In the
United States Court of Appeals For the Seventh Circuit ____________________ No. 19-1831 CHRISTINE DANCEL, Plaintiff-Appellant, v.
GROUPON, INC., Defendant-Appellee. ____________________
Appeal from the United States District Court for the Northern District of Illinois, Eastern Division. No. 1:18-cv-02027 — Ronald A. Guzmán, Judge. ____________________
ARGUED SEPTEMBER 16, 2019 — DECIDED DECEMBER 18, 2019 ____________________
Before BAUER, BRENNAN, and ST. EVE, Circuit Judges. ST. EVE, Circuit Judge. Is a person’s username a part of her identity? That is a complex question, but one that Christine Dancel proposes can be resolved categorically for all usernames and all people. The district court thought other- wise and declined to certify a class because it would have to be decided username-by-username whether each one is an as- pect of a given class member’s identity, at least as that word is defined by the Illinois Right of Publicity Act (IRPA), 765 2 No. 19-1831
ILCS 1075/5. Dancel contends this rejection of her theory was an improper decision on the merits of her and the class’s claims, and the court therefore abused its discretion at the class-certification stage. We see no such mistake in the district court’s reasoning and affirm the order denying certification. I. Background Groupon, Inc. is an online marketplace that sells vouchers for other businesses. Groupon’s website gives each business its own page with information about the business and the dis- counts available. Between April 2015 and February 2016, some visitors to the site could scroll down each page to see a “Photos” section that displayed up to nine pictures. If the vis- itor clicked a button, the page would reveal up to eighteen more. Groupon collected and displayed these pictures auto- matically using what it calls the “Instagram Widget.” As its name suggests, this Widget pulled publicly available pictures from the social networking service Instagram. It selected which pictures to assign to each page based on data linking the photos to the advertised business’s location. When a Groupon visitor hovered her cursor over a displayed photo, the Widget would show the unique, user-selected username of the Instagram account whose photo was being displayed and a caption, if the user had attached one to the photo. Dancel is an Instagram user, and like all Instagram users, her account had a username—namely, “meowchristine.” In 2015, Dancel uploaded to her account a picture of herself and her boyfriend visiting Philly G’s, a restaurant in Vernon Hills, Illinois. This picture was one of several Groupon displayed on Philly G’s page while the Widget was active. No. 19-1831 3
Dancel alleges that Groupon’s inclusion of her photo and username on Philly G’s page violated the IRPA, which pro- hibits the use of a person’s identity—meaning an “attribute of an individual that serves to identify that individual to an or- dinary, reasonable viewer or listener”—for commercial pur- poses without consent. 765 ILCS 1075/5, 30. She filed suit in the Circuit Court of Cook County and sought to maintain the action on behalf of a class of “Illinois residents” whose photo- graphs Groupon had similarly shown on its pages. The par- ties litigated in that posture for two years until Dancel moved to certify a different class, defined as “[a]ll persons who main- tained an Instagram Account and whose photograph (or pho- tographs) from such account was (or were) acquired and used on a groupon.com webpage for an Illinois business.” The class also had a sub-class: “All members of the Instagram Class whose likeness appeared in any photograph acquired and used by Groupon.” In response to these new class definitions, Groupon re- moved the case to federal court. Dancel tried to argue this re- moval came too late, but the district court disagreed, denied her motion to remand to state court, and proceeded to decide whether to certify her proposed class under Federal Rule of Civil Procedure 23(b)(3). That provision requires, among other things, “that the questions of law or fact common to class members predominate over any questions affecting only individual members.” Dancel identified the common ques- tion that united the class as “whether Instagram usernames categorically fall within the statutory definition of ‘identity.’” She distinguished this inquiry from the individualized ques- tion “whether any particular username identifies an individ- ual.” 4 No. 19-1831
