SVOBODA v. AMAZON COM INC LLC (2025)

Tanya Svoboda and Antonella Ortiz Colosi used a Virtual Try-On feature through Amazon's mobile site and app to virtually test facial products like lipstick and eyewear. They later brought this class action against Amazon alleging that the VTO feature violates the Illinois Biometric Information Privacy Act through its capture and use of their facial data. The district court certified a class of Amazon VTO users within Illinois. Seeing no abuse of discretion, we affirm.

I

A

Amazon's mobile website and application offer the VTO feature to customers to virtually “try on” makeup and eyewear products. The VTO software operates to render facial products over an image or video of a user's face so they can see what a product will look like before making a purchase decision.

Amazon used two VTO programs during the relevant period—one developed in-house and another developed by a company named ModiFace. Both programs function similarly. An Amazon customer shopping on a mobile device can preview a product by clicking a “try on” button on the product page. On the first VTO use, the application prompts the user to give Amazon permission to access the camera on their device. Once users grant access, they can opt to preview the product on a live video or actual photo of their face. They can also preview a product in “model mode,” which displays the product on a model's face rather than the user's. If users elect to preview on their own image, the VTO feature activates the camera on their device to capture their facial geometry from the live video or photo and then analyzes it to determine where to overlay the product. The VTO application sends information about the product to the device, which then renders the product on the user's image in the appropriate area.

Svoboda and Colosi's claims under the Illinois Biometric Information Privacy Act arise out of their use of Amazon's VTO feature. Biometric identifiers such as facial data and fingerprints are unique to each person and unalterable. If compromised, the privacy harm cannot be mitigated by changing them. The Illinois General Assembly enacted BIPA to strengthen biometric data security. See 740 ILCS 14/5; see also Fox v. Dakkota Integrated Sys., LLC, 980 F.3d 1146, 1149–50 (7th Cir. 2020) (explaining BIPA's history and purposes). To this end, the enactment regulates the “collection, use, safeguarding, handling, storage, retention, and destruction of biometric identifiers and information.” 740 ILCS 14/5(g).

Two provisions of BIPA are relevant here. Section 15(a) requires private entities in possession of biometric data to develop, publish, and comply with a written policy that establishes a data retention schedule and guidelines for destroying biometric identifiers and information. See 740 ILCS 14/15(a). Section 15(b), in turn, prohibits private entities from “collect[ing], captur[ing], purchas[ing], receiv[ing] through trade, or otherwise obtain[ing]” a person's biometric data unless they first: (1) provide written notice that biometric data will be collected or stored, (2) inform the person of the length and reason for which the biometric data will be collected, stored, and used, and (3) receive a written release signed by the person. 740 ILCS 14/15(b).

The parties agree that BIPA is not extraterritorial—any statutory violation must occur in Illinois to warrant relief. Illinois Supreme Court precedent gives us no reason to question that proposition. See Avery v. State Farm Mut. Auto. Ins. Co., 216 Ill.2d 100, 296 Ill.Dec. 448, 835 N.E.2d 801, 852 (2005) (observing that under Illinois law, a “statute is without extraterritorial effect unless a clear intent ․ appears from the express provisions of the statute.” (quoting Dur-Ite Co. v. Indus. Comm'n, 394 Ill. 338, 68 N.E.2d 717, 722 (1946))).

BIPA provides for statutory damages. “A prevailing party may recover” $1,000 for each negligent violation and $5,000 for each intentional or reckless violation. 740 ILCS 14/20(a)(1)–(2). A separate statutory violation generally accrues with “every scan” of biometric data. Cothron v. White Castle Sys., Inc., 466 Ill.Dec. 85, 216 N.E.3d 918, 929 (Ill. 2023), as modified on denial of reh'g (July 18, 2023). The statute also entitles successful plaintiffs to recover reasonable attorneys’ fees and costs. See 740 ILCS 14/20(a)(3).

