JOHN M. WALKER, JR., Circuit Judge:

This case presents two distinct questions that arise from the transmittal of musical works over the Internet: First, whether a download of a digital file containing a musical work constitutes a public performance of that musical work; and, second, whether the district court, acting in its capacity as the rate court, was reasonable in its assessment of the blanket license fees of Yahoo! Inc. and RealNetworks, Inc. (collectively, "the Internet Companies") to publicly perform any of the millions of musical compositions in the American Society of Composers, Authors and Publishers ("ASCAP") repertory.

For the reasons set forth below, we affirm the district court's ruling that a download of a musical work does not constitute a public performance of that work, but we vacate the district court's assessment of fees for the blanket ASCAP licenses sought by the Internet Companies and remand for further proceedings.

BACKGROUND

I. FACTS

The Internet Companies seek separate blanket licenses to publicly perform the entirety of the ASCAP repertory for certain of their websites and services. A blanket license is a license that gives the licensee the right to perform all of the works in the repertory for a single stated fee that does not vary depending on how much music from the repertory the licensee actually uses. United States v. Am. Soc'y of Composers, Authors & Publishers, No. 41-1395(WCC), 2001 WL 1589999, at *1 (S.D.N.Y. June 11, 2001). ASCAP licenses the non-dramatic, public performance rights in copyrighted musical works. More than 295,000 composers, songwriters, lyricists, and music publishers in the United States participate exclusively in licensing their music through ASCAP. ASCAP licenses approximately 45% of all of the musical works that are played on-line.

The Internet Companies perform music in myriad audio and audio-visual contexts. Yahoo! provides music content in various ways across its website. For example, a user can enjoy the specific song or music video he desires from an "on-demand" stream in Yahoo! Search, listen to a radio-style webcast in Yahoo! Music, view audio-visual clips from movies and television shows in Yahoo! Movies and Yahoo! TV, or upload and share his own videos using Yahoo! Video.1 However, only a small portion of the activity on Yahoo!'s website involves performances of musical works, and not all of the areas on Yahoo!'s website offer audio or audio-visual content.

RealNetworks performs music in audio and audio-visual contexts through a number of websites and subscription services.2 Like Yahoo!, these sites and services publicly perform musical works in numerous formats, including, inter alia, radio, television, movie, game, and music-video formats. Also like Yahoo!, only a portion of the content on RealNetworks' sites and services consist of performances of musical works.

In addition to performing music on websites and through services, the Internet Companies offer to users copies of recordings of musical works through download transmittals. A download is a transmission of an electronic file containing a digital copy of a musical work that is sent from an on-line server to a local hard drive. See United States v. Am. Soc'y of Composers, Authors & Publishers (Application of Am. Online, Inc., RealNetworks, Inc., and Yahoo! Inc.) ("RealNetworks and Yahoo! I"), 485 F.Supp.2d 438, 441 (S.D.N.Y.2007). With a download, the song is not audible to the user during the transfer. Id. at 442, 446. Only after the file has been saved on the user's hard drive can he listen to the song by playing it using a software program on his local computer. Id.

The Internet Companies primarily generate revenue from performances of musical works in two ways. On their websites, they make available, at no cost to users, performances of music, music videos, television programming, and the like that generate revenue from advertisements on the web page or in the audio or audio-visual player.3 The district court found that, in all of the forms of website advertising it considered, one principle is common: the larger the audience and the more times a site is visited, the greater the revenue generated. See United States v. Am. Soc'y of Composers, Authors & Publishers (Application of Am. Online, Inc., RealNetworks, Inc., and Yahoo!) ("RealNetworks and Yahoo! II"), 559 F.Supp.2d 332, 338 (S.D.N.Y.2008). For example, advertisers typically pay for display advertising based on the number of "impressions," or views, of the advertisement by users of the page on which an advertisement appears. The second primary way that the Internet Companies generate revenue from performing musical works is through subscription-based services.

II. PROCEDURAL BACKGROUND

Acting in its capacity as the rate court,4 the district court (William C. Conner, Judge)5 issued rulings in April 2007, April 2008, and January 2009 resolving the issues presented on appeal. In its 2007 decision, the district court held that a download of a digital file containing a musical work does not constitute a public performance of that work. In its 2008 decision, the district court determined a method for calculating the fees for the blanket licenses payable to ASCAP for the Internet Companies' performances of musical works in the ASCAP repertory. In two separate opinions issued in January 2009, the district court, applying the method it determined in 2008, issued Final Fee Determinations for Yahoo! and RealNetworks, respectively.

In its second opinion, issued in 2008, the district court arrived at a license fee formula that multiplied a royalty rate by the percentage of revenue attributable to the performance of music. The district court applied a uniform royalty rate to the Internet Companies' varying music uses that did not fluctuate over the different types of performances on the Internet Companies' sites and services. In ultimately determining a royalty rate of 2.5% for both of the Internet Companies, the district court relied upon several benchmark agreements, including ASCAP's agreements with Music Choice, terrestrial radio stations, the broadcast television networks, and the cable television networks.

