dissenting.
The chief difficulty we confront in this case arises from the fact that some of the members of the class have not suffered the antitrust injury upon which this entire case is predicated. This percentage, while small, could constitute as many as 24,000 consumers29 who would have no valid *33claim against the defendants under the state antitrust laws even if the named . plaintiffs win on the merits.
The majority correctly recognizes that certification of a class that includes uninjured consumers hinges on there being a method of identifying and removing those consumers prior to entry of judgment, and that any such method must be both administratively feasible and protective of the defendants’ Seventh Amendment and due process rights. Op. at 19-20. The majority also correctly recognizes that the district court has not identified-much less rigorously analyzed-any method for identifying and excluding these thousands of consumers prior to entry of judgment. Op. at 20. Rather, the district court certified the class because it considered the Rule 23 predominance inquiry satisfied by the fact that the vast majority of consumers in the class had been injured. As for the uninjured, the court simply kicked the can down the road by noting that the court “preserve[d] the Defendants’ right to challenge individual damage claims at trial.” In re Nexium (Esomeprazole) Antitrust Litig., 297 F.R.D. 168, 179 (2013).
The path thus marked for our court is clear. We should vacate the order certifying a class that includes uninjured consumers, and remand to the district court to proceed in accordance with the principles set forth in the majority’s opinion. To the extent that certification remains relevant, given the posture of the case, the possibility would remain that the plaintiffs might yet propose and the district court approve some method of culling uninjured consumers from the class in an administratively feasible manner that protects defendants’ rights. Instead, the majority dons the hats of both plaintiffs’ counsel and the district court by first proposing, sua sponte, a culling method that no party has proposed — limiting recovery to consumers who file affidavits — and then announcing itself quite satisfied with that method. Op. at 21. I therefore respectfully dissent. By upholding the district court on the basis of a culling method that it itself has fashioned, the majority errs both on the merits and as a matter of appellate procedure.
First, on the merits of the majority’s proposed culling method, at least one sister circuit has twice noted the limitations of using affidavits in the manner proposed by the majority. See Carrera v. Bayer Corp., 727 F.3d 300, 304, 307 (3d Cir.2013) (remanding an order certifying a class of all purchasers of a weight-loss supplement in Florida where documentary proof of purchase was “unlikely” and noting that the method of ascertaining whether someone is in the class must be “administratively feasible” and that affidavits of purchase are not sufficient); Marcus v. BMW of N. Am., LLC, 687 F.3d 583, 594 (3d Cir.2012) (remanding a class certification order on the grounds that a class of original purchasers of BMWs with run-flat tires during the class period was not readily ascertainable via a “reliable, administratively feasible” method, and cautioning against including class members based on mere affidavits that their tires had gone flat).
The majority’s response to the persuasive force of this precedent is fashioned out of a vacuum. The majority cites Madison v. Chalmette Refining, LLC, 637 F.3d 551, 556 (5th Cir.2011), a case that both makes no mention of affidavits and actual*34ly reverses a class certification order because the district court failed to analyze in detail how individual issues would be resolved at trial, and instead took a “figure-it-out-as-we-go-along approach.”30 637 F.3d at 557 (quoting Robinson v. Texas Auto. Dealers Ass’n, 387 F.3d 416, 426 (5th Cir.2004)).
But regardless of whether or not affidavits may have a role to play in this or any class action, the larger issue is that a court of appeals should not assume that Rule 23 has been satisfied on the basis of a culling method that it itself has proposed. Many circuit court judges have little to no substantial experience with the nuts and bolts of class litigation, so fashioning litigation management devices is not in our institutional wheelhouse. Many of the facts relevant to assessing whether a certain management procedure will achieve a certain objective will not be discernible by a court until one party proposes it, and the other has a chance to critique it. At that point, the district court would usually weigh the pros and cons of the procedure and make a decision, employing the fair amount of discretion assigned to it. We, in turn, come along at or near the end of the process. And we are pretty good at explaining the principles that cabin the district court’s exercise of discretion, and at analyzing what by that point have usually become stationary targets presented in competing briefs.
