LOWELL v. LYFT INC (2026)

Plaintiffs Harriet Lowell and Westchester Disabled on the Move, Inc. (“WDOMI”) (collectively “Plaintiffs”), appeal from a judgment of the United States District Court for the Southern District of New York, entered following a bench trial before Philip M. Halpern, Judge, dismissing their claims that defendant Lyft, Inc. (“Lyft”), in its ridesharing business, discriminated against persons with mobility-related disabilities, in violation of the Americans with Disabilities Act (“ADA”), 42 U.S.C. § 12101 et seq., and the New York State Human Rights Law (“NYSHRL”), N.Y. Exec. Law § 290 et seq., by not making wheelchair accessible vehicles (“WAVs”)--i.e., vehicles that accommodate fixed-frame wheelchairs--available in all of the regions it serves, as it currently does in nine cities. The district court found principally that Plaintiffs failed either to show that the Lyft App's limited menu of choices constituted a barrier to Plaintiffs’ access to WAVs or to propose an effective modification for removal of such a barrier; that the evidence Plaintiffs produced at trial did not establish that their proposed modifications to Lyft's policy, practice, or procedure were likely to achieve effective WAV transportation; and that Lyft's proof at trial established that those proposed modifications would be neither effective nor reasonable. See Lowell v. Lyft, Inc., No. 17-CV-06251, 2024 WL 4350159 (S.D.N.Y. Sept. 30, 2024). On appeal, Plaintiffs contend that the judgment should be reversed, arguing principally that the district court erred in (1) imposing on them the burden of persuasion as to the effectiveness of their proposed modifications; (2) finding their evidence as to effectiveness insufficient; and (3) failing to require Lyft to prove its affirmative defense that making Plaintiffs’ proposed modifications would fundamentally alter the nature of its business. Finding no basis for reversal, we affirm.

I. BACKGROUND

Lowell is a citizen of White Plains in Westchester County, New York (“Westchester”), who relies on a motorized scooter for travel. WDOMI is a nonprofit community-based organization headquartered in Yonkers, New York, that advocates for individuals with disabilities in Westchester. Lyft is a business corporation that operates, in all 50 states, a peer-to-peer marketplace for on-demand ridesharing, offering access to a variety of transportation options, although not all options in all of the regions in which it provides service.

Plaintiffs commenced the present action against Lyft in 2017 on behalf of themselves and others similarly situated. The amended complaint, filed in 2017, asserted several causes of action under the ADA or state and local laws. In 2023, after several years of discovery and the dismissal of one claim on consent of the parties, the district court certified three plaintiff classes of persons “who require WAVs for vehicular transportation,” “who are denied equal access to Lyft's transportation services,” and who are “residents of or visitors to,” respectively,

[1] any and all regions serviced by Lyft, aside from Lyft's Access Regions or NYC ․ (the Non-Access Region Class) (represented by Lowell and WDOMI and asserting claims under the ADA)[;]

[2] any and all regions serviced by Lyft in New York State aside from NYC ․ (the New York State Other Than NYC Class) (represented by Lowell asserting claims under the ADA and NYSHRL and WDOMI asserting claims under the ADA); and

[3] Westchester County ․ (the Westchester Class) [(]represented by Lowell asserting claims under the ADA and NYSHRL and WDOMI asserting claims under the ADA).

Lowell v. Lyft, Inc., No. 17-CV-06251, 2023 WL 2622925, at *11-*12 (S.D.N.Y. Mar. 24, 2023) (other ellipses omitted) (emphases added).

The parties thereafter submitted a proposed joint pretrial order, approved by the court, containing, inter alia, a stipulation as to facts that were undisputed (“SF”); statements as to each side's positions with respect to expected exhibits, deposition testimony, and anticipated witnesses; and each side's proposed findings of fact and conclusions of law. Affidavits of certain witnesses accepted as direct testimony were eventually marked as joint exhibits (“JE”).

After a four-day bench trial, the district court issued an Opinion and Order dated September 30, 2024, see Lowell v. Lyft, Inc., No. 17-CV-06251, 2024 WL 4350159 (S.D.N.Y. Sept. 30, 2024) (“Lyft”), familiarity with which is assumed. Findings of fact after a bench trial “must not be set aside unless clearly erroneous.” Fed. R. Civ. P. 52(a)(6). The following summary is based primarily on the district court's findings of fact--none of which are challenged by Plaintiffs as clearly erroneous.

A. The Facts

Using Lyft's terminology for the descriptions of its operations and the services it offers, the district court found, inter alia, as follows.

“Lyft is a business corporation” that in 2012 “launched what it refers to as its peer-to-peer marketplace for on-demand ridesharing,” operating “multimodal transportation networks in the United States that offer access to a variety of transportation options through Lyft's platform and App” and that are “designed to address a wide range of mobility needs.” Lyft, 2024 WL 4350159, at *2 (quoting SF ¶¶ 6-9). Lyft's stated mission is to “improve people's lives with the world's best transportation.” Lyft, 2024 WL 4350159, at *2 (quoting SF ¶ 10) (internal quotation marks omitted).