The Illinois General Assembly amended BIPA in 2024 to cap statutory damages in certain circumstances. See Act of Aug. 2, 2024, Pub. Act 103-0769, 2024 Ill. Laws 6757, 6759–60. Section 20(b) now specifies that an entity that collects the same biometric data from the same person using the same method of collection in violation of Section 15(b) only commits a single violation. See 740 ILCS 14/20(b). Whether this amendment ultimately may cap the damages that some class members can recover is not a question before us, as the answer does not affect our review of the district court's class certification decision.

Tanya Svoboda alleges that she used Amazon's VTO in Illinois while shopping for lipstick on its mobile website in 2020. In September 2021, she filed a putative class action against Amazon in Illinois state court. Amazon removed the case to federal court under the Class Action Fairness Act. See 28 U.S.C. §§ 1332(d), 1453(b). Antonella Ortiz Colosi joined the action in an amended complaint in May 2022, alleging that she too used Amazon's VTO for makeup products while in Illinois. Colosi testified that she used the VTO in 2019, 2020, 2021, and 2022.

Together, Svoboda and Colosi allege that Amazon's use of the VTO feature violated Sections 15(a) and (b) of BIPA. More specifically, Amazon allegedly collected, captured, stored, and used their own facial geometry and associated personal identifying information and that of “thousands if not millions” of people who used the VTO feature in Illinois without providing notice and the required information, obtaining informed written consent, or creating written publicly available data retention and destruction guidelines. They seek statutory damages, injunctive relief, and fees and costs.

B

Svoboda and Colosi moved for class certification under Federal Rule of Civil Procedure 23(b)(3). The district court granted the motion and certified a class of “all individuals who used a [VTO] feature on Amazon's mobile website or app while in Illinois on or after September 7, 2016.”

The district court determined that the proposed class satisfied Rule 23(a)’s threshold requirements for certification. Only commonality is relevant to this appeal and, even then, only insofar as it relates to predominance under Rule 23(b)(3). The district court found that the same alleged course of conduct unites the alleged BIPA violations: Amazon's unlawful use of the VTO application. The court identified questions common to the class that, if answered, would resolve issues central to each claim “in one stroke.” Those questions generally concern the functionality of the VTO software, Amazon's use of the class members’ biometric data, and legal questions about whether that use violated Sections 15(a) and 15(b) of BIPA.

The district court also found that the class satisfied the predominance and superiority requirements of Rule 23(b)(3). After identifying common questions, it acknowledged that individual inquiries might be needed to assess whether potential class members used the VTO in Illinois. But it determined that this issue was manageable enough to justify certification without violating Amazon's due process right to challenge particular class members’ showing of a statutory violation or entitlement to a particular damages award. Nor did Amazon's proposed affirmative defenses defeat predominance because they either did not apply to BIPA claims or reflected common issues.

The district court further determined that a class action was superior to individual litigation because individuals would be unlikely to sue on their own and common issues made class resolution efficient. The prospect of high aggregate or individual damages under BIPA did not alter this conclusion because the district court observed that it retains equitable discretion to reduce unconstitutionally excessive damages.

Amazon timely filed an interlocutory appeal challenging the district court's class certification ruling. See 28 U.S.C. § 1292(e); Fed. R. Civ. P. 23(f).

II

Rule 23(a) imposes four requirements for class certification: numerosity, commonality, typicality, and adequacy of representation. See Fed. R. Civ. P. 23(a)(1)–(4). Once a class satisfies Rule 23(a), it must also fit into one of three types of class actions enumerated in Rule 23(b). The plaintiff bears the burden of showing that the proposed class satisfies Rule 23 by a preponderance of the evidence. See Eddlemon v. Bradley Univ., 65 F.4th 335, 338 (7th Cir. 2023). If the class meets all requirements, the district court must certify it. See id.

Svoboda and Colosi sought class certification under Rule 23(b)(3), which requires that “questions of law or fact common to class members predominate” over individual questions and that “a class action is superior to other available methods for fairly and efficiently adjudicating the controversy.” Fed. R. Civ. P. 23(b)(3).