For Yahoo!, because only a portion of the revenue generated from its website is attributable to performances of musical works, the district court decided to measure Yahoo!'s music-use revenue by multiplying the company's total revenue from its licensed services—defined as those business units that publicly perform music—less certain customary costs (such as for advertising sales commissions and traffic acquisition expenses) by a music-use-adjustment factor ("MUAF"). The MUAF was a fraction that reflected the amount of time users spent streaming performances of musical works relative to their overall time on the website; its numerator was the number of hours of music streamed from the licensed sites and services, and its denominator was the number of hours that the company's licensed sites and services were utilized.

For RealNetworks, the district court at first accepted ASCAP's argument that it was unnecessary to apply a MUAF because, unlike Yahoo!, "the vast majority of RealNetworks's revenue subject to fee is generated from subscription music services and advertising-supported sites where music is the central theme." RealNetworks and Yahoo! II, 559 F.Supp.2d at 399; see also id. at 411. The district court, however, did reduce RealNetworks' revenue figure by subtracting revenue attributable to RealNetworks' Technology Products and Solutions business unit, which develops and markets software products and services that enable wireless carriers, cable companies, and other media communication companies to distribute media content to PCs, mobile phones, and other non-PC devices. Id. at 359-60, 411-12. In its 2009 Final Fee Determination, the district court altered course and applied certain MUAFs to RealNetworks' various sites and services, but without explaining how it arrived at these MUAFs. (RealNetworks Judgment Order 2-4)

DISCUSSION

I. Public Performance Right as Applied to Downloads

The Copyright Act confers upon the owner of a copyright "a bundle of discrete exclusive rights," each of which may be transferred or retained separately by the copyright owner. N.Y. Times Co. v. Tasini, 533 U.S. 483, 495-96, 121 S.Ct. 2381, 150 L.Ed.2d 500 (2001) (internal quotation marks omitted). Section 106 of the Copyright Act sets forth these various rights, including the right "to reproduce the copyrighted work in copies" and the right "to perform the copyrighted work publicly." 17 U.S.C. § 106(1), (4). In this case, the Internet Companies offer their customers the ability to download musical works over the Internet. It is undisputed that these downloads create copies of the musical works, for which the parties agree the copyright owners must be compensated. However, the parties dispute whether these downloads are also public performances of the musical works, for which the copyright owners must separately and additionally be compensated. The district court held that these downloads are not public performances, and we agree.6

In answering the question of whether a download is a public performance, we turn to Section 101 of the Copyright Act, which states that "[t]o `perform' a work means to recite, render, play, dance, or act it, either directly or by means of any device or process."7 17 U.S.C. § 101. A download plainly is neither a "dance" nor an "act." Thus, we must determine whether a download of a musical work falls within the meaning of the terms "recite," "render," or "play."

"As in all statutory construction cases, we begin with the language of the statute. The first step is to determine whether the language at issue has a plain and unambiguous meaning with regard to the particular dispute in the case." Barnhart v. Sigmon Coal Co., 534 U.S. 438, 450, 122 S.Ct. 941, 151 L.Ed.2d 908 (2002) (internal quotation marks omitted). "[U]nless otherwise defined, statutory words will be interpreted as taking their ordinary, contemporary, common meaning." United States v. Piervinanzi, 23 F.3d 670, 677 (2d Cir.1994) (internal quotation marks and alterations omitted). "When the language of a statute is unambiguous, judicial inquiry is complete."8 Marvel Characters, Inc. v. Simon, 310 F.3d 280, 290 (2d Cir. 2002) (internal quotation marks omitted); accord Barnhart, 534 U.S. at 450, 122 S.Ct. 941.

The ordinary sense of the words "recite," "render," and "play" refer to actions that can be perceived contemporaneously. To "recite" is "to repeat from memory or read aloud esp[ecially] before an audience," Webster's Third New International Dictionary 1895 (1981); to "render" is to "say over: recite, repeat,"9 id. at 1922; and to "play" is to "perform on a musical instrument," "sound in performance," "reproduce sound of recorded material," or "act on a stage or in some other dramatic medium," id. at 1737. All three actions entail contemporaneous perceptibility.

These definitions comport with our common-sense understandings of these words. Itzakh Perlman gives a "recital" of Beethoven's Violin Concerto in D Major when he performs it aloud before an audience. Jimmy Hendrix memorably (or not, depending on one's sensibility) offered a "rendition" of the Star-Spangled Banner at Woodstock when he performed it aloud in 1969. Yo-Yo Ma "plays" the Cello Suite No. 1 when he draws the bow across his cello strings to audibly reproduce the notes that Bach inscribed. Music is neither recited, rendered, nor played when a recording (electronic or otherwise) is simply delivered to a potential listener.