Lyft's ridesharing platform is a two-sided technology platform premised on principles of supply and demand. (JE 9 ¶ 5). Members of the general public may download the Lyft App and agree to the Terms of Service. (SF ¶ 11). Drivers on the Lyft ridesharing platform must agree to the Terms of Service and Driver Addendum and meet applicable state and/or local regulatory vehicle and driver requirements. (Id. ¶ 12).

Lyft, 2024 WL 4350159, at *2. The Lyft platform attempts to match specific ride requests by passengers (e.g., from where, to where, at what time), with acceptance by qualified Lyft drivers who are able and willing to agree.

1. Lyft's Standard Service Mode

Lyft offers a variety of ride modes. Its “Standard mode” is offered on the App for ordinary cars, in contrast to, e.g., SUVs (offered in “XL mode”), luxury vehicles (offered in “Lux mode”), or WAVs (offered in “Access mode”). See Lyft, 2024 WL 4350159, at *3 n.6. “Lyft organizes its coverage areas into more than 300 geographic Regions nationwide,” “but does not offer every ride mode in every Region.” Id. at *3.

Lyft makes Standard mode available in all regions where it operates, regardless of population density, wait time, number of available vehicles, or any other considerations or metrics. ([SF] ¶ 17). Standard mode appears in every Region where Lyft operates as an available ride mode when a user inserts their destination address. (Id. ¶ 18).

Lyft, 2024 WL 4350159, at *3.

Lyft offers Standard mode solely through an independent contractor (“IC”) model. ([SF] ¶ 24). The IC model refers to drivers who personally own or rent a car and use Lyft's App to connect with riders. (Id. ¶ 25). These IC drivers set their own hours and drive where they want to drive. (Id.).

Lyft, 2024 WL 4350159, at *3.

When any mode (other than Access mode) is available on the Lyft platform in a Region, riders are able to request a ride in that “mode” at any time, 24 hours a day, but Lyft makes no guarantee that the ride request will be matched or that a driver will be available or accept the ride request. ([SF] ¶ 19).

Lyft, 2024 WL 4350159, at *3.

Depending on a user's location, the App displays information specific to that location, such as available ride modes, price, and availability. ([SF] ¶ 20). The App, in addition to ridesharing options, may show local information such as public transportation options, bike share, or rental car offerings. (Id.). Wait times (i.e., the amount of time between the time a ride is requested and the time when the driver arrives) and completion rates (i.e., the percentage of requested rides that are completed) on the Lyft platform vary by location and by mode. (Id. ¶ 21). Lyft does not guarantee any service levels or outcomes to its users, including wait times, completion rates, or that a requested ride will be completed. (Id. ¶ 22). Lyft does not restrict its Standard mode in any Region based upon the level of service, including wait times and completion rates. (Id. ¶ 23).

Lyft, 2024 WL 4350159, at *3.

2. Lyft's Access Mode

Lyft offers Access mode only in nine cities in the United States: Boston, Massachusetts; Chicago, Illinois; Dallas, Texas; Los Angeles, California; New York, New York (“NYC”); Philadelphia, Pennsylvania; Portland, Oregon; Phoenix, Arizona; and San Francisco, California (the “Access Regions”). And while it continues to offer that service in Dallas where it is no longer required to do so, it has initiated WAV service only “in Access Regions with regulatory requirements or where it is subsidized, either through a contractual obligation such as healthcare and transit partnerships,” or where it could participate in regulatory proceedings such as those before the California Public Utilities Commission. Lyft, 2024 WL 4350159, at *6.

“Lyft has never launched WAV service anywhere where it was not either required or subsidized to do so,” because of supply-chain issues. Id. The supply side of the Lyft platform has two facets: To meet demand for Access mode rides, there must be a sufficient number of WAVs, and there must be a sufficient number of WAV drivers. Each facet presents obstacles to efficient predictions as to the existence of adequate supply.

3. Supply of Vehicles

There is little information as to the number of WAVs that could be available to supply Access mode rides in Westchester.

WAVs are vehicles built to accommodate fixed-frame wheelchairs, and typically feature a wheelchair ramp or lift, a lowered floor to accommodate the equipment, and a securement device to keep the wheelchair in place when the vehicle is in motion.

Lyft, 2024 WL 4350159, at *3. “The number of” existing WAVs is “small, as the number of people who rely on WAVs is 1% or less of the population.” Id. at *30.

Moreover, the number of existing WAVs in any given region is difficult to ascertain, as WAVs are “not manufactured” by the car makers; “WAVs are created in the after-market by taking a standard minivan and modifying it to add accessible equipment.” Id.; see, e.g., id. at *9 (a trial witness whose business since 1952 has been “selling WAVs and manufacturing equipment to convert vehicles into WAVs,” was “unaware of any car manufacturer in the United States that manufactures vehicles with the wheelchair accessible sort of modifications already installed. (JE 7 at 2)”).