Rule 23(b)(3)’s predominance requirement “builds on commonality.” Eddlemon, 65 F.4th at 338 (quoting Howard v. Cook County Sheriff's Off., 989 F.3d 587, 607 (7th Cir. 2021)); see also Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 623–24, 117 S.Ct. 2231, 138 L.Ed.2d 689 (1997) (“[T]he predominance criterion is far more demanding [than commonality].”). While commonality requires that common questions exist, predominance requires that they also predominate. See Eddlemon, 65 F.4th at 338. Predominance tests whether the class is “sufficiently cohesive” to justify class adjudication. Tyson Foods, Inc. v. Bouaphakeo, 577 U.S. 442, 453, 136 S.Ct. 1036, 194 L.Ed.2d 124 (2016) (quoting Amchem, 521 U.S. at 623, 117 S.Ct. 2231). Central to this inquiry is whether the claims “arise from a common nucleus of operative facts and issues.” Scott v. Dart, 99 F.4th 1076, 1092 (7th Cir. 2024) (quoting Beaton v. SpeedyPC Software, 907 F.3d 1018, 1029 (7th Cir. 2018)).

The Supreme Court has emphasized the rigor required in a proper predominance analysis. See Comcast Corp. v. Behrend, 569 U.S. 27, 33–34, 133 S.Ct. 1426, 185 L.Ed.2d 515 (2013) (observing that class certification requires “rigorous analysis” and that “Rule 23(b)(3)’s predominance criterion is even more demanding than Rule 23(a)”); see also Schroeder v. Progressive Paloverde Ins. Co., 146 F.4th 567, 573 (7th Cir. 2025) (emphasizing the same point); Eddlemon, 65 F.4th at 338 (same). Predominance does not entail a mechanical tallying of common questions against individual questions. Instead, a district court must identify the elements of each claim, the proof necessary for each, the manageability of those issues at trial, and alternative procedural tools. See Schroeder, 146 F.4th at 574; Simer v. Rios, 661 F.2d 655, 672–75 (7th Cir. 1981). From there the court must scrutinize the relationship between common and individual questions and assess their relative importance. See Tyson Foods, 577 U.S. at 453, 136 S.Ct. 1036.

Rule 23(b)(3) separately requires a showing of superiority. The plaintiff must demonstrate that a class action is “the most practical and efficient way to resolve the litigation.” Jacks v. DirectSat USA, LLC, 118 F.4th 888, 898 (7th Cir. 2024). The superiority inquiry is comparative: instead of considering the cost of a class action “in a vacuum,” a court must consider its costs and benefits in light of other available methods. Mullins v. Direct Digital, LLC, 795 F.3d 654, 663–64 (7th Cir. 2015).

To be sure, predominance and superiority analyses often overlap. Indeed, Rule 23(b)(3) provides a non-exhaustive list of considerations relevant to both:

(A) the class members’ interests in individually controlling the prosecution or defense of separate actions;

(B) the extent and nature of any litigation concerning the controversy already begun by or against class members;

(C) the desirability or undesirability of concentrating the litigation of the claims in the particular forum; and

(D) the likely difficulties in managing a class action.

Fed. R. Civ. P. 23(b)(3)(A)–(D); see Amchem, 521 U.S. at 615, 117 S.Ct. 2231 (stating that the list of factors in Rule 23(b)(3) is non-exhaustive).

III

Our review of a district court's class certification decision is deferential. We give the district court considerable leeway to exercise its discretion unless it commits legal error or makes clearly erroneous factual findings. See Jacks, 118 F.4th at 894–95.

Amazon limits its challenge to the district court's determination on predominance and superiority under Rule 23(b)(3). We take each in turn and, in the end, see no abuse of discretion by the district court.

We begin with predominance. The district court concluded that six common questions about Amazon's liability predominate over three individual questions: the locational element of a BIPA claim, affirmative defenses, and damages.

