal Rule of Civil Procedure 23(a) requires that the members of a proposed class share a common question of law or fact. Fed.R.Civ.P. 23(a)(2). That commonality requirement is, along with numerosity, typicality, and adequacy of representation, one of Rule 23(a)'s four "threshold requirements" for class certification, Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 613, 117 S.Ct. 2231, 138 L.Ed.2d 689 (1997), which are intended to "limit the class claims to those fairly encompassed by the named plaintiff's claims," Dukes, 131 S.Ct. at 2550 (quoting Gen. Tel. Co. of Sw. v. Falcon, 457 U.S. 147, 156, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982)) (internal quotation marks omitted).6 Although they apply to all putative classes, those requirements are of "vital importance" in the settlement context because they protect absent class members "by blocking unwarranted or overbroad class definitions." Amchem, 521 U.S. at 620, 117 S.Ct. 2231. For that reason, the Supreme Court held in Amchem that, although certain Rule 23 considerations, such as "whether the case, if tried, would present intractable management problems," are not applicable in the settlement class context, the Rule's other requirements "demand undiluted, even heightened, attention" when class action representatives are seeking certification for the purpose of settlement. Id. at 620, 117 S.Ct. 2231. Thus, in addition to determining whether a proposed settlement is "fair, reasonable, and adequate," Fed.R.Civ.P. 23(e), district courts must ensure that each of Rule 23(a)'s requirements, including commonality, is satisfied before certifying a class and approving a class settlement agreement.7 See Sullivan v. DB Invs., Inc., 667 F.3d 273, 296 (3d Cir.2011) (en banc) ("[B]efore approving a class settlement agreement, a district court first must determine that the requirements for class certification under Rule 23(a) and (b) are met." (internal quotation marks omitted)); In re Cmty. Bank of N. Va., 418 F.3d 277, 300 (3d Cir.2005) ("[R]egardless of whether a district court certifies a class for trial or for settlement, it must first find that the class satisfies all the requirements of Rule 23.").

Plaintiffs agree that the requirements of Rule 23(a) remain intact in the settlement context, but they maintain that, "when a settlement class ... is presented for consideration," those requirements "operate in tandem with a strong presumption in favor of voluntary settlement agreements." (Petitioners' Opening Br. at 25-26.) They say the District Court disregarded that presumption and "undermin[ed] the well-established policy interest in promoting settlements" (id. at 28), by "delving into the merits of Class Members' claims" (id. at 24) and "speculat[ing] regarding evidence not on the record" (id. at 28). They argue that the Court should have occupied a more "limited" role in conducting its Rule 23 inquiry — one that respected the bargain the parties had struck in reaching a negotiated settlement.

Laying particular emphasis on Sullivan v. DB Investments, Inc., 667 F.3d 273 (3d Cir.2011) (en banc), and Ehrheart v. Verizon Wireless, 609 F.3d 590 (3d Cir.2010), Plaintiffs correctly assert that we have, on several occasions, articulated a policy preference favoring voluntary settlement in class actions. Sullivan instructed that assessing whether individual class members have viable claims is inappropriate in the context of reviewing a proposed settlement class because such an inquiry would "seriously undermine" our strong preference for settlement agreements. 667 F.3d at 311; see id. at 305 (explaining that "[a]n analysis into the legal viability of asserted claims is properly considered through a motion to dismiss ... or summary judgment..., not as part of a Rule 23 certification process"). In Ehrheart, we refused to vacate a settlement despite an intervening change in the law that eliminated the plaintiffs' underlying cause of action. 609 F.3d at 595-96. We concluded that the settlement agreement was a binding contract, and that permitting a party to "back out of an agreement at any time before court approval" would render the settlement process "meaningless." Id. at 594. Thus, we held the parties to their bargain, upholding our "strong judicial policy in favor of class action settlement." Id. at 595.

But while that policy is indeed strong, it cannot alter the strictures of Rule 23. The Supreme Court explained in Amchem that courts must be "mindful that the Rule as now composed sets the requirements they are bound to enforce," as the Federal Rules of Civil Procedure may be amended only through the "extensive deliberative process" Congress prescribed. 521 U.S. at 620, 117 S.Ct. 2231. Rule 23 has been amended to provide for settlement classes, but solely to add the additional hurdle of Rule 23(e), which mandates that the reviewing court find the settlement to be "fair, reasonable, and adequate." Fed. R.Civ.P. 23(e)(2). That amendment "was designed to function as an additional requirement, not a superseding direction" respecting the class-qualifying criteria of Rule 23(a) and (b). Amchem, 521 U.S. at 621, 117 S.Ct. 2231. The "dominant concern" of Rule 23 — that "a proposed class has sufficient unity so that absent members can fairly be bound by decisions of class representatives" — "persists when settlement, rather than trial, is proposed." Id. Therefore, whether class action representatives are seeking certification for the purpose of settlement or with the intent to litigate, the members of the proposed class must meet the threshold requirements of Rule 23(a), and our policy preference for voluntary settlement cannot and does not alter that demand.