Thus, what comes from car manufacturer assembly lines are vehicles that then must be modified or altered to accommodate fixed-frame wheelchairs. See, e.g., Lyft, 2024 WL 4350159, at *9 (companies in the business of vehicle conversion charged from $16,000 “to $18,000 although the cost to convert a vehicle has increased in the last six months to approximately $25,000.”). “[T]here is no public database” showing how many vehicles have “been modified to become” WAVs. Id. at *30.

Although Plaintiffs’ expert witnesses expressed opinions that there were more than 500 WAV drivers driving into Westchester who could be available to accept requests for Access mode rides, the district court gave their estimates no weight because they lacked sufficient factual data, relied on inapposite experience, or were “predicated upon assumptions not verified by the creator of the underlying data.” Id. at *28; see also id. at *30, *32.

4. Supply of Drivers

The availability of sufficient Lyft drivers to supply Access mode rides in Westchester involves additional questions, and Lyft looks to multiple sources for drivers. As it does with respect to Standard mode, “the IC model for Access mode depends on drivers who personally own or rent WAVs.” Lyft, 2024 WL 4350159, at *3. “The IC WAV drivers, like other IC drivers on the Lyft platform, set their own hours and drive where they want to drive.” Id.

In an effort to “incentivize IC WAV drivers to accept WAV rides, Lyft pays a per-ride bonus to the drivers on top of their regular earnings for each WAV ride completed.” Id. at *6. Its success in recruiting IC drivers with WAVs has been limited. At one point in 2020,

Lyft sent a promotional text and in-app console card to drivers in the Philadelphia and Delaware markets, looking for drivers with WAVs who might be interested in driving on the Lyft platform in Philadelphia. (JE 2 ¶ 18). Lyft received, out of over 26,000 drivers requested, fewer than 130 responses and 30 referrals; and of that amount, only one driver ended up driving a WAV on the Lyft platform. (Id.).

Lyft, 2024 WL 4350159, at *6.

In addition to the IC model, for Access mode drivers Lyft uses what it calls a “W-2 model,” in which “Lyft pays a third-party company on a per-hour, per-vehicle basis regardless of the number of WAV rides requested. (JE 2 ¶ 21).” Lyft, 2024 WL 4350159, at *7. It uses the IC model and the W-2 model

to attempt to artificially create a supply of WAVs in the Access Regions. (JE 2 ¶¶ 15, 20, 22). Lyft, in some of the Access Regions, uses the IC model. (Trial Tr. at 56). This method also permits drivers to cross-dispatch, which means the vehicle can operate in multiple modes: the driver can match with a Standard mode user or an Access mode user. (Id. at 107). Lyft, in NYC for example, permits all IC drivers with WAVs to provide both WAV and Standard mode rides. (Id. at 111).

Lyft, 2024 WL 4350159, at *6 (emphasis added). In all Lyft regions where IC WAV drivers are used (e.g., NYC, Boston, Philadelphia, Chicago), they are allowed to cross-dispatch--of necessity, since in NYC, for example, only one-tenth of one percent of the rides are WAV rides. These drivers normally spend the bulk of their time driving Standard mode passengers and clustering in areas with the highest density of Standard mode demand; this reduces their availability to Access mode passengers. Id. at *20.

WAV service in other Access Regions is provided only through the W-2 model, such as San Francisco where W-2 drivers are a dedicated WAV supply and cannot cross-dispatch (i.e., the WAV drivers cannot offer rides in any mode other than Access mode). (Trial Tr. at 56, 107). The W-2 model is expensive because Lyft pays a third-party company on a per-hour, per-vehicle basis regardless of the number of WAV rides requested. (JE 2 ¶ 21). Under the W-2 model, Lyft can specify locations for the WAVs and can instruct its third-party partners to require drivers to drive solely in Access mode. (Id. ¶ 20). In other words, the W-2 model offers greater reliability than the IC model by guaranteeing a supply of WAV drivers. (Id. ¶ 21).

Lyft, 2024 WL 4350159, at *7.

Lyft also tried a third model in NYC, in which Lyft contracted with a rental car company to procure WAVs for rental to independent drivers (the “Rental model”). ([Trial Tr.] 187; JE 2 ¶¶ 22-23). The Rental model was in place in NYC from 2019 to September 2021 and Lyft determined that it did not work. (JE 2 ¶¶ 22, 25). Lyft is no longer using this model. (Id.). Lyft had to pay its rental partner a fee for every WAV (that was actually sitting idle) and realized as a result of an analysis performed, that the IC WAV drivers who rented their WAVs through a Lyft-sponsored rental program spent only 2% of their time providing Access mode rides--a result that did not work. (Id. ¶ 24).

Lyft, 2024 WL 4350159, at *7.

As discussed further below, Plaintiffs proposed numerous modifications to Lyft's operations in order to provide WAV transportation in Westchester. The district court did not find that any of them would likely be effective to accomplish that goal.

B. Issues on Appeal

As noted earlier, Plaintiffs on this appeal have not argued that any of the court's findings of fact are clearly erroneous. In addition, while Plaintiffs’ notice of appeal stated without limitation that they appealed the September 30, 2024 judgment that dismissed their claims after trial, their brief on appeal expressly abandons any challenge to several aspects of the judgment. First, although as described above, the court certified three classes of plaintiffs living in or visiting areas where Lyft did not offer Access mode service, and the claims of all three classes were pursued at trial, Plaintiffs inform us that they “are not appealing the district court's decision as it applies to the nationwide class or the New York subclass.” (Plaintiffs brief on appeal at 18 n.4.) Thus, only the class of Westchester residents or visitors remains at issue.