The final clause of the § 101 definition of "to perform" further confirms our interpretation. It states that "[t]o `perform'... a motion picture or other audiovisual work ... [is] to show its images in any sequence or to make the sounds accompanying it audible." 17 U.S.C. § 101. The fact that the statute defines performance in the audio-visual context as "show[ing]" the work or making it "audible" reinforces the conclusion that "to perform" a musical work entails contemporaneous perceptibility. ASCAP has provided no reason, and we can surmise none, why the statute would require a contemporaneously perceptible event in the context of an audio-visual work, but not in the context of a musical work.

The downloads at issue in this appeal are not musical performances that are contemporaneously perceived by the listener. They are simply transfers of electronic files containing digital copies from an on-line server to a local hard drive. The downloaded songs are not performed in any perceptible manner during the transfers; the user must take some further action to play the songs after they are downloaded. Because the electronic download itself involves no recitation, rendering, or playing of the musical work encoded in the digital transmission, we hold that such a download is not a performance of that work, as defined by § 101.

ASCAP, pointing to the definition of "publicly" in § 101, argues that a download constitutes a public performance. Section 101 defines "[t]o perform or display a work `publicly'" as follows:

Id. § 101. ASCAP argues that downloads fall under clause (2) of this definition because downloads "transmit or otherwise communicate a performance," id., namely the initial or underlying performance of the copyrighted work, to the public. We find this argument unavailing. The definition of "publicly" simply defines the circumstances under which a performance will be considered public; it does not define the meaning of "performance." Moreover, ASCAP's proposed interpretation misreads the definition of "publicly." As we concluded in Cartoon Network LP v. CSC Holdings, Inc., "when Congress speaks of transmitting a performance to the public, it refers to the performance created by the act of transmission," not simply to transmitting a recording of a performance. 536 F.3d 121, 136 (2d Cir. 2008). ASCAP's alternative interpretation is flawed because, in disaggregating the "transmission" from the simultaneous "performance" and treating the transmission itself as a performance, ASCAP renders superfluous the subsequent "a performance... of the work" as the object of the transmittal. See Duncan v. Walker, 533 U.S. 167, 174, 121 S.Ct. 2120, 150 L.Ed.2d 251 (2001) ("It is our duty to give effect, if possible, to every clause and word of a statute." (internal quotation marks omitted)). In contrast, our interpretation in Cartoon Network recognizes that a "transmittal of a work" is distinct from a transmittal of "a performance"—the former being a transmittal of the underlying work and the latter being a transmittal that is itself a performance of the underlying work. See 536 F.3d at 134 ("The fact that the statute says `capable of receiving the performance,' instead of `capable of receiving the transmission,' underscores the fact that a transmission of a performance is itself a performance.").

The Internet Companies' stream transmissions, which all parties agree constitute public performances, illustrate why a download is not a public performance. A stream is an electronic transmission that renders the musical work audible as it is received by the client-computer's temporary memory. This transmission, like a television or radio broadcast, is a performance because there is a playing of the song that is perceived simultaneously with the transmission. See, e.g., Twentieth Century Music Corp. v. Aiken, 422 U.S. 151, 158, 95 S.Ct. 2040, 45 L.Ed.2d 84 (1975). In contrast, downloads do not immediately produce sound; only after a file has been downloaded on a user's hard drive can he perceive a performance by playing the downloaded song.10 Unlike musical works played during radio broadcasts and stream transmissions, downloaded musical works are transmitted at one point in time and performed at another. Transmittal without a performance does not constitute a "public performance." Cf. Columbia Pictures Indus., Inc., v. Prof'l Real Estate Investors, Inc., 866 F.2d 278, 282 (9th Cir. 1989) (holding that renting videodiscs to a hotel guest for playback in the guest's room does not constitute the "transmission" of a public performance).

ASCAP misreads our opinion in NFL v. PrimeTime 24 Joint Venture, 211 F.3d 10, 11-13 (2d Cir.2000), to hold that the Copyright Act does not, in fact, require a contemporaneously perceptible performance to infringe on the public performance right. In NFL, defendant PrimeTime, a satellite television provider, captured protected content in the United States from the NFL, transmitted it from the United States to a satellite ("the uplink"), and then transmitted it from the satellite to subscribers in both the United States and Canada ("the downlink"). PrimeTime had a license to transmit NFL games to its subscribers in the United States but not to Canada. The NFL sought to enjoin the transmissions sent to Canada by arguing that the uplink in the United States constituted unauthorized public performances of the games in the United States. The relevant issue was whether the uplink transmission was a public performance even though the uplink was only to a satellite and could not, itself, be perceived by viewers. Id. at 12. We determined that PrimeTime's uplink transmission of signals captured in the United States amounted to a public performance because it was an integral part of the larger process by which the NFL's protected work was delivered to a public audience. Id. at 13.