A

Recall that the parties start from a common point of legal agreement: class members must show that they were in Illinois when they used Amazon's VTO to establish a statutory violation of BIPA. The district court primarily analyzed this location requirement as an issue of class identification and manageability. This approach was incomplete. Location is relevant beyond just identifying class members. For the purposes of this appeal, we assume it is also a requirement of a BIPA claim. As a result, the district court needed to go a step further and assess whether proof of location generates an individualized question of fact that predominates over common questions. See Eddlemon, 65 F.4th at 339–40 (holding that the district court abused its discretion in finding that common issues predominated where it failed to identify and separately analyze the elements of the claims and conducted only a scant analysis); Santiago v. City of Chicago, 19 F.4th 1010, 1016–17 (7th Cir. 2021) (holding the same where the district court failed to identify the claims or discuss their elements and other clarity issues in the record complicated appellate review).

But this shortcoming was not fatal because the substance of the district court's reasoning touched sufficiently on all components required for a proper predominance analysis. Remand is therefore unnecessary, as it is clear from our own review that common questions predominate. See United States v. Nonahal, 338 F.3d 668, 671 (7th Cir. 2003) (“[R]emand is unnecessary because the district court's reasons ․ are apparent.”).

Location Presents an Individual Question

Sections 15(a) and (b) of BIPA both require proof that Amazon collected, captured, obtained, or possessed biometric data. As the district court recognized, this requirement, in turn, raises two common questions that may require expert discovery on the VTO software: (1) whether the facial data used by the operation of the VTO application constitutes biometric identifiers or information within the meaning of BIPA and (2) whether the use of the VTO means Amazon collected, captured, obtained, or possessed it. If the answer to both questions is yes, Amazon violated Section 15(a) if it failed to develop and publish a policy for data retention and destruction, with the latter inquiry itself presenting a common question. Amazon separately violated Section 15(b) if it failed to disclose information required by Sections 15(b)(1) and (b)(2) or failed to obtain signed written releases from VTO users per Section 15(b)(3). Those statutory requirements likewise present common questions.

Location—the geographic place from which a class member used the VTO—is the countervailing individual question. That observation follows from the evidence that would be used to prove location at trial. See Tyson Foods, 577 U.S. at 453, 136 S.Ct. 1036 (“An individual question is one where ‘members of a proposed class will need to present evidence that varies from member to member․’ ” (quoting 2 W. Rubenstein, Newberg on Class Actions § 4:50 (5th ed. 2012))).

Svoboda and Colosi proposed proving location with three sources of information: VTO user billing addresses, IP addresses and geolocation data, and personal affidavits from class members in which they attest to using the VTO application within Illinois. The record developed in the district court shows that Amazon maintains VTO usage data with accompanying billing address information associated with the user's account and, in certain circumstances, corresponding IP address and geolocation data. If a user is signed into an Amazon account or, in the case of the ModiFace VTO, the user adds the tested product to their virtual shopping cart, Amazon records the IP address associated with the VTO use.

A cybersecurity expert testifying for Amazon in the district court explained how IP addresses reveal (or can be analyzed to yield) geolocation information associated with VTO usage. IP addresses, the expert explained, contain indicators that roughly correlate with the location of the associated device. It also appears that Amazon contracts with a third-party provider to generate geolocation information, which includes an approximation of the state corresponding to the IP address and to the VTO usage data. It is not clear from the record at what level of completeness Amazon captures IP addresses. That fact matters because it bears on how closely the captured IP address information correlates with the location of the device using the VTO application. Based on our review of the record, what stands out is the testimony offered by an Amazon witness during a deposition. The witness stated that Amazon captures and maintains VTO usage records with full IP address information. This leads us to believe that, at least in some circumstances, reasonable findings about location information will follow.

What all of this means as a practical matter is that common proof of location including billing address and geolocation information will be available for some claimants. But, as the district court acknowledged, individualized inquiries will be necessary for others. For example, where billing address and geolocation data point to different states, or are unavailable for an alleged VTO use, individual affidavits or other proof will be necessary to show that the claimant used the VTO in Illinois.