The district court found Dancel’s categorical theory “un- persuasive.” Though the court accepted that a common ques- tion existed as to whether any username identifies an individ- ual, that question was not enough to certify a class because it “ignore[d] the individual inquiry that is the essence of deter- mining ‘identity’ under the IRPA.” The district court listed several of the usernames belonging to class members—e.g., eawhalen, artistbarbie, isa.tdg, loparse, johanneus—and con- cluded that it was “simply impossible to make any type of across-the-board determination as to whether these names ‘identify’ a particular person, as that term is defined by the IRPA.” Because it determined that the IRPA, as applied to the facts of this case, required a “username-by-username (photo- by-photo)” inquiry, the court found that common questions would not predominate over individual ones and denied cer- tification. Dancel petitioned for review of that decision, and we granted the petition. Fed. R. Civ. P. 23(f). II. Jurisdiction We start our analysis where we left off after our previous opinion: the district court’s jurisdiction. Dancel v. Groupon, Inc., 940 F.3d 381, 386 (7th Cir. 2019). The Class Action Fair- ness Act, 28 U.S.C. § 1332(d)(2), extends federal subject-matter jurisdiction to a proposed class action that is, among other things, minimally diverse—meaning one member of the plaintiff class is a citizen of a state different from any one de- fendant. Groupon is a Delaware corporation with its principal place of business in Illinois; it is therefore a citizen of those two states. Instead of alleging the citizenship of even one di- verse class member, Groupon asserted only that the class “un- doubtedly would include at least some undetermined num- ber of non-Illinois and non-Delaware citizens as class No. 19-1831 5
plaintiffs.” We agreed with Dancel that this allegation failed to carry Groupon’s burden and directed Groupon to amend its notice of removal. Dancel, 940 F.3d at 384–85. Though 28 U.S.C. § 1653 permits a party to amend jurisdictional allega- tions on appeal, see Heinen v. Northrop Grumman Corp., 671 F.3d 669, 670 (7th Cir. 2012), Groupon said it required discov- ery to make those amendments, so we ordered a limited re- mand for the district court to oversee that discovery and con- firm its jurisdiction. Dancel, 940 F.3d at 386. On remand, Groupon amended its allegations and supple- mented its notice of removal with three affidavits in which the affiants attested to their significant ties to their home states, which are not Illinois or Delaware. Groupon also provided the incorporation and registration documents for two Instagram- using businesses based in California and Missouri. Dancel ar- gued that these businesses and one of the affiants fell outside the class definition, but she conceded that the other two affi- davits belonged to putative class members domiciled in Vir- ginia and California, respectively. See Myrick v. WellPoint, Inc., 764 F.3d 662, 664 (7th Cir. 2014) (“Citizenship means domicile (the person’s long-term plan for a state of habitation) rather than just current residence.”). The district court saw no reason to conduct an evidentiary hearing to peer beyond Groupon’s allegations and Dancel’s concession (which was well sup- ported by Groupon’s evidence anyway), and neither do we. See Tilden v. Comm'r, 846 F.3d 882, 887 (7th Cir. 2017) (explain- ing that parties can agree to the fact of the parties’ citizen- ship—as distinct from the existence of jurisdiction—if the agreement is not collusive). With these amended allegations that at least one class member was a citizen of Virginia and another of California, Groupon has shown minimal diversity between it and the proposed class, and the district court’s 6 No. 19-1831
exercise of jurisdiction was proper. We may now move on to the class-certification decision. III. Class Certification Dancel argues that the district court improperly addressed the merits of her IRPA claim at the class-certification stage, and that, to the extent those merits are not off limits, the court resolved them incorrectly. She posits that the individual con- tent of each username is irrelevant to the class’s claims. In her view, an Instagram username is inherently an identity under the IRPA because it is unique—no two accounts can use the same username at the same time. Because this theory ignores the usernames’ content, Dancel maintains it cannot depend on the individual usernames of the class members but only on the common evidence describing the general function of In- stagram usernames and the Instagram Widget. She