The majority’s opinion, in contrast, skips all that. It simply assumes that the question of how uninjured consumers can be identified and excluded can be answered with affidavits.31 Untested by the adversary system, unexamined by any trial judge, and fashioned without awareness of its fit to the parties’ needs and goals, the majority’s method raises more questions than it answers. Will it require two forms of notice to class members-one to TPPs and one to consumers?- What happens to those consumers who do not return an affidavit (of whom there may be many, *35given the low dollar amount of any potential recovery)? Will they be deemed to have opted out of the class? Or will they be deemed to have remained in, but lost their claims due to lack of injury? Even more daunting, what happens if tens or hundreds of thousands of Nexium purchasers file affidavits? How exactly will defendants exercise their acknowledged right to “challenge individual damage claims at trial”? Will the defendants seek to depose everyone who has returned an affidavit, effectively challenging plaintiffs’ counsel to a discovery game of chicken? The majority simply hedges on these questions by assuming — without any basis at all, and likely unreasonably — that the affidavits will be “unrefuted.”
Throwing up an idea to see if it might stick is just not what courts of appeals do best. Rather, it is only after the adversaries have gone to the mat and the dust has settled that we can fairly review a district court’s assessment of whether a proposed method would be feasible. For this reason, if the district court does not identify a culling method to ensure that the class, by judgment, includes only members who were actually injured, this court has no business simply hoping that one will work. See Gen. Tel. Co. of the Sw. v. Falcon, 457 U.S. 147, 160, 102 S.Ct. 2364, 72 L.Ed.2d 740 (1982) (noting that “actual, not presumed, conformance” with the rule is “indispensable”); In re New Motor Vehicles Can. Exp. Antitrust Litig., 522 F.3d 6, 28 (1st Cir.2008) (requiring the district court to evaluate a proposed model for proving fact of injury prior to certification). In this important respect, any attempt to reconcile the majority’s holding with the approach taken by our circuit in New Motor Vehicles will result in hopeless confusion unless one concludes that the dissent in New Motor Vehicles has become the law without en banc review.
On a related note, I must also part company with the majority’s dalliance with a percentage-based rule inspired by the Seventh Circuit’s decision in Messner v. Northshore University HealthSystem, 669 F.3d 802, 826 (7th Cir.2012). The majority quite rightly says that the test for determining whether the inclusion of uninjured class members should defeat class certification is “functional.” Op. at 31. But then it backslides: it notes that a 2.4% decrease in the size of the class due to the removal of uninjured members was not so large as to defeat certification in Messner, 669 F.3d at 824, and concludes that the number of uninjured members here “seems comparable” to the number in Messner. Op. at 31.
If 2.4% is okay, why not 5.7%? Or any number under 50%? The percentage tells one almost nothing about the functional sufficiency of the method. The relevant inquiry for a court considering certifying a class that includes uninjured members is whether the court will be able to feasibly cull out those members before entry of judgment. It may be relatively easy to cull 5% out of a class of 30. Culling out 5% of 1 million is almost certainly not. Here, “just 2.4%” is likely to be at least 24,000 people. Moreover, nobody knows who the 24,000 are. So the culling process may need to review individually all the affidavits of class members who return them. How this is feasible, the majority does not explain.32
*36I also take issue with the majority’s suggestion that when a proposed class includes some uninjured members who will have to be removed post-certification, it is the defendants who bear the burden of demonstrating that it cannot be done. The Supreme Court has been clear that the party seeking certification bears the burden of demonstrating that the requirements of Rule 23 are satisfied. Wal-Mart Stores, Inc. v. Dukes, — U.S. -, 131 S.Ct. 2541, 2551, 180 L.Ed.2d 374 (2011). So, too, has this circuit. Smilow v. Sw. Bell Mobile Systems, Inc., 323 F.3d 32, 38 (1st Cir.2003).
The majority acknowledges this, Op. at 18, 26, and yet goes on to suggest the opposite: “Defendants have merely speculated that a mechanism for exclusion cannot be developed later. This is not enough to overcome plaintiffs’ case for having met the requirements of Rule 23.” Op. at 21. But plaintiffs have not met their burden because, as the majority acknowledges, the proposed class includes some number of uninjured members, Op. at 30, and the plaintiffs have not explained how they will be removed before judgment, Op. at 19-20. (It is also notable that the majority felt compelled to propose a culling method sua sponte — if the plaintiffs had indeed met their burden, this step would not be necessary.) When the plaintiffs have only shown that the number of uninjured members is relatively small, the class still cannot satisfy Rule 23 unless there exists a method for excluding those uninjured members prior to judgment. In such a context, it is no more the defendants’ burden to prove that this cannot be done than it is this court’s job to come up with a way that it can. Cf. Wallace B. Roderick Revocable Living Trust v. XTO Energy, Inc., 725 F.3d 1213, 1218 (10th Cir.2013) (vacating a certification order in part because the district court appeared to shift the burden to the defendant to prove lack of commonality).