In Sullivan and Ehrheart, we recognized the constant applicability of Rule 23. We said in Sullivan that, although settlement is clearly favored, "global settlements may nevertheless be rejected for failing to meet the requirements of Rule 23." 667 F.3d at 311 n. 40. In Ehrheart too we acknowledged that district courts have a responsibility to review settlements, and we reversed the district court not for any Rule 23 determination but for granting judgment on the pleadings after the parties had entered into a binding contract. 609 F.3d at 593 (describing the district court's role and explaining that the court in that case "never considered whether to approve the settlement"). In neither case did we excuse a failure to establish commonality, nor did we bar the district court from exercising its independent judgment in deciding whether the Rule 23 requirements were met. See In re Cmty. Bank of N. Va., 418 F.3d at 301 (requiring that a district court "exercise[] independent judgment" in the Rule 23 determination).

Furthermore, neither case lessened the burden required to demonstrate that putative class members share a common question of law or fact. As we have repeatedly stated, the Rule 23(a) requirements are "not mere pleading rules." Marcus, 687 F.3d at 591 (quoting In re Hydrogen Peroxide, 552 F.3d at 316) (internal quotation marks omitted). Rather, at the threshold, "[t]he party seeking certification bears the burden of establishing each element of Rule 23 by a preponderance of the evidence," which requires demonstrating "actual, not presumed, conformance" with the Rule. Id. (alterations and internal quotation marks omitted). The district court "must conduct a `rigorous analysis' of the evidence and arguments put forth," id. (quoting In re Hydrogen Peroxide, 552 F.3d at 316), a task that sometimes involves "a preliminary inquiry into the merits" of the plaintiffs' claims to ensure they can be "properly resolved as a class action," Newton, 259 F.3d at 168; see also Dukes, 131 S.Ct. at 2551 ("Frequently that rigorous analysis will entail some overlap with the merits of the plaintiff's underlying claim." (internal quotation marks omitted)).

Relying on Sullivan, Plaintiffs imply that a "rigorous analysis" is inappropriate in the context of a class action settlement. They note that Sullivan instructed courts not to delve into the underlying merits to determine if individual claims are viable. See Sullivan, 667 F.3d at 306 ("[T]he merits inquiry is particularly unwarranted in the settlement context...."). Ehrheart made similar statements, emphasizing the "restricted, tightly focused role that Rule 23 prescribes for district courts," 609 F.3d at 593, and holding that the district court abused its discretion by rescinding the parties' settlement agreement due to a change in the law that made plaintiffs' claims nonviable, id. at 595-96. But while both decisions advised courts not to assess whether plaintiffs' claims would be capable of succeeding if the case were to go to trial, neither limited the ability of district courts to consider the merits of a case when necessary for a Rule 23 determination. In fact, Sullivan specifically explained that "an examination of the elements of plaintiffs' claim is sometimes necessary, not in order to determine whether each class member states a valid claim, but instead to determine whether the requirements of Rule 23 ... are met." 667 F.3d at 306. Put more simply, our policy in favor of voluntary settlement does not alter the "rigorous analysis" needed to ensure that the Rule 23 requirements are satisfied. Dukes, 131 S.Ct. at 2551; see also Sullivan, 667 F.3d at 335 ("The same analytical rigor is required for litigation and settlement certification....") (Scirica, J., concurring).

Plaintiffs' other arguments regarding the proper role of the district court in settlement certification are similarly unavailing. They take particular issue with the District Court's alleged "conjecture" regarding evidence not in the record. (Petitioners' Opening Br. at 34.) Noting that "the parties in this case agreed to a settlement before the record was as developed as it would have been in a fully contested motion on class certification" (id. at 29), they argue that the District Court should not have "quibble[d] with and concentrate[d] on the quantity of evidence at settlement" because "[s]uch considerations do not apply in a settlement class" (id. at 30). They contend that by "speculat[ing]" about nondiscriminatory explanations for individual loan officers' decisions (id. at 28), the District Court improperly elevated the evidentiary showing needed for certification of a settlement class.