In addition, while Plaintiffs unsuccessfully sought to prove at trial that the Lyft platform contained “a discriminatory barrier” that “Lyft failed to remove,” Plaintiffs no longer pursue that contention. (Id. at 18 n.6.) Finally, “[a]t trial, the plaintiffs [also] proposed three ․ modifications” to Lyft's policy, practice, or procedure (in addition to five others dealt with by the court) that they do not pursue on appeal. (See id.)

With respect to Plaintiffs’ claim that Lyft discriminated against them by failing to make reasonable modifications of its policy, practice, or procedure, Plaintiffs’ five surviving proposed modifications--numbered 1-4 and 6 as Plaintiffs pursued them at trial--were to

(1) remove the current categorical preclusion (“blocker”) of WAVs on Lyft's platform in the Non-Access Regions and prevent implementation of Lyft's hidden Access mode toggle to view WAV options on the App's home screen;

(2) ask all drivers on its platform in the Non-Access Regions and those onboarding in the Non-Access Regions whether they have access to WAVs and whether they are interested in driving WAVs on Lyft's platform;

(3) utilize cross-dispatching (i.e., allow WAV drivers to receive and accept ride requests for different ride modes) for all drivers;

(4) implement prioritization logic for Access mode (i.e., whereby Lyft prioritizes dispatching WAVs to riders seeking WAVs to ensure efficient vehicle allocation);

․ [and]

(6) offer potential drivers baseline bonuses and incentives for guaranteed periods of time.

Lyft, 2024 WL 4350159, at *27 (emphases added). The court concluded that, in light of facts established by Lyft, none of Plaintiffs’ proposed modifications were reasonable.

II. DISCUSSION

The ADA, which prohibits discrimination against any individual “on the basis of disability in the full and equal enjoyment of,” inter alia, “specified public transportation services provided by a private entity that is primarily engaged in the business of transporting people and whose operations affect commerce,” 42 U.S.C. § 12184(a), defines “discrimination” to include

a failure to make reasonable modifications in policies, practices, or procedures, when such modifications are necessary to afford such ․ services ․ to individuals with disabilities, unless the entity can demonstrate that making such modifications would fundamentally alter the nature of such ․ services.

42 U.S.C. § 12182(b)(2)(A)(ii) (emphases added); see also id. § 12184(b)(2)(A) (incorporating § 12182(b)(2)(A)(ii) by reference). A defendant's failure to make a proposed modification does not constitute ADA-prohibited discrimination under this definition if the proposed modification was not both (a) “reasonable” and (b) “necessary” to accommodate the disability.

“An ineffective ‘modification’ or ‘adjustment’ will not accommodate a disabled individual's limitations.” US Airways, Inc. v. Barnett, 535 U.S. 391, 400, 122 S.Ct. 1516, 152 L.Ed.2d 589 (2002) (“Barnett”) (emphases in original). Thus, a proposed modification that would be ineffective would neither be necessary nor be reasonable. Even a proposed effective modification, however, may be found unreasonable because of other factors (including, as specified in the above definition, a defendant's showing that making such a modification would fundamentally alter the nature of its business). The issue of reasonableness plainly encompasses more than simply whether the proposed modification would be effective.

On appeal, Plaintiffs contend that the district court misapplied this Court's precedents with regard to an ADA plaintiff's burden to prove that a plausible accommodation exists and is facially reasonable, erred in finding Plaintiffs’ evidence insufficient, and erred in assessing the rebuttal evidence presented by Lyft. Finding Plaintiffs’ contentions unpersuasive, we affirm.

A. This Court's Precedents

Although Plaintiffs contend that the district court ignored or misapplied this Court's precedents as to the burdens of persuasion and production in disability cases, we see no error. The parties agreed prior to trial, in their Joint Proposed Findings of Fact and Conclusions of Law (“JPFC”), that

[p]laintiffs ‘bear[ ] the initial burdens of both production and persuasion as to the existence of an accommodation’ and the effectiveness of the modification.

(JPFC ¶ 85 (quoting Dean v. University at Buffalo School of Medicine & Biomedical Sciences, 804 F.3d 178, 189-90 (2d Cir. 2015) (“Dean”) (emphases ours), and citing Borkowski v. Valley Central School District, 63 F.3d 131, 138 (2d Cir. 1995) (“Borkowski”), and Staron v. McDonald's Corp., 51 F.3d 353 (2d Cir. 1995))). This stipulation referred to “effectiveness,” not reasonableness.