primarily insists that she is entitled to test her theory, regardless of its validity, on a class-wide basis because Rule 23 prohibits a court from resolving the merits of a claim at the class-certifi- cation stage. To the extent that the court can consider the va- lidity of her theory, she contends the district court imposed a higher evidentiary burden than the IRPA requires. We review the district court’s certification decision for an abuse of discretion, but a district court necessarily abuses its discretion if its reasoning is based on a legal error, like those Dancel asserts. See Beaton v. SpeedyPC Software, 907 F.3d 1018, 1025 (7th Cir. 2018), cert. denied, 139 S. Ct. 1465 (2019). We con- clude that the district court did not legally err in its interpre- tation of the Federal Rules of Civil Procedure or the IRPA and did not otherwise abuse its discretion. No. 19-1831 7
A. Rule 23(b)(3) A plaintiff seeking class certification bears the burden of proving that her proposed class meets the four requirements of Federal Rule of Civil Procedure 23(a), as well as those for one of the three types of classes identified in Rule 23(b). Bea- ton, 907 F.3d at 1025. We, like the district court, need not reach the Rule 23(a) elements, because we agree that Dancel has failed to meet those of Rule 23(b)(3)—the provision on which she relies. See Riffey v. Rauner, 910 F.3d 314, 318 (7th Cir. 2018), cert. denied sub nom. Riffey v. Pritzker, 139 S. Ct. 2745 (2019). Federal Rule of Civil Procedure 23(b)(3) requires a district court to find that “the questions of law or fact common to class members predominate over any questions affecting only indi- vidual members.” This distinction is often easier stated than applied: An individual question is one where “members of a proposed class will need to present evidence that var- ies from member to member,” while a common ques- tion is one where “the same evidence will suffice for each member to make a prima facie showing [or] the issue is susceptible to generalized, class-wide proof.” Tyson Foods, Inc. v. Bouaphakeo, 136 S. Ct. 1036, 1045 (2016) (al- teration in original) (quoting 2 W. Rubenstein, Newberg on Class Actions § 4:50 (5th ed. 2012)); Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 815 (7th Cir. 2012) (“If, to make a prima facie showing on a given question, the members of a proposed class will need to present evidence that varies from member to member, then it is an individual question. If the same evidence will suffice for each member to make a prima facie showing, then it becomes a common question.” (quoting 8 No. 19-1831
Blades v. Monsanto Co., 400 F.3d 562, 566 (8th Cir. 2005))). One way of establishing that common evidence can supply an an- swer to a common question is to consider whether an individ- ual class member could have relied on that same evidence in an individual action. Bouaphakeo, 136 S. Ct. at 1046. In resolving whether a common question exists or pre- dominates, the Supreme Court has emphasized that “Rule 23 grants courts no license to engage in free-ranging merits in- quiries at the certification stage.” Amgen Inc. v. Conn. Ret. Plans & Tr. Funds, 568 U.S. 455, 466 (2013); see also Messner, 669 F.3d at 811 (“[T]he court should not turn the class certification pro- ceedings into a dress rehearsal for the trial on the merits.”). Critically, “Rule 23 allows certification of classes that are fated to lose as well as classes that are sure to win.” Schleicher v. Wendt, 618 F.3d 679, 686 (7th Cir. 2010). Still the merits are not completely off limits. Rule 23 is more than “a mere pleading standard,” and the court must satisfy itself with a “rigorous analysis” that the prerequisites of certification are met, even if that analysis has “some overlap with the merits of the plain- tiff’s underlying claim.” Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350–51 (2011). We have described this analysis as involv- ing a “peek at the merits” that is “limited to those aspects of the merits that affect the decisions essential under Rule 23.” Schleicher, 618 F.3d at 685. The Supreme Court discussed the limits and extent of this peek in Amgen Inc. v. Connecticut Retirement Plans and Trust Funds, 568 U.S. 455. There the Court confronted a fraud-on- the-market securities class action. The underlying securities fraud claim required a plaintiff to prove, among other things, a material misstatement or omission and reliance on that mis- representation. Id. at 460–61. Ordinarily, reliance is a fact- No. 19-1831 9