ASCAP seizes on the fact that the uplink to the satellite was not contemporaneously perceptible to argue against a contemporaneous perceptibility requirement in this case. ASCAP's argument, however, fails to accord controlling significance to the fact that the immediately sequential downlink from the satellite to Canadian PrimeTime subscribers was a public performance of the games. Id. at 11-13; see also David v. Showtime/The Movie Channel, Inc., 697 F.Supp. 752, 758-60 (S.D.N.Y.1988) (finding that because "Showtime and The Movie Channel both broadcast television programming ... to cable system operators," which, in turn, "pass[ed] the signal along to their individual customers," the initial transmissions constituted public performances because they were a "step in the process by which a protected work wends its way to its audience"); Melville B. Nimmer & David Nimmer, 2 Nimmer on Copyright § 8.14[C][2] at 190.6 & n. 63 (2009) (explaining that when a transmission is made "to cable systems that will in turn transmit directly to the public," the earlier transmission is a public performance despite the absence of any contemporaneous perceptibility). In holding the transmission in Cartoon Network not to be a public performance, we distinguished NFL on the basis that in that case the final act in the sequence of transmissions was a public performance. See 536 F.3d at 137. That same distinction applies here. Just as in Cartoon Network, the Internet Companies transmit a copy of the work to the user, who then plays his unique copy of the song whenever he wants to hear it; because the performance is made by a unique reproduction of the song that was sold to the user, the ultimate performance of the song is not "to the public." See id. at 137, 138; see also United States v. Am. Soc'y of Composers, Authors & Publishers (Application of Cellco P'ship), 663 F.Supp.2d 363, 371-74 (S.D.N.Y.2009).

Accordingly, we affirm the district court's grant of partial summary judgment on the basis that downloads do not constitute public performances of the downloaded musical works.11

II. Fee Determination for Using the ASCAP Repertory

We now turn to the district court's determination of the appropriate fees payable by the Internet Companies for blanket licenses to publicly perform any of the millions of musical compositions in the ASCAP repertory. The district court determined these blanket license fees by applying a uniform royalty rate of 2.5% to the Internet Companies' music-use revenue, which was calculated by multiplying the total revenue from licensed services by a MUAF (music-use-adjustment factor), a fraction that reflected the amount of time users spent streaming performances of musical works relative to their overall time on the website. We conclude that the district court's analysis was flawed in two major respects.

First, the district court did not adequately support the reasonableness of its method for measuring the value of the Internet Companies' music use. Second, the district court did not adequately support the reasonableness of the 2.5% royalty rate applied to the value of the Internet Companies' music use. Accordingly, we remand to the district court so that it may redetermine reasonable fees for the licenses in light of the following discussion.

A. Standards of Review

"In order to find that the rate set by the District Court is reasonable, we must find both that the rate is substantively reasonable (that it is not based on any clearly erroneous findings of fact) and that it is procedurally reasonable (that the setting of the rate, including the choice and adjustment of a benchmark, is not based on legal errors)." United States v. Broad. Music, Inc. (Application of Music Choice) ("Music Choice IV"), 426 F.3d 91, 96 (2d Cir.2005). Fundamental to the concept of "reasonableness" is a determination of what an applicant would pay in a competitive market, taking into account the fact that ASCAP, as a monopolist, "exercise[s] disproportionate power over the market for music rights." Id.

"A rate court's determination of the fair market value of the music is often facilitated by the use of benchmarks—agreements reached after arms' length negotiation between other similar parties in the industry." Id. at 94. Determinations by the district court that particular benchmarks are comparable and particular factors are relevant are questions of law reviewed de novo. Am. Soc'y of Composers, Authors & Publishers v. Showtime/The Movie Channel, Inc. ("Showtime"), 912 F.2d 563, 569-71 (2d Cir.1990). However, factual findings as to each factor under consideration or those underlying a proposed benchmark agreement, as well as findings with respect to fair market value, are reviewed for clear error. Music Choice IV, 426 F.3d at 96; Showtime, 912 F.2d at 569; United States v. Am. Soc'y of Composers, Authors & Publishers (Application of Capital Cities/ABC, Inc., CBS Inc., & Nat'l Broad. Co.), 157 F.R.D. 173, 195-96 (S.D.N.Y.1994).

B. The District Court's Determination of the Music-Use-Adjustment Factor

Fundamental to the reasonableness of a fee for music use under a license is the reasonableness of the determination of the revenue attributable to the actual uses by the applicant of the music to which the rate percentage is to be applied. See United States v. Am. Soc'y of Composers, Authors and Publishers (Applications of Capital Cities/ABC, Inc. & CBS, Inc.) ("Capital Cities"), 831 F.Supp. 137, 156-57 (S.D.N.Y.1993); United States v. Am. Soc'y of Composers, Authors and Publishers (Application of Nat'l Cable Television Ass'n.), No. 41-CV-1395 (WCC)(MHD), 1999 WL 335376, at *12 (S.D.N.Y. May 26, 1999). In this case, the value of the applicants' uses could not be premised on total revenue without an adjustment for the fact that some revenues were not at all attributable to any use of ASCAP music. The district court decided to make this adjustment by using a MUAF that discounted the total revenue to reflect only those revenues attributable to music use. We have no quarrel with the use of a MUAF here, but we find error in the district court's method of determining its components.