Finally, it bears noting that in the time that this interlocutory appeal was pending, the district court tried most of the liability issues in this case, leaving the end payors’ fact-of-injury for a future proceeding. That trial concluded with a defense verdict just as these opinions were about to issue.' The district court solicited no affidavits from consumers, nor does it appear that there was a plan to do so. So even if the majority’s proposed culling method were tenable, we know that the district court did not employ it. In short, the majority affirms a certification order based entirely on a fiction that we know to be false. And unless one-way intervention is allowed, but see Am. Pipe & Constr. Co. v. Utah, 414 U.S. 538, 546-49, 94 S.Ct. 756, 38 L.Ed.2d 713 (1974) (discussing the history and application of the one-way intervention rule), it is likely too late to let class members self-identify after taking a peek at the verdict.
These changing facts on the ground warrant caution before affirming a class certification order based on a possibility that the district court might do something that it did not do, and which it is likely that it could not do. Will there be an appeal from the verdict that will succeed? Is there any plan to send notice?33 Is the basis for interlocutory review now eliminated? Although I agree entirely with the key principle that serves as the predicate for the majority’s opinion — that certification of a class that includes uninjured members is possible if the district court identifies a feasible method for culling *37those members prior to entry of judgment in a way that protects defendants’ rights— I do not believe that the majority has properly applied that principle in this case. I respectfully dissent.
.Neither side has precisely defined the size of the class, but the defendants, without challenge, suggest that it includes over a million consumers. Appellants' Br. at 20 (noting that 7.8% to 9.8% of the consumers in the class would constitute a group of over 100,000, meaning that the class would number more *33than a million). The majority’s careful analysis suggests, in turn, that the percentage of uninjured consumers may be comparable to the 2.4% in Messner v. Northshore University HealthSystem, 669 F.3d 802, 826 (7th Cir.2012). In a putative class including over a million consumers, that's at least 24,000 people.
. The majority gleans a potential blessing of affidavits from Chalmette by noting that among the criticisms of the district court by the Fifth Circuit was the failure to consider use of a so-called “Lone Pine " order, a device used in mass accident litigation to streamline a case. Lore v. Lone Pine Corp., No. L-33606-85, 1986 WL 637507 (N.J.Super.Ct. Law Div. Nov. 18, 1986). A Lone Pine order, in turn, can include a requirement that those willing to sue first produce "some evidence to support a credible claim," which may include affidavits from a physician or real estate appraiser as evidence of injury. See Steering Comm. v. Exxon Mobil Corp., 461 F.3d 598, 604 n. 2 (5th Cir.2006). Lone Pine orders are for mass accident cases, which the drafters of Rule 23(b)(3) recognized are generally not certifiable. Fed.R.Civ.P. 23 advisory committee’s note to subdiv. (b)(3) (1966) ("A 'mass accident’ resulting in injuries to numerous persons is ordinarily not appropriate for a class action because of the likelihood that significant questions, not only of damages but of liability and defenses to liability, would be present, affecting the individuals in different ways.”). So while a court might consider how a trial might be held in a mass accident case with use of a Lone Pine order, and by potentially requiring affidavits, nothing in Chalmette remotely suggests that affidavits would suffice as an administratively feasible tool for establishing injury in a manner protective of defendant’s jury trial rights.
. The majority also toys with the idea that courts could create a presumption that a consumer would buy a generic if it was available. Even assuming we could do so in a federal question case, but see 28 U.S.C. § 2072(b) (prohibiting the use of any procedural device to "abridge, enlarge, or modify any substantive right”), given that indirect purchasers cannot sustain an antitrust claim under federal law, see Illinois Brick Co. v. Illinois, 431 U.S. 720, 746-48, 97 S.Ct. 2061, 52 L.Ed.2d 707 (1977), the only time such a presumption could be employed is in a state-law diversity suit where a federal court is without authority to create such a presumption. See Erie R. Co. v. Tompkins, 304 U.S. 64, 78, 58 S.Ct. 817, 82 L.Ed. 1188 (1938).
. Oddly, the majority avoids any discussion of how affidavits will, as a practical matter, affect trial, even while citing Vega v. T-Mobile USA, Inc., 564 F.3d 1256, 1270 (11th Cir.2009) for the proposition that there is no predominance when the addition or subtraction of class members has a substantial effect on the quantity of the evidence offered. See Op. at 31.
. This is a Rule 23(b)(3) action in which notice to class members is mandatory. Fed. R.Civ.P. 23(c)(2)(B).