intensive individual question that is not conducive to class- wide resolution. Id. at 462–63. Under the fraud-on-the-market theory, however, a plaintiff can demonstrate reliance by com- mon proof, on the inference that a security traded in an effi- cient market reflects all publicly available material infor- mation, so all buyers presumptively relied on false public statements when purchasing the security. Id. at 461–62. In or- der to certify a class under this theory, a plaintiff is required to affirmatively prove that the market is efficient and that an alleged misrepresentation was publicly made. Id. at 473. Not so with materiality, the Supreme Court held. The Court gave two reasons why a plaintiff does not need to establish materiality in order to certify a class but can wait to prove it at the merits stage. First, materiality is an objective question that can be proven through common evidence. Id. at 467. Second, the Court emphasized that “there is no risk what- ever that a failure of proof on the common question of mate- riality will result in individual questions predominating.” Id. at 467–68. This is so because materiality is an element of every plaintiff’s securities-fraud claim and not just a prerequisite to the class’s fraud-on-the market theory, as publicity and mar- ket efficiency are. Id. at 468, 473–74. “As to materiality, there- fore, the class is entirely cohesive: It will prevail or fail in unison. In no event will the individual circumstances of par- ticular class members bear on the inquiry.” Id. at 460. The parties agree that the IRPA's identity element is like materiality in that it, too, is an objective-inquiry—it asks whether an attribute would identify a plaintiff to the “ordi- nary, reasonable viewer or listener,” 765 ILCS 1075/5. Dancel concedes, however, that her theory is different from material- ity, in that if a class is certified and it is proven that a username 10 No. 19-1831
is not categorically an identity, Dancel and the class’s claims would “remain live,” Amgen, 568 U.S. at 474. Each person could still try to prove, after decertification, that his or her username is an identity in a way different from usernames generally. This distinction is not fatal to class certification, Dancel ar- gues, because this court has permitted certification even when there was a risk the court might later need to decertify the class. In Bell v. PNC Bank, N.A., 800 F.3d 360 (7th Cir. 2015), we affirmed a district court’s certification of a class alleging PNC had a policy requiring employees to work unpaid over- time. Id. at 372. We recognized that, if the policy were found not to exist then the class would be decertified, and the class members could still bring individual claims. Those individual claims, though, “would be based on an entirely different legal theory”—“Employee A alleging that her manager, Manager B, forced her to work off-the-clock.” Id. at 378. The class’s pol- icy-based theory would still “prevail or fail for the class as a whole.” Id. Similarly, in Costello v. BeavEx, Inc., 810 F.3d 1045 (7th Cir. 2016), we reversed the denial of certification to a class of couriers arguing that they were employees—not independ- ent contractors—of a delivery company, BeavEx. Id. at 1048– 49. Under state law, all workers were employees unless the employer could prove three elements, but the district court thought only the second element (whether the plaintiffs per- formed work outside the usual course of the employer’s busi- ness) was subject to common proof. Id. at 1059. The district court refused to consider whether this common question pre- dominated for fear of deciding the merits of the other ele- ments. Id. at 1060. We explained that even if only this single element was subject to common proof, that did not prevent certification. Id. If the plaintiffs won on that element, then all No. 19-1831 11
the couriers would be employees notwithstanding the other elements; if BeavEx won, then it would not have to relitigate the second element against each plaintiff. Id. Regardless of the way the evidence turned out, this common answer would rep- resent a significant aspect of the case. Id. Dancel’s theory is different from those in Bell and Costello because it presents a question that is common only if she is right. If the class in Bell lost on its theory, no member could argue that PNC had a policy only with respect to her over- time; nor could an individual courier in Costello argue that BeavEx’s usual course of business included delivery in his sit- uation but not the class’s. The question—does this policy ex- ist? is couriering BeavEx’s usual business?