That argument misunderstands the burden of proof required for class certification. It is not enough that the parties agreed to settle and believed that "a more fully developed record would show that there were questions of law and fact common to all Class Members." (Id. at 29-30.) One cannot say, in effect, "we could show commonality, if we had to." The short answer is, "you do have to." See Marcus, 687 F.3d at 591 (holding that the party seeking class certification must demonstrate the putative class's conformance with Rule 23). That burden is not onerous. It does, however, require an affirmative showing that the class members share a common question of law or fact. Sullivan, 667 F.3d at 306. The mere possibility that evidence of commonality could have been produced does not satisfy that burden. Therefore, the District Court did not err by requiring actual evidence that the putative class satisfies the requirements of Rule 23(a).

Plaintiffs further contend that the District Court erred by "fail[ing] to fulfill its fiduciary role" to protect the interests of the unnamed members of the class. (Petitioners' Opening Br. at 26.) They argue that that role defines the scope of a district court's responsibilities in certifying a settlement class, and they imply that the District Court here acted beyond the scope of those responsibilities by denying class certification. Both contentions are incorrect. Although we have indeed highlighted the district court's role as a protector of absent members of the plaintiff class, see Sullivan, 667 F.3d at 319 ("[A] district court acts as a fiduciary for absent class members[.]" (internal quotation marks omitted)), our emphasis on that role has never been meant as a substitute for the requirements of Rule 23. Rule 23(a) explicitly requires that a putative class possess commonality, and courts are bound to enforce that requirement regardless of whether it benefits plaintiffs or defendants. In any event, there is no indication that the District Court's decision to decertify the class here failed to protect the interests of absent class members. When a class of plaintiffs does not share a common question of law or fact, it may well include individuals who did not actually experience the harm allegedly caused by the defendants. If that class is certified, those individuals will nonetheless partake in the recovery, which diminishes the relief for class members who actually were harmed. The Supreme Court stated in Amchem that the Rule 23(a) requirements are of "vital importance" in the settlement context precisely because they are "designed to protect absentees by blocking unwarranted or overbroad class definitions." 521 U.S. at 620, 117 S.Ct. 2231 (emphasis added). Thus, denial of certification to a class that lacks commonality falls squarely within the district court's prescribed role in considering the propriety of a settlement class.

Finally, Plaintiffs argue that the parties entered into their agreement knowing that Dukes might alter the legal landscape, and the District Court should have respected their decision "to settle and achieve certainty" rather than gamble on what the Supreme Court would decide. (Petitioners' Opening Br. at 33.) They again cite Ehrheart and Sullivan, this time for the proposition that the "choice to settle implicitly acknowledges calculated risks and, in the end, reflects deliberate decisions of both parties to opt for certainty in terminating their litigation." (Id. (quoting Ehrheart, 609 F.3d at 594) (internal quotation marks omitted).) See also Sullivan, 667 F.3d at 312 ("[A] district court's certification of a settlement simply recognizes the parties' deliberate decision to bind themselves according to mutually agreed-upon terms...."). According to Plaintiffs, "[t]he District Court's decision does ... injustice to the Parties' bargain" by "render[ing][the] settlement void, seemingly because the Parties had agreed to put a halt to this litigation before class, expert, factual and merits issues were fully litigated, an eventuality that the Parties consciously chose to avoid...."8 (Petitioners' Opening Br. at 35-36.)

Yet again, that argument fails because, as much as they might like to, parties cannot choose to avoid the judicial scrutiny demanded by Rule 23. Before approving the settlement of a class action, a district court must certify that the settlement comports with Rule 23 and is "fair, reasonable, and adequate." Fed.R.Civ.P. 23(e). Nothing less will suffice. It is true that we have advised courts not to "intrude upon the parties' bargain" after a settlement agreement is reached, Ehrheart, 609 F.3d at 593, but a denial of class certification does not constitute such an intrusion, as it is the result of a required inquiry that both parties had to have contemplated from the outset of their agreement. The parties in this case may well have considered in their bargaining that the Supreme Court had granted certiorari in Dukes, but that consideration does not insulate them from the District Court's responsibilities under Rule 23.

At base, Plaintiffs' argument regarding the proper role of the District Court seems to be that the Court was required to conduct its commonality review in a manner that did not upset the parties' settlement agreement. Such a review, though, is no review at all. The Rule 23 inquiry is certainly not meant to discourage settlement, but it is more than a rubber stamp, and thus it will sometimes result in the undoing of a settlement. The fact that it did so in this instance is therefore not in itself a basis for reversing the denial of class certification. The District Court stayed fully within its prescribed role and conducted the inquiry required of it by our precedent, by the Supreme Court, and by the Federal Rules of Civil Procedure.