Location Does Not Predominate Over Common Questions

The district court did not abuse its discretion in finding that individualized questions connected to proof of location will not predominate over common questions related to BIPA liability. But, if we accept that BIPA requires proof of location, not all elements of liability must be resolved on a classwide basis to satisfy Rule 23. Supreme Court precedent is clear on this point. See Amgen Inc. v. Conn. Ret. Plans & Tr. Funds, 568 U.S. 455, 469, 133 S.Ct. 1184, 185 L.Ed.2d 308 (2013) (“Rule 23(b)(3) ․ does not require a plaintiff seeking class certification to prove that each element of her claim is susceptible to classwide proof.” (cleaned up)). We have emphasized the same on many prior occasions. See, e.g., Beaton, 907 F.3d at 1029–30 (“[Defendant] misreads Supreme Court precedent in arguing that liability with regard to all class members must be resolved in a single stroke.” (citing Wal-Mart Stores, Inc. v. Dukes, 564 U.S. 338, 350, 131 S.Ct. 2541, 180 L.Ed.2d 374 (2011))); Bell v. PNC Bank, Nat'l Ass'n, 800 F.3d 360, 380–81 (7th Cir. 2015) (observing that Rule 23 does not require a class action to resolve all liability issues and discussing cases); Suchanek v. Sturm Foods, Inc., 764 F.3d 750, 759 (7th Cir. 2014) (holding that it was an “error of law” to conclude that the individual question of causation, “as required to establish liability” in a consumer action, predominated).

To the contrary, it is not uncommon for class actions to have a “final phase” for class members to submit individualized proof of a claim. Suchanek, 764 F.3d at 756. After proving a defendant's actions relevant to liability, each claimant must show that those actions harmed them. See id. at 757. A phase requiring individual presentations of proof on all (or part of) an element of a claim does not defeat predominance. Stated another way, an individual question does not predominate where common questions of law and fact relevant to liability otherwise generate significant efficiencies and the individual question is manageable. See Dukes, 564 U.S. at 350, 131 S.Ct. 2541 (“What matters to class certification ․ [is] the capacity of a class-wide proceeding to generate common answers apt to drive the resolution of the litigation.” (cleaned up)); see, e.g., Scott, 99 F.4th at 1092–93 (holding in a class action alleging inadequate dental care that common questions predominated even though individualized trials would be needed to prove causation and seek damages and not every class member will have suffered a denial of adequate care); Bell, 800 F.3d at 378–80 (holding the same in a class action alleging failure to pay overtime wages where each class member would need to individually prove they worked overtime and were entitled to relief) (discussing cases); Pella Corp. v. Saltzman, 606 F.3d 391, 394 (7th Cir. 2010) (holding the same in a product liability class action where each class member would need to individually prove causation and damages).

The district court did not abuse its discretion in concluding that the same reasoning applies here. One alleged course of conduct sufficiently unites the class's claims—Amazon's unlawful use of the VTO feature. And the entire class will “prevail or fail in unison” on common questions related to that conduct. Amgen, 568 U.S. at 460, 133 S.Ct. 1184. Some common questions require expert discovery to analyze the VTO software. None will benefit from repeated adjudication. And the potential class is estimated to be over 100,000 people. These factors point to class resolution as an efficient method of adjudication. See Suchanek, 764 F.3d at 760; Scott, 99 F.4th at 1092–93. If the common questions are all answered in favor of the class, it will have established a statutory injury. Even if location is necessary for individual recovery under BIPA, it is more characteristic of the individual questions permitted and often isolated in class actions than a core liability question based on Amazon's alleged statutory violations.

The district court's manageability analysis, though only part of predominance, is also relevant. See Rule 23(b)(3)(D); Simer, 661 F.2d at 677 (“[A]ny discussion of predominance must incorporate the concept of manageability because the issues for individual adjudication may be adequately managed with certain procedural devices.”). Our view on manageability parallels much of our reasoning on how we see class members proving the locational element. In short, the district court did not abuse its discretion in finding that the combination of billing addresses, IP and geolocation data, and affidavits make the locational element manageable.