—was common re- gardless of its answer. If the common proof failed as to that answer, the plaintiffs could not use individualized proof in its stead. Any individual class member with a claim remaining would have to rely on different proof to answer a different question to establish the defendant’s liability. In contrast, the answer to Dancel’s proposed common question—are usernames categorically identities under the IRPA?—matters not only for the merits but for commonality. If the answer to that question is yes, then one significant as- pect of the case can be resolved in the class’s favor (if, as a factual matter, Groupon used the usernames within the meaning of the IRPA). But if usernames are not categorically an identity under the IRPA, and the court decertified the class, then the same element would remain entirely subject to dis- pute for each plaintiff. Nothing significant will have been de- cided, because each member could provide individualized evidence—her username’s content—on top of the common evidence to prove her username is an identity. Indeed, 12 No. 19-1831
because each username necessarily has some content, whether it be meowchristine or isa.tdg or loparse, the individ- ual claims would not be hindered by the preclusion of a con- tent-ignorant theory. Class members would be prohibited only from arguing again that all usernames are categorically an identity—that is, from certifying another class. This difference arises from the fact that Dancel’s proposed common question is whether identity is a common or an indi- vidualized question in the first place. Must identity be proven through “evidence that varies from member to member” (the username’s content) or can “the same evidence” (its being a username) “suffice for each member to make a prima facie showing”? Bouaphakeo, 136 S. Ct. at 1045. This is not a question that can be saved for after certification; it is precisely the ques- tion the court must answer at certification. As we said in Bell, “[a] proposed class of plaintiffs must prove the existence of a common question … but it need not prove that the answer to that question will be resolved in its favor.” 800 F.3d at 376. Like PNC, though, Dancel has “conflated [these] two inquir- ies,” id., by failing to separate her argument for why a com- mon question exists with an argument that the question should be resolved in her favor. A court cannot decide the lat- ter, but it must decide the former—and it must police this fine line or else reduce certification to a mere pleading standard. The Third Circuit faced a situation like this in Gonzalez v. Corning, 885 F.3d 186 (3d Cir. 2018). The plaintiff there alleged that the defendant’s shingles were defectively designed be- cause some would not last as long as advertised. Id. at 196. Whether any individual shingle would fail, though, could be determined only on review of that shingle. Id. The plaintiff came up with a theory to avoid this problem: the defect was No. 19-1831 13
not that the individual shingle would fail, but the chance that any shingle would fail. Id. at 197. This theory presented a com- mon, shared defect (the chance of failure) distinct from the in- dividualized defect (the failure), and the plaintiff insisted that, under Amgen, he could test the merits of his common theory on a class-wide basis. Id. at 200. The Third Circuit saw this request for what it was, though—an “attempt to circum- vent the need to identify a common defect by, in effect, rede- fining the concept.” Id. at 198. By checking the plaintiff’s su- perficially common theory for “coherence and legal founda- tion,” the district court had not found that any class member’s shingles were not defective and thereby decided the merits in violation of Amgen; it “merely applied Rule 23’s predomi- nance requirement.” Id. at 201. The same reasoning applies here. Dancel is trying to define the concept of identity in a common way so that it covers up individual questions that each class member might raise. Her use of the word “categorically” makes her theory universal, but only in the way “any question becomes universal when it includes the word ‘all.’” Id. This is not a disingenuous tactic under Rule 23—it is no different from how a securities-fraud class can paper over the individualized question of reliance with the fraud-on-the-market theory under which all buyers are presumed to rely on material public statements. See Amgen, 568 U.S. at 461. Whether the tactic succeeds, however, must be determined before class certification by considering the substantive law underlying the putative class’s claims. See Richard A. Nagareda, Class Certification in the Age of Aggregate Proof, 84 N.Y.U. L. Rev. 97, 104 (2009). The district court, then, was right to identify the starting point as “the substantive elements of plaintiffs’ cause of action 14 No. 19-1831