B. The Commonality Determination

Because we conclude that the District Court properly fulfilled its prescribed role in conducting a Rule 23 review, we turn to the question of whether the Court rightly concluded that commonality is lacking in the class proposed in this case. A putative class satisfies Rule 23(a)'s commonality requirement if "the named plaintiffs share at least one question of fact or law with the grievances of the prospective class." Baby Neal v. Casey, 43 F.3d 48, 56 (3d Cir.1994). Again, that bar is not a high one. We have acknowledged commonality to be present even when not all plaintiffs suffered an actual injury, id., when plaintiffs did not bring identical claims, In re Prudential, 148 F.3d at 311, and, most dramatically, when some plaintiffs' claims may not have been legally viable, Sullivan, 667 F.3d at 305-07. In reaching those conclusions, we explained that the focus of the commonality inquiry is not on the strength of each plaintiff's claim, but instead is "on whether the defendant's conduct was common as to all of the class members." Id. at 299; see also In re Warfarin Sodium Antitrust Litig., 391 F.3d 516, 528 (3d Cir.2004) (focusing the commonality inquiry on the defendant's conduct, not "on the conduct of individual class members"); Newton, 259 F.3d at 183 (identifying common questions regarding the defendant's conduct); Baby Neal, 43 F.3d at 57 (considering whether the defendant "engag[ed] in a common course of conduct toward" the class members). In other words, there may be many legal and factual differences among the members of a class, as long as all were subjected to the same harmful conduct by the defendant. Baby Neal, 43 F.3d at 56.

In Dukes, the Supreme Court explained how the commonality standard applies when the complained-of conduct is a discretionary corporate policy that allegedly has a discriminatory effect. The putative class in that case consisted of "all women employed at any Wal-Mart domestic retail store at any time since December 26, 1998, who have been or may be subjected to Wal-Mart's challenged pay and management track promotions policies and practices." 131 S.Ct. at 2549 (alteration and internal quotation marks omitted). That enormous class of about 1.5 million women alleged that Wal-Mart's policy "allowing discretion by local supervisors over employment matters" produced a disparate discriminatory impact, evidenced by a statistical analysis of the company's employment information.9 Id. at 2547, 2554 (emphasis omitted). The Supreme Court concluded that that evidence was insufficient to establish commonality. While acknowledging that "giving discretion to lower-level supervisors can," in some circumstances, "be the basis of Title VII liability under a disparate impact theory," id. at 2554, the Court quoted Watson v. Fort Worth Bank & Trust, 487 U.S. 977, 994, 108 S.Ct. 2777, 101 L.Ed.2d 827 (1988), to emphasize that such claims must do more than "merely prov[e] that the discretionary system has produced a racial or sexual disparity" — they must also identify "the specific employment practice that is challenged," id. at 2555 (internal quotation marks omitted).10 Moreover, to bring a case as a class action, the named plaintiffs must show that each class member was subjected to the specific challenged practice in roughly the same manner. Dukes, 131 S.Ct. at 2555-56. The Dukes plaintiffs were all subjected to the discretion of their supervisors, but they had not demonstrated "a common mode of exercising discretion that pervades the entire company," id. at 2554-55, such that the policy could be considered a "uniform employment practice" that all members of the putative class had experienced, id. at 2554. Rather, the Dukes plaintiffs encountered different managers making different types of employment decisions for different reasons, many of them likely nondiscriminatory in nature. They therefore had not been subjected to a common harm, and the proposed class lacked commonality. Id. at 2555.

This case bears a striking resemblance to Dukes. Here, the class proposed by the Plaintiffs consists of "[a]ll African-American and Hispanic persons who obtained a Mortgage Loan" from National City between January 1, 2004 and the date the class was preliminarily certified. (J.A. at 250.) On behalf of those 153,000 class members, the named plaintiffs allege that National City's "Discretionary Pricing Policy" had the effect of charging African-American and Hispanic borrowers "a disproportionately greater amount in non-risk-related charges than similarly-situated Caucasian persons." (J.A. at 102, 117.) More specifically, Plaintiffs argue that National City granted brokers and loan officers the discretion to increase or decrease loan prices after an objective determination of loan eligibility, which discretion produced an overall disparate discriminatory impact. Therefore, in order to demonstrate that they have suffered a common harm, the putative class here must show that National City's grant of discretion to individual loan officers constitutes a "specific practice" that affected all the class members in the same general fashion. In other words, Plaintiffs must identify some "common mode" in which those brokers exercised their discretion. 131 S.Ct. at 2554.