The district court quoted the parties’ agreement as to this standard, see Lyft, 2024 WL 4350159, at *24, and accepted it as “undisputed and supported by the record and law,” id. at *24 n.20. While Plaintiffs argue that it was error to accept that stipulation, we disagree. Although they proffer a citation to the record in an effort to show that they attempted--in their posttrial briefing to the district court--to disavow their pretrial stipulation that they had the burden of persuasion as to their proposed modifications’ effectiveness (see Plaintiffs reply brief on appeal at 10), Plaintiffs’ citation is to an argument that focuses not on effectiveness but on reasonableness (see Plaintiffs Posttrial Memorandum at 7 (describing Plaintiffs’ “burden of production to establish” an accommodation that is “reasonable”)). And while, as discussed below, a proposed modification must be effective to be reasonable, as our existing precedents show, the issue of reasonableness includes other considerations.

1. The Parties’ Evidentiary Burdens in a Disability Case

Neither the pertinent statutes nor the regulations implementing them specify which side bears the burden of demonstrating whether or not a requested accommodation or modification for a given disability is reasonable. See, e.g., Borkowski, 63 F.3d at 136 (noting that the Rehabilitation Act, 29 U.S.C. § 701 et seq., with disability provisions applicable to recipients of federal funds, and its implementing regulations “say nothing about who bears the burden of demonstrating that an accommodation is reasonable”); Roberts v. Royal Atlantic Corp., 542 F.3d 363, 369 (2d Cir. 2008) (“Roberts”) (“Neither the [ADA] nor the regulation specifies the allocation of burdens of production and persuasion between the parties in establishing whether a facility has been altered.”), cert. denied, 556 U.S. 1104, 129 S.Ct. 1581, 173 L.Ed.2d 675 (2009). Our cases have developed the following framework, based on the principle that in order to be reasonable, a proposed modification must at least be effective to accommodate or alleviate the disability.

In the employment context, where a plaintiff with a disability is seeking an accommodation--i.e., a modification of the normal job conditions--“the plaintiff bears the burden of production and persuasion on the issue of whether she is otherwise qualified for the job in question.” Borkowski, 63 F.3d at 137 (emphasis added). In other words, the modification proposed must be effective to counteract or ameliorate the disability. This is the cornerstone of the plaintiff's claim, and “[i]t follows that,” in proposing a modification, “the plaintiff bears the burden of proving” that the modification “permits her to perform the job's essential functions,” id. at 138, i.e., that the proposed modification would be effective. However,

[w]hether a proposed accommodation is reasonable ․ is another question. “Reasonable” is a relational term: it evaluates the desirability of a particular accommodation according to the consequences that the accommodation will produce. This requires an inquiry not only into the benefits of the accommodation but into its costs as well.

Id. (emphases added). Thus, while effectiveness is a necessary element of reasonableness, the assessment of reasonableness requires consideration of other factors as well.

As to the requirement that an accommodation be reasonable, we have held that the plaintiff bears only a burden of production. Gilbert [v. Frank], 949 F.2d [637,] 642 [(2d Cir. 1991)]. This burden, we have said, is not a heavy one. Id. It is enough for the plaintiff to suggest the existence of a plausible accommodation, the costs of which, facially, do not clearly exceed its benefits. Once the plaintiff has done this, she has made out a prima facie showing that a reasonable accommodation is available, and the risk of nonpersuasion falls on the defendant. Id.

At this point the defendant's burden of persuading the factfinder that the plaintiff's proposed accommodation is unreasonable merges, in effect, with its burden of showing, as an affirmative defense, that the proposed accommodation would cause it to suffer an undue hardship.

Borkowski, 63 F.3d at 138 (emphases added).

The analytical framework adopted in Borkowski, for the Rehabilitation Act claim involving a disability affecting employment, was used in Dean, 804 F.3d at 189-90, for an ADA claim involving a requested accommodation in the context of education, see, e.g., id., at 187 (the accommodation “provisions of the ADA and the Rehabilitation Act are generally equivalent”). In Dean, we noted that we had

not previously addressed the allocation of the burdens of production and persuasion with respect to establishing the third prong of a prima facie violation of the ADA or Rehabilitation Act--here the purported denial of a reasonable accommodation--in the education context․ [I]n employment-related claims based on a failure to accommodate, the plaintiff bears the initial burdens of both production and persuasion as to the existence of an accommodation that would allow the plaintiff to perform the essential functions of the position in question ․

Id. at 189-90 (emphasis added); see also id. at 191 (noting that the “trier of fact could find that Dean met his initial burdens of production and persuasion as to the existence of an accommodation” that could alleviate his disability (emphases added)).

In contrast, “as to the facial reasonableness of the accommodation,” the plaintiff has only

a “light burden of production” ․ [McBride v. BIC Consumer Prods. Mfg. Co., 583 F.3d 92,] 97 n.3 [(2d Cir. 2009)]; see Borkowski v. Valley Cent. Sch. Dist., 63 F.3d 131, 138 (2d Cir. 1995) (plaintiff satisfies “burden of production” by showing “plausible accommodation”). The burden of persuasion then shifts to the defendant to rebut the reasonableness of the proposed accommodation. McBride, 583 F.3d at 97 n.3. This burden of non-persuasion is in essence equivalent to the “burden of showing, as an affirmative defense, that the proposed accommodation would cause [the employer] to suffer an undue hardship.” Borkowski, 63 F.3d at 138; see Barnett, 535 U.S. at 402, 122 S.Ct. 1516 ․ (finding summary judgment in favor of defendant appropriately granted where a plaintiff fails to present evidence from which a jury may infer that an accommodation “seems reasonable on its face[)”] ․

․ Once the plaintiff has met the light burden of producing evidence as to the facial reasonableness or plausibility of the accommodation, the burden falls to the defendant educational-institution to persuade the fact-finder that the proposed accommodation is unreasonable.