No proxy alone may suffice or be perfect to prove location. Billing addresses may be out of date. Geolocation data has a margin of error, especially for cellular data, according to Amazon's cybersecurity expert. But the same expert reported that even the company's vendor's lowest estimate of geolocation accuracy correctly resolves a cellular IP address to a city 54% of the time within a 100-kilometer radius. On this record, we cannot say the district court abused its discretion in finding that this data, combined with billing addresses and affidavits, can prove a claimant was more likely than not in Illinois when they used the VTO. See Beaton, 907 F.3d at 1030 (observing that affidavits and auditing procedures can be used to handle individualized inquiries).

Amazon's challenge to predominance fares no better as a due process argument. The company will still have the opportunity to challenge class members’ individual proof of location. It can choose to challenge the reliability of the location data on the merits, contest the credibility of subgroups—such as class members for whom there is no geolocation data—or present evidence challenging particular affidavits. See id. (explaining that a defendant's due process rights are not harmed by case management tools for individual inquiries); Mullins, 795 F.3d at 669 (determining that due process does not require a “cost-effective procedure for challenging every individual claim”). But at this stage, individual questions about location neither predominate nor threaten the fundamental fairness due to Amazon. See Mullins, 795 F.3d at 669 (“[A] defendant has a due process right ․ to present individualized defenses if those defenses affect its liability.” (citing Dukes, 564 U.S. at 365–67, 131 S.Ct. 2541)).

B

Amazon's affirmative defenses also convince us that the district court did not commit any error at class certification. The company raised, and the district court rejected, four common law defenses as challenges to predominance—consent, failure to mitigate, estoppel, and waiver. As we understand it, Amazon's position is that these defenses require individualized inquiries to determine if they apply to each class member. For example, a class member who fully understood the VTO software but chose to use it anyway, so the company's argument goes, should not be able to recover for BIPA violations under a failure to mitigate theory or other defense. Individual application of these defenses may impact individual damages calculations.

Assuming that common law defenses can operate to preclude BIPA liability, which the parties contest and we have not decided, we see no error. Consent presents a common question because class members uniformly opt into the VTO by clicking the “try on” button. It is therefore appropriate for classwide resolution. See Gorss Motels, Inc. v. Brigadoon Fitness, Inc., 29 F.4th 839, 844 (7th Cir. 2022) (observing that prior express permission may be amenable to class-wide proof when given through uniform means).

Even if individual questions about the remaining defenses and damages remain, they do not defeat class certification, particularly where other common questions predominate. See Tyson Foods, 577 U.S. at 453–54, 136 S.Ct. 1036 (concluding that when one or more common issues predominate, class certification is proper even where individual questions about affirmative defenses and damages exist); Arandell Corp. v. Xcel Energy Inc., 149 F.4th 883, 887 (7th Cir. 2025) (same for damages); see also Riffey v. Rauner, 910 F.3d 314, 315, 317–19 (7th Cir. 2018) (holding that the district court did not abuse its discretion in denying class certification where the only issue remaining, damages, would require highly individualized inquiries).

Here, damages, and affirmative defenses bearing on measures of harm, are ancillary to the central question—liability—and can be handled at a separate phase of litigation. See 7AA Charles Alan Wright & Arthur R. Miller, Federal Practice and Procedure § 1778, Westlaw (database updated Sept. 2025) (“[I]f the main issues in a case require the separate adjudication of each class member's individual claim or defense, a Rule 23(b)(3) action would be inappropriate.” (emphasis added)); see, e.g., Gorss Motels, 29 F.4th at 844–45 (affirming denial of class certification where liability turned on an affirmative defense requiring individual resolution). Remember, too, that the district court has the flexibility to alter certification or decertify should damages become unmanageable. See Fed. R. Civ. P. 23(c)(1); Carnegie v. Household Int'l, Inc., 376 F.3d 656, 661 (7th Cir. 2004).

IV

In closing, we turn to Rule 23(b)(3)’s superiority requirement. Given that key common questions related to Amazon's liability predominate and will streamline this litigation, superiority is less likely to present a barrier to class certification. See Messner v. Northshore Univ. HealthSystem, 669 F.3d 802, 814 n.5 (7th Cir. 2012) (“There are so many common issues of law and fact relating to the issue of Northshore's liability ․ that the superiority requirement likely poses no serious obstacle to class certification here.”). And we agree with the district court that the possible size of aggregate and individual damages does not alter that calculus.