Plaintiffs claim they have done so. They conducted regression analyses of National City's loan data, which they say demonstrate the Discretionary Pricing Policy's disparate impact even after controlling for legitimate factors affecting the price of loans.11 From what Plaintiffs characterize as "the objective nature of a loan pricing decision," they argue that, by "eliminat[ing] all objective credit and risk factors impacting loan pricing," they have shown that the only function the discretionary policy served was to produce a discriminatory effect. (Petitioners' Opening Br. at 12.) Therefore, they say, the regression analyses show that the loan officers' "common mode of exercising discretion," Dukes, 131 S.Ct. at 2554, was discriminatory.

But that conclusion is simply unsupported by the evidence. Even if Plaintiffs had succeeded in controlling for every objective credit-related variable — something no court could have reviewed because the analyses are not of record — the regression analyses do not even purport to control for individual, subjective considerations. A loan officer may have set an individual borrower's interest rate and fees based on any number of non-discriminatory reasons, such as whether the mortgage loans were intended to benefit other family members who were not borrowers, whether borrowers misrepresented their income or assets, whether borrowers were seeking or had previously been given favorable loan-to-value terms not warranted by their credit status, whether the loans were part of a beneficial debt consolidation, or even concerns the loan officer may have had at the time for the financial institution irrespective of the borrower.12 While those possibilities do not necessarily rebut the argument that the Discretionary Pricing Policy opened the door to biases that individual loan officers could have harbored, they do undermine the assertion that there was a common and unlawful mode by which the officers exercised their discretion.

Even assuming, however, that Plaintiffs had succeeded in identifying a specific employment policy that could be sufficiently distinguished from the discretionary policy in Dukes, they still have not shown that it affected all class members in all regions and bank branches in a common way. Another significant problem with the proposed class in Dukes was that the statistical disparity was based on an average of national data that was not necessarily representative of regional or store disparities. The Court explained that "a regional pay disparity ... may be attributable to only a small set of Wal-Mart stores, and cannot by itself establish the uniform, store-by-store disparity upon which plaintiffs' theory of commonality depends." Id. at 2555.

The proposed class in this case is also national, with 153,000 plaintiffs who obtained loans at more than 1,400 bank branches. As in Dukes, the application of the Discretionary Pricing Policy may have resulted in a disparity in some regions or branches but not at all in others. Accordingly, a very significant disparity in one branch or region could skew the average, producing results that indicate a national disparity, when the problem may be more localized. If the national disparity is not reflective of regional or even individual branch data, the putative class cannot show the policy affected each individual plaintiff in the same general fashion.

Plaintiffs contend that they controlled for regional differences in their regression analyses, but they must show that the putative class meets the commonality requirement by a preponderance of the evidence. In re Hydrogen Peroxide, 552 F.3d at 320. They did not introduce their data, regression analyses, or any other evidence to support a finding of commonality. Although Plaintiffs moved for class certification before the Supreme Court issued the Dukes opinion, the District Court requested the parties to submit briefs on class certification in light of the guidance given in that decision, and still Plaintiffs did not give the District Court a factual foundation for a commonality finding in their favor.13

Whether an appropriate foundation could be laid in a case like this is a question we leave for another day.14 We note, however, that, when faulting the Dukes plaintiffs for failing to account for regional differences that could undermine their claim of commonality, the Supreme Court went on to say: "There is another, more fundamental, respect in which respondents' statistical proof fails. Even if it established (as it does not) a pay or promotion pattern that differs from the nationwide figures or the regional figures in all of Wal-Mart's 3,400 stores, that would still not demonstrate that commonality of issue exists." Dukes, 131 S.Ct. at 2555. The Court then explained why, emphasizing that, as we have already noted, Watson requires that "the plaintiff must begin by indentifying the specific employment practice that is challenged." Id. (quoting Watson, 487 U.S. at 994, 108 S.Ct. 2777) (internal quotation marks and alteration omitted). "Other than the bare existence of delegated discretion," the Court observed, "respondents have identified no `specific employment practice' — much less one that ties all their 1.5 million claims together. Merely showing that Wal-Mart's policy of discretion has produced an overall sex-based disparity does not suffice." Id. at 2555-56.

Here, as in Dukes, the exercise of broad discretion by an untold number of unique decision-makers in the making of thousands upon thousands of individual decisions undermines the attempt to claim, on the basis of statistics alone, that the decisions are bound together by a common discriminatory mode.15 Plaintiffs therefore have not met their burden of demonstrating that the "defendant's conduct was common as to all of the class members," Sullivan, 667 F.3d at 299, and thus the District Court was correct to conclude that they do not share a common question of law or fact.16

IV. Conclusion

Because the putative class lacks commonality, the District Court did not abuse its discretion by denying the Plaintiffs' motion for final approval of the settlement and certification of the settlement class. Accordingly, we will affirm the Court's order.