and … the proof necessary for the various elements.” Simer v. Rios, 661 F.2d 655, 672 (7th Cir. 1981) (emphasis added). If the pro- posed common proof could be enough to carry a prima facie case on an element for an individual plaintiff, in an individual case, then that element presents a common question that can predominate for a class action. Bouaphakeo, 136 S. Ct. at 1045– 47; Messner, 669 F.3d at 815; see also Costello, 810 F.3d at 1060 (“Plaintiffs have demonstrated that common questions pre- dominate by making out a prima facie claim … based on evi- dence common to the class.”). Here the present dispute is pre- cisely what evidence is needed to make a prima facie case for the identity element under the IRPA. Would facts common to all Instagram usernames (i.e., their innate uniqueness) suffice, or would a plaintiff need to show something more, something about her username, to meet her initial burden? Though this question overlaps with the merits, it is also “vital to any sen- sible decision about class certification,” Szabo v. Bridgeport Machs., Inc., 249 F.3d 672, 674 (7th Cir. 2001), and the district court was right to decide it before certifying a class. We must now determine whether the court correctly decided it. B. The IRPA We start with the substantive elements of an IRPA claim. The IRPA prohibits “us[ing] an individual’s identity for com- mercial purposes during the individual’s lifetime without having obtained previous written consent from the appropri- ate person.” 765 ILCS 1075/30(a). A plaintiff must thus prove (1) the appropriation of one’s identity, (2) without one’s con- sent, (3) for another’s commercial benefit. Trannel v. Prairie Ridge Media, Inc., 987 N.E.2d 923, 929 (Ill. App. Ct. 2013). The IRPA defines identity as “any attribute of an individual that serves to identify that individual to an ordinary, reasonable No. 19-1831 15
viewer or listener, including but not limited to (i) name, (ii) signature, (iii) photograph, (iv) image, (v) likeness, or (vi) voice.” 765 ILCS 1075/5. Name, in turn, “means the actual name or other name by which an individual is known that is intended to identify that individual.” Id. Dancel argues that all Instagram usernames are names or identities under the IRPA because an ordinary, reasonable viewer would know that a username is intended to identify one and only one user. Because a username is innately unique, Dancel contends, the viewer would not need to know what the username is to know that it identifies an individual. Dancel supports her categorical rule by comparing usernames to email addresses, but her analogy cuts against her. In United States v. Hastie, 854 F.3d 1298 (11th Cir. 2017), the Eleventh Circuit concluded that email addresses can qual- ify as “personal information” under the Driver’s Privacy Pro- tection Act, 18 U.S.C. § 2721, which defines the term as “infor- mation that identifies an individual.” Hastie, 854 F.3d at 1301 (citing 18 U.S.C. § 2725(3)). This was so, the court explained, because “[e]mail addresses often expressly include the ac- count holder’s name, affiliated organization, or other identi- fying information.” Id. at 1303. As is critical for our purposes, the court did not hold that email addresses categorically iden- tify an individual, only that they often did, and they often did so only because of their content, not their inherent nature as email addresses. See also Gov't Accountability Project v. U.S. Dep't of State, 699 F. Supp. 2d 97, 106 (D.D.C. 2010) (recogniz- ing that email addresses “can be identified as applying to par- ticular individuals” (emphasis added)). A username’s content, and not its nature, is likewise nec- essary to decide whether it is an identity under the IRPA. 16 No. 19-1831