Svoboda and Colosi seek statutory damages for each violation of BIPA. They tell us that this means that each time a class member used the VTO feature, Amazon could be liable for up to another $1,000 or $5,000 depending on the nature of the violation. See 740 ILCS 14/20(a)(1)–(2), (b). Even if we credit Amazon's contention that the aggregate damages could extend into the tens of millions of dollars, that observation is not enough to defeat superiority. BIPA reflects a permissible judgment of the Illinois General Assembly that the misuse of biometric data amounts to a privacy violation warranting the imposition of money damages. See 740 ILCS 14/5. And sanctions may also deter future violations. To the extent that the total damages disproportionately outweigh any proven harm, the district court can reduce the award to ensure it is not unconstitutionally punitive. See Murray v. GMAC Mortg. Corp., 434 F.3d 948, 954 (7th Cir. 2006).

We recognize that high individual damages cut against the primary aim of classwide adjudication—to facilitate relief where claimants cannot or will not sue individually. See Amchem, 521 U.S. at 616–17, 117 S.Ct. 2231; Nagel v. ADM Inv. Servs., Inc., 217 F.3d 436, 443 (7th Cir. 2000); see also Fed. R. Civ. P. 23(b)(3)(A)–(B). This often occurs when the collective injury is great but individual damages are too small to incentivize any person to act alone. See Murray, 434 F.3d at 953. But Rule 23 does not bar class certification where individual damages are high. See Amchem, 521 U.S. at 617, 117 S.Ct. 2231 (observing that “the text of Rule 23(b)(3) does not exclude from certification cases in which individual damages run high”).

Other factors, including the lack of awareness of a claim, inability to access an attorney, or complex, costly discovery, can also hamper individual action regardless of the potential award. In these circumstances, a class may be a superior vehicle to vindicate individual injuries. See Mace v. Van Ru Credit Corp., 109 F.3d 338, 344 (7th Cir. 1997) (explaining that a plaintiff's awareness of their right to sue, willingness to take on the burden of litigation, and ability to find an attorney are relevant to superiority analysis); Suchanek, 764 F.3d at 760 (observing that the district court might find that “no rational individual plaintiff would be willing to bear the costs of this lawsuit” due to costly discovery including survey evidence and expert testimony); see also In re Facebook Biometric Info. Priv. Litig., 326 F.R.D. 535, 548 (N.D. Cal. 2018) (“While not trivial, BIPA's statutory damages are not enough to incentivize individual plaintiffs given the high costs of pursuing discovery on Facebook's software and code base and Facebook's willingness to litigate the case.”), aff'd sub nom. Patel v. Facebook, Inc., 932 F.3d 1264 (9th Cir. 2019).

The district court did not abuse its discretion in finding that Amazon VTO users lack sufficient incentive to bring individual suits. The plaintiffs reported having spent over $100,000 on expert discovery related to the VTO software. See Suchanek, 764 F.3d at 760 (observing that “[common] class issues often will be the most complex and costly to prove”). And the absence of any meaningful individual VTO-related BIPA litigation against Amazon to date is also telling. See Fed. R. Civ. P. 23(b)(3)(B). We see no infirmity in the district court's conclusion based on these observations that affected users may be unable or unwilling to undertake this costly litigation alone without any guarantee of recovery.

Even if some individuals eventually sue, a class action may still proceed. The superiority requirement entails a balancing analysis. See Mullins, 795 F.3d at 663–64. And as the district court observed, this class action will resolve virtually identical claims for potentially thousands of people. It will ensure uniform resolution of complex issues about the operation of the VTO software, preserve judicial resources, and promote efficiency. We agree with the district court that this suffices to show superiority.

Finally, it warrants emphasis that class certification does not carry any permanency. The district court should remain vigilant in monitoring the propriety of certification as the case develops.

For these reasons, we AFFIRM.

Scudder, Circuit Judge.