The district court began by calculating the total revenue of the licensed sites and services—defined as those business units that publicly perform music—less certain customary deductions, to arrive at the revenue base to which the MUAF was applied. The Internet Companies argue that total revenue, when used as the revenue base, bears no relation to the value they derive performing musical works on the Internet. This argument is unavailing because it overlooks the function of the MUAF, which can be understood as accounting for the value of the Internet Companies' music use in relation to their overall revenue. We find nothing wrong in concept with using a formula that reduces the total revenue of the licensed sites and services by a factor that, with substantial accuracy, accounts for music-use revenue, as the MUAF can be understood to do here.12 With respect to the components of the MUAF, however, we find error.

1. Yahoo!

The district court's MUAF accounts for the value of Yahoo!'s music use by using the amount of time that music is streamed. Streaming time, however, neither drives nor correlates with Yahoo!'s advertising revenue. The record evidence makes plain that Yahoo!'s advertising revenue model more accurately correlates with the number of times a particular page is accessed by users than to the duration of streaming time. To the extent that the district court's MUAF relied on an imprecise metric for determining advertising revenue attributable to music use when a superior metric was apparently available and practicable,13 the district court's method for calculating the MUAF was unreasonable.14 E.g., Capital Cities, 831 F.Supp. at 156-57 (accepting amount of music used in television program as an adequate measure of the value of the music to the program because it was the only measure available).

Display advertising on the Internet is sold on a cost-per-thousand model that counts the total number of page impressions, i.e., how many times a particular page is accessed. Pages that are accessed a greater number of times occasion higher advertising rates because the advertisements on these pages are viewed with greater frequency. It is, thus, unreasonable to use streaming time, which has no necessary correlation with page views, as a proxy for the number of times a page is viewed; time spent on-line is not reflective of how a user interacts with a particular page.

A user may have a page open that he is not viewing at all, either because he has multiple pages open and is interacting with another page, website, or computer program, or because he has walked away from the computer altogether but has left that page open. For example, user A, who is listening to a four-minute song, may view only the page on which the song is playing in that four minutes because his exclusive focus is on the song, while user B, who is listening to the same song on as background music, may be simultaneously clicking on links and reading articles throughout Yahoo!'s website, and thus may be seeing multiple advertisements on multiple pages during that same four-minute period. The streaming time for users A and B is the same, but the advertising exposure of each differs widely.

The advertising marketplace takes account of the foregoing consideration by applying different advertising rates to different areas of the website. For example, advertising for radio-style webcasts in Yahoo! Music is priced at a rate lower than similarly placed ads on Yahoo!'s homepage, because users normally have the Yahoo! Music window minimized (and thus not viewable) for much of the time the radio-style music is playing, in contrast to other areas on Yahoo!, like the homepage, that command greater viewer attention.15 The district court erred by constructing a MUAF that failed to take into account these various realities of Internet advertising.16

In sum, the district court erred by adopting an imprecise metric—music streaming time rather than page views—as the basis of its MUAF, without providing a sufficient rationale for that decision. We will not specify a particular method of developing a formula for determining music-use revenue on remand; we also leave it to the district court to determine whether it should proceed with a variant of its current formula (revenue multiplied by a MUAF) or in some other way altogether, in light of the foregoing discussion.

Yahoo! has also faulted the district court for using statistics with differing methodologies in the MUAF's numerator and denominator. In calculating the numerator, the district court used Yahoo!'s statistics for the number of hours of music streamed, which give no effect to the specific window engaged by the user. In calculating the denominator, the district court used statistics based on a different methodology provided by comScore Media Metrix for the total hours of use of the licensed sites and services.17 Unlike Yahoo!, comScore measures hours used only for the specific window that is engaged. The difference is that, in an instance in which a user has multiple windows open at one time, comScore will count the time for only the single window that is in active use, while Yahoo! will count the time for all the windows open. Yahoo!'s statistics thus reflect considerably greater use time than comScore's because, anytime a user has multiple windows open, Yahoo! is counting the time use for each of the Yahoo! windows open while comScore is counting the time use for only the single window engaged by the user. Without addressing whether one method is preferable to the other, we conclude that constructing a MUAF by using Yahoo!'s statistics for the numerator and comScore's statistics for the denominator is unreasonable because these statistics are not comparable, with the result that their comparison overstates music-streaming time.

The district court's opinion states that, as of the date of its issuance, Yahoo! had failed to supply any site-hours data comparable to that supplied by comScore that could have been used for the denominator, but that Yahoo! was free to do so before the district court ordered its Final Fee Determination. The parties disagree over whether Yahoo! ultimately furnished adequate total site-hours data. Because we remand for reconsideration of the MUAF, we refer this dispute to the district court. We do, however, note that in calculating any MUAF, the district court must strive to use measurements that are as consistent and as precise as practicable.