Categorically, Instagram usernames identify only Instagram accounts. The IRPA, however, requires more than that. It de- mands that an attribute, even a name, serve to identify an in- dividual. And not just an individual but “that individual,” the one whose identity is being appropriated. This is a compara- tive exercise that depends on both the specific individual and the specific appropriated attribute in question and cannot be solved categorically here. The common evidence that Dancel proposes she will provide—that usernames are unique, the Instagram Widget collected usernames, and Groupon’s deal pages delivered those usernames to visitors—does nothing to answer the question whether any given username identifies that specific individual who is behind that username and its associated account. So, when Dancel insists that she is not ask- ing “whether any particular username identifies an individ- ual,” she gives up the game. Whether the appropriated attrib- ute serves to identify a particular person is, as the district court said, the “essence of determining ‘identity’ under the IRPA.” Dancel cannot answer this question for herself or for any putative class member with only her proposed common evidence, and so she cannot develop, for each class member, a common prima facie case under the identity element of an IRPA claim. Dancel insists that this individualized inquiry is incon- sistent with the IRPA’s broad scope, which is intended to pro- tect lay people’s identities. She argues that the content of a username could matter only if the person needed to know to whom the username belonged from just the username itself. The Illinois common law had just such a “notoriety” element, whereby the plaintiff needed to show her identity had “intrin- sic value,” like a celebrity’s might. Dwyer v. Am. Exp. Co., 652 N.E.2d 1351, 1356 (Ill. App. Ct. 1995). As Dancel rightly notes, No. 19-1831 17
though, the IRPA has supplanted the common law right of publicity. 765 ILCS 1075/60; Trannel, 987 N.E.2d at 928. So, a plaintiff need not show her identity has value to prove a claim under the IRPA. She still must show, however, that it is her identity that the defendant appropriated. The Appellate Court in Tranell v. Prairie Ridge Media, thus, did not ask whether a reasonable person would know who Karen Tranell, winner of the 2009 McHenry County Living gardening contest, was based only on her photograph. 987 N.E.2d at 926. It did, however, emphasize that the photograph “depict[ed] plaintiff and her daughter smiling at the camera,” and reviewed sev- eral other photographs for whether the subjects were recog- nizable. Id. An individual’s photograph, like her name, is ex- pressly protected by the IRPA, but that protection extends only so far as that photograph, that name, that username “serves to identify that individual to an ordinary, reasonable viewer.” 765 ILCS 1075/5 (emphasis added). Not all photo- graphs, names, or usernames will do so, and whether any does must be answered with individual proof beyond the cat- egory into which the attribute falls. Even if individual proof is necessary, Dancel proposes that she could provide it through class members’ affidavits attest- ing to their ownership of their usernames. We have permitted district courts, in their discretion, to use individual affidavits to identify class members after liability is established class- wide, Mullins v. Direct Dig., LLC, 795 F.3d 654, 672 (7th Cir. 2015), or to resolve individual merits questions after resolu- tion of class issues, Beaton, 907 F.3d at 1030. Dancel’s proposal goes much further than this. She asks to use class members’ affidavits not to clean up individual questions left over after some other, common question has been decided, but to pro- vide individual proof for the same question that she says can 18 No. 19-1831
be resolved through common proof. Given this contradiction, we need not decide whether it would ever be permissible to resolve the IRPA’s identity element this way (for example if she had identified some other, predominating common ques- tion). The district court here did not abuse its discretion. We emphasize that neither we nor the district court have decided the merits of any putative class member’s claim. Some Instagram usernames—maybe a great many of them— might qualify as identities under the broad definition in the IRPA. An ordinary, reasonable viewer might recognize that meowchristine serves to identify Christine Dancel, or that isa.tdg, artistbarbie, and loparse serve to identify their respec- tive users. We cannot say. What we can say is, under the IRPA, that ordinary, reasonable viewer would need to have evi- dence of which username, which account, and which person it was linking before it could make that decision. This individ- ualized evidentiary burden prevents identity from being a predominating common question under Rule 23(b)(3). IV. Conclusion We affirm the district court’s order denying class certifica- tion.