Dean, 804 F.3d at 190 (emphases added).

In Roberts, we also used the Borkowski framework in dealing with ADA claims that wheelchair access was unreasonably denied in the context of existing buildings at the defendants’ resort. We disagree with Plaintiffs’ contention (see Plaintiffs reply brief on appeal at 5-6) that Roberts stands instead for the proposition that plaintiffs have only the initial burden of production on the issue of a proposed modification's effectiveness, not the burden of persuasion. If that had been intended, the opinion would surely have so stated; but the Roberts opinion contains no such statement. And as we read that opinion, the principal concerns were feasibility, regulatory constraints, and cost, with little question as to effectiveness. The opinion noted that the plaintiffs had proffered “plans--proposed themselves or with the aid of [an] independent architect--that would permit facially cost-effective wheelchair access to at least one of the pool areas,” including “a ramp leading to the pool.” Roberts, 542 F.3d at 377-78 & n.9. Whether or not such ramps--meeting “slope” and “width ․ mandate[s of] the [ADA],” id. at 378 (internal quotation marks omitted)--would be “cost”-effective (a component of the ultimate issue of reasonableness), we do not see that there was an issue as to whether such a ramp would be effective to provide wheelchair access to the pool (or to any area of the resort on a different level, such as the check-in desk, which was otherwise accessed by stairs). For individuals using wheelchairs, the presence of ADA-compliant ramps would clearly be a benefit.

Accordingly, Roberts had little need to discuss the obvious effectiveness of ramps to assist individuals needing wheelchairs. And, adopting the Borkowski framework, it thus principally discussed other components of the ultimate issue of reasonableness, along with the parties’ respective burdens with regard to that issue. It noted that a plaintiff has only a light burden to suggest a “plausible accommodation, the costs of which, facially, do not clearly exceed its benefits”; that the plaintiff would thereby have made “a prima facie showing that a reasonable accommodation is available”; and that “the risk of nonpersuasion” would then “fall[ ] on the defendant.” Id. at 370 (internal quotation marks omitted); see also id. at 371 (joining Borkowski’s view that placing the ultimate burden as to reasonableness on a defendant is warranted by defendants’ superior access to information about their own businesses, as well as about the needs and practices of their industry as a whole).

In sum, as indicated in Borkowski and Dean, and consistent with Roberts, plaintiffs who propose a modification have the burden of persuasion as to the modification's effectiveness; they have only a light burden of production--essentially a showing of plausibility--with respect to whether the modification they propose is reasonable. And if the plaintiffs meet these standards, “the burden falls to the defendant ․ to persuade the fact-finder that the proposed accommodation is unreasonable.” Dean, 804 F.3d at 190.

2. The District Court's Recognition of These Principles

These principles were reflected in the parties’ pretrial stipulations and in the district court's decision. The district court was well aware that effectiveness is a necessary element of reasonableness, and that assessment of reasonableness requires consideration of other factors as well. It stated that the “determination of whether a particular modification is ‘reasonable’ involves a fact-specific, case-by-case inquiry that considers, among other factors, the effectiveness of the modification in light of the nature of the disability in question and the cost to the organization that would implement it.” Lyft, 2024 WL 4350159, at *24 (other internal quotation marks omitted) (emphasis ours). We are unpersuaded by Plaintiffs’ contentions that the district court either erected too high a hurdle for what constitutes “effectiveness,” or placed on them the burden of proving reasonableness.

The parties had stipulated that “[a]n effective modification is one that accommodates Plaintiffs’ and the Class's disabilities.” (JPFC ¶ 86 (citing Barnett, 535 U.S. at 399-400, 122 S.Ct. 1516).) Contrary to Plaintiffs’ contentions that the district court tethered effectiveness to Lyft's own internal benchmarks or required the proposed modifications to result in better service for Access mode than for Standard mode in order to be effective (see Plaintiffs brief on appeal at 46-47), the district court correctly described the standard, stating that

[a]n effective modification is one that accommodates the plaintiff's disabilities, and affords access to the service allegedly denied. See US Airways, Inc. v. Barnett, 535 U.S. 391, 399-400, 122 S.Ct. 1516 (2002) (“An ineffective ‘modification’ or ‘adjustment’ will not accommodate a disabled individual's limitations.”).

Lyft, 2024 WL 4350159, at *24. And, accordingly, it stated that “[a]n ‘effective’ modification in the context of this action is one that w[ould] provide Plaintiffs with access to transportation by WAVs.” Id. at *27 (quoting PGA Tour, Inc. v. Martin, 532 U.S. 661, 688, 121 S.Ct. 1879, 149 L.Ed.2d 904 (2001)). What the court said Plaintiffs were required to “demonstrate[ ] by a preponderance of the evidence,” was “whether ․ the modification will result in effective WAV transportation based upon the basic principles of supply and demand which are the bedrock of the rideshare platform industry.” Lyft, 2024 WL 4350159, at *27 (emphases added).