2. RealNetworks

Turning to RealNetworks, as noted previously the district court in 2009 determined and applied differing MUAFs to RealNetworks' various licensed sites and services despite its decision not to apply MUAFs in 2008, without any explanation for the basis of these MUAFs. Accordingly, we remand for explanation (or reconsideration if the current MUAFs cannot be justified) because the district court's rationale was insufficient.

The district court applied the following MUAFs to RealNetworks. For its Rhapsody subscription service,18 the district court defined the MUAF's numerator as total number of plays of audio-music and music-video streams, and the denominator as the total number of Rhapsody streams plus the number of deliveries of conditionally downloaded music files to subscribers. (RealNetworks Judgment Order 2-3) For its SuperPass subscription service,19 the numerator was fixed as the total number of hours of streams to users in each month by means of the music-radio portion of SuperPass, and the denominator as the total number of SuperPass music hours plus the total number of hours of all other streams to users by means of SuperPass. (RealNetworks Judgment Order 3) And for RealNetworks' Music, and Media Software and Services groups, the district court used as the numerator the total number of hours of streams that non-subscription, on-demand, and radio-music users receive from RealNetworks' licensed sites, and as the denominator the total number of hours users spend on RealNetworks' licensed sites. (RealNetworks Judgment Order 4)

Because the district court failed to explain the basis for these MUAFs, and in light of the issues we raised with respect to the MUAF applied to Yahoo!, we remand for the district court to explain or reconsider the MUAFs applied to RealNetworks. In addition to consideration of any issues that it deems appropriate, the district court should address the following: (1) whether its method for calculating the MUAF for the Rhapsody subscription service is more precise or practicable than the method used in the benchmark agreements in the record; (2) whether there is a more precise way, that is also practicable, to account for the value of the music use for the SuperPass subscription service in light of the fact that some components of the subscription do not involve the streaming of content to users; and (3) whether there is a more precise and still practicable way to measure RealNetworks' advertising revenue, in light of the issues we raised in our discussion of Yahoo!'s MUAF.

C. The Royalty Rate Applied by the District Court

1. Yahoo!

The district court arrived at Yahoo!'s royalty rate by relying on benchmark agreements for blanket licenses that ASCAP entered into with Music Choice, terrestrial radio stations, the broadcast television networks, and cable television providers, as well as the rates that Yahoo! itself pays to the major record companies for music-video rights. The district court's factual findings that support selecting these benchmarks were not clearly erroneous. After reviewing the district court's analysis of these benchmarks in relation to this case, however, we hold that the district court unreasonably arrived at its decision to apply a uniform 2.5% royalty rate and that, in setting the royalty rate, the district court must follow an approach more tailored to the varying nature and scope of Yahoo!'s music use.

Beginning with the Music Choice benchmark, the district court found that Music Choice provides channels of music to listeners on a subscription basis via cable and satellite television and the Internet. Music Choice's channels are organized around genres of music, and, on the Internet, listeners have the option to create up to ten personalized audio channels. ASCAP's current blanket license with Music Choice, for the period January 1, 2006 through December 31, 2010, calls for payment of a royalty rate that is 2.5% of Music Choice's gross revenues, where gross revenue is defined as all revenues derived from the licensed services.

We conclude that the royalty rate agreed to by Music Choice provides strong support for applying a 2.5% royalty rate to those Yahoo! sites and services that provide access to music channels organized around music genre, similar to those on Music Choice. Additionally, it provides a basis for a 2.5% royalty rate, or higher, for Yahoo! sites and services that permit an interactive music experience, in which the user may control the selection of music he is hearing, for example if a user tunes into a more customized station or uses Yahoo! Search to listen to songs on-demand. See RealNetworks and Yahoo! II, 559 F.Supp.2d at 413.

However, the Music Choice benchmark does not justify applying a 2.5% royalty rate to all of Yahoo!'s music uses, because Yahoo! offers numerous sites and services that are less music intensive than Music Choice's offerings. The district court finessed the Music Choice benchmark's limited relevance by concluding that there are other benchmarks in the record that support applying a 2.5% royalty rate to all of Yahoo!'s performances of the ASCAP repertory. Because we conclude that these other benchmarks do not adequately support an across-the-board 2.5% royalty rate, we remand for reconsideration of a reasonable royalty rate.

Turning to the major broadcast television networks' agreements with ASCAP, the district court found that these agreements differed from Yahoo!'s because the major networks pay a flat rate for their ASCAP usage without regard to revenue (as specified in Yahoo!'s application) and that "[a similar] flat-fee structure is unsuitable for the online music industry," RealNetworks and Yahoo! II, 559 F.Supp.2d at 401. The district court nevertheless concluded that, by comparing the percentages of broadcast revenue that the television companies pay ASCAP under their flat royalty rate to the percentage of licensed services revenue that Yahoo! would pay if a 2.5% royalty rate (as well as a MUAF) were applied to its license, the networks' agreements could be "useful to gauge the reasonableness of the fee range ASCAP seeks from . . . Yahoo!, and [Yahoo!'s] ability to pay the blanket fees rather than resorting to less efficient licensing options or foregoing the use of ASCAP music altogether." Id.