B. Application of the Principles to This Case

1. Plaintiffs Failed to Prove Effectiveness

The district court assessed the effectiveness of all of Plaintiffs’ proposed modifications and was unpersuaded “that any of the[m] ․ would result in effective WAV transportation in Westchester.” Lyft, 2024 WL 4350159, at *35. As to Plaintiffs’ first proposed modification--that Lyft be required to have its Access mode icon displayed in Westchester--the district court found that Plaintiffs did not “establish how the option of merely displaying an icon in the App would create any service at all.” Id. at *30.

As to Plaintiffs’ second proposed modification--which consisted of “asking existing and onboarding drivers if they have a WAV,” id. at *14--the district court was unpersuaded that it would likely be effective to get WAV service for Westchester, for several reasons. For example, one expert's opinion endorsing that modification was based on his experience working at the New York City Taxi and Limousine Commission (or “TLC”). He testified that the TLC had acquired information as to how many WAVs existed in NYC by having its “licensing division”--as “a requirement for renewal of vehicle licensing and [for] new drivers being licensed by the TLC”-- “require[ ] ․ all drivers to identify whether they had a WAV.” Id. (emphases added). A successful poll by a private company cannot reasonably be predicted on the basis of the success of a poll by a municipal entity that has (and has threatened to exercise) the power to withhold needed licenses.

Moreover, the court noted that the number of existing WAVs is “small, as the number of people who rely on WAVs is 1% or less of the population”; and simply “asking the question will not create WAVs that do not otherwise exist.” Id. at *30.

Finally, the unlikelihood that such an inquiry by Lyft would gain either a significant amount of information as to the number of existing WAVs--or a significant increase in the supply of WAV drivers--was suggested by Lyft's attempt to obtain WAV-supply information in the Philadelphia-Delaware area. Lyft “sent to over 26,000 drivers” in that area “promotional messages [ ]asking drivers whether or not they owned a WAV.” Id. at *31 (citing JE 2 ¶ 18 (Lyft's solicitation offered “incentives, including referral and sign-on bonuses”)). Lyft received “fewer than 130 responses and 30 referrals,” Lyft, 2024 WL 4350159, at *6; and it managed to onboard only one WAV driver, id. at *31.

The court also found insufficient evidence that any of Plaintiffs’ other proposed modifications would have the effect of obtaining WAV transportation for Westchester--a conclusion apparently consistent with the views of the expert who had fashioned Plaintiffs’ requested modifications. When he was asked at trial by Plaintiffs’ counsel whether any proposed modifications other than the first and second were “essential to the provision of WAV service by Lyft,” he answered “No.” Id. at *16 (quoting Trial Tr. 429). He testified, “My other modifications ․ really have to do with efficiency” in terms of generally “allow[ing] Lyft to provide access mode service in a more efficient manner.” Lyft, 2024 WL 4350159, at *16 (quoting Trial Tr. 468). The district court was given no basis for finding that those supposedly “efficiency-promoting” proposals would be effective in acquiring WAV service for Westchester.

As to Plaintiffs’ contention that the district court improperly imposed on them the burden of persuasion on the issue of reasonableness in terms of their proposed modifications’ cost and burdens to Lyft, we are unpersuaded. First, Plaintiffs assert that the district court improperly required them to provide precise “detailed costs” with respect to their sixth proposed modification that would have required Lyft to offer potential WAV drivers baseline bonuses and incentives. (See, e.g., Plaintiffs brief on appeal at 5, 22, 26, 34, 46, 51-52.) In fact, however, the court had noted that “ ‘[t]he plaintiff's burden does not require him or her to furnish exact or highly detailed cost estimates,’ ” Lyft, 2024 WL 4350159, at *24 (quoting Roberts, 542 F.3d at 371 (emphasis ours)); and while noting what types of information could have been useful in this case, the court observed that, for Westchester, “Plaintiffs did not provide evidence of any cost,” Lyft, 2024 WL 4350159, at *33 (emphasis added).

Second, notwithstanding the district court's section title, “Plaintiffs Fail to Establish Reasonable Modifications,” id. at *25, we do not interpret the district court's discussion as placing the burden of persuasion on Plaintiffs with regard to the issue of reasonableness. With regard to “ ‘reasonable[ness]’ in terms of effectiveness, costs, and benefits,” the district court stated only that Plaintiffs “ha[d] not met their initial burden.” Id. at *27 (emphasis added). That “initial” burden as to reasonableness, as discussed above, is only to show that the proposed modification's “costs ․ facially, do not clearly exceed its benefits.” Borkowski, 63 F.3d at 138 (emphases added). And in every instance in which the district court referred to what Plaintiffs had “failed to prove” or “failed to establish,” the court was discussing the issue of one or more proposed modification's “effective[ness].” Lyft, 2024 WL 4350159, at *32 (“Plaintiffs have failed to prove that their third and fourth modifications would result in effective WAV transportation” (emphasis added)); id. at *33 (same with regard to Plaintiffs’ fifth, sixth, and seventh proposed modifications); see also id. at *1 (Plaintiffs failed to prove “that the proposed modifications which Plaintiffs seek would result in WAV transportation services in the Non-Access Regions” (emphasis added)); id. at *35 (“Plaintiffs have not established under the ADA or NYSHRL that any of the proposed modifications would result in effective WAV transportation in Westchester County, New York State, or across Lyft's 300 other Non-Access Regions.” (emphasis added)).