The district court found that "[t]he three television networks, ABC, CBS, and NBC, have annual revenues in the range of $3 to $4 billion[,] comparable to current . . . Yahoo! revenues" and that the networks pay a percentage of broadcast revenue for their ASCAP license that is comparable to the amount Yahoo! would have paid if a 2.5% royalty rate (as well as a MUAF) was applied to the total domestic revenue from the licensed services for the same period. From this finding, the district concluded that, to the extent that music is not much less important to Yahoo! than it is to the television networks, a 2.5% royalty rate is reasonable.

The district court similarly looked to percentage of total revenue when assessing ASCAP's agreements with the general-entertainment and music-intensive cable television networks. The district court found that the general-entertainment cable television networks pay 0.375% of total revenue derived from licensed services and that the music-intensive cable television networks pay 0.9% of licensed-services revenue derived from licensed services. As it did with the major broadcast television networks, the district court compared the percentage of total revenue that Yahoo! would pay under the court's formula to the 0.375% and 0.9% paid by the cable television networks to conclude that a flat 2.5% royalty rate, when multiplied by a MUAF, is reasonable for Yahoo!'s license.

We are unconvinced that percentage of revenue comparisons between broadcast and cable television networks, and Yahoo! are useful in determining a reasonable fee for Yahoo!'s public performances of the ASCAP repertory. Nearly every program on a television station somehow utilizes musical works. In contrast, only a fraction of the traffic on Yahoo!'s website uses music; much of Yahoo!'s website does not implicate any music use whatsoever. Given that Yahoo!'s revenue base relies far less on ASCAP content than the television networks' revenue base, we believe that comparing percentages of overall revenue bases is of little probative value in this benchmark analysis.20

These comparisons, moreover, indicate a deeper flaw in the district court's analysis: its inclination to lump all of Yahoo!'s varying musical uses together, instead of looking to the nature and scope of Yahoo!'s different types of uses and applying a rate that reflects (or rates that reflect) the varying nature of Yahoo!'s music use. See United States v. Am. Soc'y of Composers, Authors & Publishers (Application of Nat'l Cable Television Ass'n.), No. 41-CV-1395 (WCC)(MHD), 1999 WL 335376, at *12 (S.D.N.Y. May 26, 1999). The district court's finding that cable television networks pay 0.375% of their revenue for their ASCAP licenses is a good example. The most direct conclusion to be drawn from this benchmark is that providers of cable-style television will pay 0.375% of their revenue, in a competitive market, to license ASCAP music.21 However, the district court looked past this conclusion in setting the royalty rate and instead used this benchmark to justify applying a 2.5% rate to all of Yahoo!'s music use based on an imprecise comparison of the percentages of overall revenue. We think a better approach would be to attempt, if practicable, to set a royalty rate that requires Yahoo! to pay a rate for its cable-style programming that is similar to that in the cable market.22

Our view that a reasonable royalty rate should reflect the varying values of Yahoo!'s differing music uses is supported by Yahoo!'s license with BMI.23 As previously noted, BMI is ASCAP's principal competitor in licensing the performance rights for musical works. The two rights organizations control approximately 90% of all on-line music performances, with roughly equal shares of the market. Because these two companies operate in the market in such similar manners, BMI's agreements are instructive. See e.g., United States v. Am. Soc'y of Composers, Authors & Publishers (Application of MobiTV, Inc.), 712 F.Supp.2d 206, 248-49 (S.D.N.Y. 2010).

In its negotiated agreement with BMI, Yahoo! agrees to pay the following license fees: (a) 1.75% of "Direct Revenue" from "Preprogrammed Radio"; (b) 2.5% of "Direct Revenue" from "On-Demand Streams"; (c) 2.15% of "Direct Revenue" from "User-Influenced Programming"; (d) 1.0% of "Direct Revenue" from "Audio-Visual Programming"; and (e) 0.6875% of "Yahoo! Music Advertising Revenue and Yahoo! Music Run of Site Allocation." This agreement, providing for a so-called "bucket" for each different use, supports the conclusion that the other benchmarks also suggest: the market assigns different values to Yahoo!'s different music uses and is capable of yielding a royalty rate that reflects the varying intensity of Yahoo!'s music uses.24

The district court found that effectuating the BMI structure is quite complex, because Yahoo! is required to subdivide its sites and services into 17 separate revenue categories that are apportioned into "5 buckets" and that, to report the revenue under these categories, Yahoo! is required to make dozens of calculations and collect numerous data points. "Because most of these data [points] are not ordinarily collected for other business purposes," the district court concluded that "their accuracy [i]s suspect and auditing [Yahoo!'s] reports [i]s difficult and expensive." The district court found that the BMI method is additionally complicated by how the parties decide what revenue is "directly attributable" to the streaming of music. RealNetworks and Yahoo! II, 559 F.Supp.2d at 411. We neither endorse nor challenge these conclusions, and we leave it to the district court on remand to decide whether a bucket system is feasible with alterations, or whether another system is preferable in light of our guidance today.

We further acknowledge the requirement that fee structures and the proceedings used to arrive at them comport with the provisions of the Second Amended Final Judgment, see, e.g., United States v. BMI (Application of Muzak LLC), 275 F.3d 168, 176-77 (2d Cir.2001) (assessing whether a rate structure that included "carve outs" for songs licensed by the Applicant directly from the copyright holder was within the rate-setting court's blanket-license authority), and the district court's concern that the setting of different rates for the Internet Companies' various services would be in some tension with prior case law regarding blanket licenses, see RealNetworks & Yahoo! II, 559 F.Supp.2d at 406-08. We note, however, that the district court's own proposed MUAF would have involved "reporting or . . . keeping track of . . . music use," notwithstanding the court's observation that such tracking is generally not a feature of blanket licenses, see id. at 407, and that a recent rate-setting opinion has made use of multiple rates as part of a blanket license, see MobiTV, 712 F.Supp.2d at 231-32, 247-49, 254-55. Moreover, our own precedent indicates that non-traditional fee structures can be compatible with blanket licenses. Muzak, 275 F.3d at 177. It is for the district court to consider, in the first instance, what options are available to it in setting a reasonable rate or rates in this case.

For the foregoing reasons, we remand to the district court to redetermine the royalty rate (or rates) that Yahoo! must pay ASCAP for its license, in light of our holding that the district court's valuation of Yahoo!'s use of the ASCAP repertory must reflect, as well as practicable, the varying nature and scope of Yahoo!'s music use. We do not, however, suggest that the specific royalty rates set forth in the BMI-Yahoo! agreement must be accepted by the district court on remand, nor do we suggest that a bucket system is the only method by which a reasonable fee could be calculated. Instead, we believe that there are other ways to proceed, including possibly using a "blended" uniform rate25 (e.g., taking a snapshot of Yahoo!'s current music use, valuing the different uses independently, averaging them into a blended, uniform royalty rate, and then revisiting that rate periodically), or perhaps using a modification of the BMI bucket system to avoid some of the reliability problems noted by the district court, or employing some third method not yet discussed. We leave it to the district court to determine the best way to proceed consistent with the concerns we have discussed.

2. RealNetworks

The district court applied the Music Choice and radio station benchmarks to RealNetworks. The court found that Music Choice's 2.5% royalty rate supported applying a 2.5% royalty rate to RealNetworks because, inter alia, many of RealNetworks' music uses are equally or more music intensive than Music Choice's uses, specifically referring to RealNetworks' on-demand, music-video, and music-stream uses. The district court also found that the radio station benchmarks supported a 2.5% royalty rate for RealNetworks.

We do not take specific issue with the district court's analysis as it may pertain to uses by RealNetworks that are analogous to those of Music Choice and the radio stations. However, RealNetworks objects to the district court's analysis on the basis that RealNetworks' services include music uses that are less music-oriented than those of Music Choice or the radio stations, such as uses in video games, television shows, and ring tones. We find this objection persuasive to the extent that it tracks the flaws we have identified in the district court's analysis underlying Yahoo!'s royalty rate, and we remand so that the district court may conduct a more complete analysis of the various uses of ASCAP musical works by RealNetworks and may determine, in light thereof, the appropriate method for determining RealNetworks' royalty rate (or rates).

D. Download Revenues

With respect to the district court's fee calculation, ASCAP cross-appeals one issue: whether download revenue should be included in the revenue base, even though downloads are not public performances, because public performances of music are used to generate the downloads of music. Although the district court recognized some relationship between the ability to stream music and download revenue, it made no formal factual findings about the extent and implications of that relationship. In an analogous context, the district court found that the ability to preview ringtones (via streaming) "undeniably increase[d]" the sale of ringtones (via download). United States v. Am. Soc'y of Composers, Authors & Publishers (Application of AT & T Wireless f/k/a Cingular Wireless), 607 F.Supp.2d 562, 565 (S.D.N.Y. 2009). We leave it to the district court on remand to determine the extent of any such relationship in this case.

Finally, ASCAP cross-appeals the district court's admission of the Internet Companies' alleged late evidence. This issue is moot in light of our remand to the district court for further proceedings, at which additional evidence may be considered.

CONCLUSION

For the foregoing reasons, we AFFIRM the district court's ruling that downloads of musical works do not constitute public performances of those works, and we VACATE the district court's assessment of reasonable fees for the blanket ASCAP licenses sought by the Internet Companies and REMAND for further proceedings in light of this opinion.