Plaintiffs also argue that the district court erred because it considered each of their proposed modifications in isolation, rather than in combination, but the record belies the premise. The court explicitly examined “[t]he third and fourth proposed modifications” together, noting that “[c]ross-dispatching goes hand-in-hand with prioritization logic.” Id. at *32 (internal quotation marks omitted). It found that those modifications would be ineffective not only when considered as a pair, but also when assessed in connection with Plaintiffs’ first (flawed) proposed modification (i.e., the proposal to require the Lyft App to display Access mode in Non-Access Regions, where it was not available). Id.

Given the district court's findings of discrete facts, we see no error in its ultimate finding that “Plaintiffs ha[d] not established” that implementation of these “proposed modifications would result in effective WAV transportation.” Id. at *35 (emphasis added).

2. Lyft Rebutted Effectiveness and Established Unreasonableness

Nor do we agree with Plaintiffs’ contention that the district court relieved Lyft of its burden of proof as to the ultimate issue of reasonableness. The court did not find that Lyft had simply presented evidence to support its contentions that the modifications proposed by Plaintiffs were unworkable or unreasonable; the court concluded that Lyft had “establishe[d]” that Plaintiffs’ “first proposed modification would be neither effective nor reasonable in cost.” Lyft, 2024 WL 4350159, at *29.

The court found that Lyft also “established” that it had been unable to provide Access mode service without “independently supplement[ing] the WAV supply to ensure that rides can take place, as an organic WAV supply needed to support a functioning platform does not otherwise exist,” id., and that the record “established that the number of WAVs are small,” and “established that WAVs are expensive and unique,” id. at *30; see also id. at *33 (Lyft also “established that [it] might have to start with as many as 20 contracted partner drivers if it were ordered to provide service in Westchester County alone, as it does in Los Angeles, a city close in square mileage to Westchester. (Trial Tr. at 124-125). [Lyft] also established that the costs of partnerships have ranged from $40-$65 per hour per vehicle. (JE 9 ¶ 22; JE 2 ¶ 21).”).

Thus, we see no merit in Plaintiffs’ contention that the district court did not determine whether Lyft rebutted their evidence of reasonableness for several of their proposed modifications. The court's findings as to the facts that Lyft “established” to show ineffectiveness, or unreasonableness as a whole, meant that the court found those facts existed more likely than not, implicitly finding that Plaintiffs’ proffered evidence for reasonableness was rebutted.

Finally, the court concluded that Lyft “established that” “[e]ven if access to an icon in the Lyft App constituted some benefit, ․ the risks and costs outweigh any such benefit[,] ․ includ[ing] reputational damage and the prospect of lawsuits, such as this one, which result when a service or product is held out to the public as available but ultimately is not and cannot be delivered.” Lyft, 2024 WL 4350159, at *35. Given the facts that the court found Lyft had “established,” we see neither in form nor in substance any indication that, in concluding that Plaintiffs’ proposed modifications were not reasonable, the district court had placed the burden of proving reasonableness on Plaintiffs.

In sum, plaintiffs who propose modifications to accommodate their disabilities have the initial burden and the burden of persuasion as to whether their proposed modifications would be effective; as to the multi-faceted issue of reasonableness, they have only the “initial” burden of identifying a “plausible” modification, i.e., a modification “the costs of which, facially, do not clearly exceed its benefits.” Borkowski, 63 F.3d at 138 (emphases added). “Once the plaintiff has met” these burdens, “the burden falls to the defendant ․ to persuade the fact-finder that the proposed accommodation is unreasonable.” Dean, 804 F.3d at 190 (emphasis added).

In this case, the factfinder was persuaded that Plaintiffs’ proposed modifications would not be effective to achieve WAV service. As there is no contention that any of the district court's factual findings are clearly erroneous, and we see no error of law, Plaintiffs have shown no basis for disturbing the judgment dismissing their claims. Given the district court's conclusion that Plaintiffs’ proposed modifications would not be effective to achieve WAV service, their proposals could not be found reasonable. Because the district court placed the proper evidentiary burdens on Plaintiffs and did not err in finding that Plaintiffs failed to meet these burdens, we need not address Plaintiffs’ argument that Lyft failed to prove its affirmative defense and demonstrate that making Plaintiffs’ proposed modifications would fundamentally alter the nature of its business.

CONCLUSION

We have considered all of Plaintiffs’ arguments on this appeal and have found them to be without merit. The judgment of the district court is affirmed.

KEARSE, Circuit Judge: