OPINION

THOMAS, Judge:

These consolidated antitrust actions arise out of an agreement ("Promotion Agreement") between Netflix and Walmart1 whereby Walmart transferred its online DVD-rental subscribers to Netflix, and Netflix agreed to promote Walmart's DVD sales business. The plaintiffs, individuals representing a class of Netflix subscribers ("Subscribers"), contend that this arrangement violated §§ 1 and 2 of the Sherman Act by illegally allocating and monopolizing the online DVD-rental market. The Subscribers' theory of injury is that they paid supracompetitive prices for one of Netflix's subscription plans because Netflix would have reduced the price of that plan but for its allegedly anticompetitive conduct.

We agree with the district court that the Subscribers have not raised a triable issue of fact as to whether they suffered antitrust injury-in-fact, and we affirm the district court's grant of summary judgment. We vacate in part the district court's cost award, and remand for consideration in light of this opinion.

I

In 1997, Reed Hastings and Marc Randolph co-founded Netflix, the first internet-based DVD rental service. Netflix commenced operations in 1998, offering customers through its website the option to rent or buy DVDs by mail. Netflix initially offered DVD rentals on a pay-per-rental basis, but soon replaced that system with a monthly subscription model. In 2000, it discontinued its DVD sales business altogether. Several Netflix subscription plans permitted customers to rent an unlimited number of DVDs, differing in how many DVDs a customer could borrow at a given time. For example, in 2003, Netflix offered its "3U" plan, which permitted three DVDs to be rented at a time, for $19.95 per month, while its "4U" plan cost $24.95 per month and allowed four DVDs at a time. Netflix's DVD-rental business flourished under the new model, and by 2005 it had a 77.8% share of the online DVD-rental market, rising to 92.3% by 2010.

Netflix faced no serious competition in its early years. However, in 2003, Walmart, one of the nation's largest retail companies, launched its own online DVD-rental service. Walmart initially offered its 3U plan for $18.76 a month. Although Walmart's 3U plan was cheaper than Netflix's ($19.95 per month), Netflix did not alter its 3U plan price for a full year. When Netflix eventually did change its 3U price, in June 2004, it increased the price to $21.99 per month.

Two months later, in August 2004, Blockbuster, the largest store-based DVD rental company, launched its own online DVD rental service, becoming the third major competitor in the market. Blockbuster offered its 3U plan at $19.99 per month and included with it two free coupons per month for in-store rentals.

In October 2004, in apparent response to rumors that Amazon planned to enter the online DVD-rental market as well, Netflix announced that it would lower the price of its 3U plan from $21.99 to $17.99 per month. Blockbuster responded the next day by announcing that it would cut its 3U price to $17.49 per month. In November 2004, Walmart reduced its 3U price to $17.36 per month. In December 2004, Blockbuster again reduced the price of its 3U plan, this time to $14.99 per month — the lowest 3U plan price in the market. Netflix maintained its $17.99 price until August 2007, when it lowered the price to $16.99.

During this period, Walmart's online DVD-rental business performed poorly. Walmart never had more than 60,000 subscribers. In contrast, in mid-2004 Netflix had over 2 million subscribers, and Blockbuster had 400,000 subscribers. From June 2003, when Walmart opened its online DVD-rental business, until it signed the Promotion Agreement with Netflix in March 2005, Walmart gained an average of 5,000 subscribers per quarter. Netflix added 250,000 subscribers per quarter over the same period. Walmart's subscriber share peaked at 2.4% in early 2004 and declined from that point. By February 2005, Walmart had only a 1.4% market share. In contrast, Netflix controlled 77.8% of the market in 2005. Walmart lost 7,000 subscribers during the final quarter of 2004.

In October 2004, Netflix's CEO Reed Hastings sought a meeting with Walmart CEO John Fleming. Hastings testified that he requested the meeting because he hoped to form a partnership with Walmart that would strengthen Netflix's position before Amazon entered the market. Hastings was aware that Walmart's online DVD-rental service was performing poorly, and hoped that Walmart might therefore be open to a partnership. The two CEOs met on October 27, 2004. Hastings recounts that Walmart seemed uninterested in a deal at the time and that there was no discussion about Netflix selling new DVDs. Fleming provided a similar account. No agreement was reached at the meeting.

Unbeknownst to Hastings, Walmart was entertaining other suitors. Walmart considered a potential partnership with Yahoo!, and a draft partnership agreement to that effect was prepared as early as December 1, 2004. Walmart considered a similar deal with Microsoft.

Walmart began considering alternative strategic options for its online DVD-rental business, and ultimately looked in depth at four possibilities: (1) continuing to run the business with a low subscriber amount, (2) aggressively building the business, (3) partnering with Yahoo!, and (4) exiting the online DVD-rental business. After carefully analyzing each option, Walmart concluded that none would be profitable and that, in fact, it would probably suffer multi-million dollar financial losses under all four scenarios.

Walmart made the final decision to exit the market by early January 2005. It established an impairment reserve to cover any losses incurred from the closure and stopped accepting new subscribers for its 3U and 4U plans. By February 2005, Walmart had incurred $3 million of costs associated with shuttering its online DVD-rental business. By March 2005, Netflix had 3 million subscribers. Walmart had 52,000. Walmart employees speculated that Walmart's online DVD-rental business did not succeed because Walmart devoted insufficient resources to marketing, could not match Netflix's guaranteed one- to two-day delivery, had a poorly designed website, and offered a relatively limited selection of DVDs.

Aware of Walmart's market share decline, but unaware of its plan to discontinue its online DVD-rental business, Hastings renewed his efforts to meet with Fleming. The two CEOs met on February 9, 2005. Fleming did not inform Hastings that Walmart had decided to leave the online DVD-rental business. Although no agreement was reached at the meeting, Hastings's efforts did eventually bear fruit. By March 17, 2005, Hastings and Fleming had reached a verbal agreement, the key terms of which were that: (1) Walmart DVD-rental subscribers and their rental queues would be transferred to Netflix, for those customers who so chose, free of charge, and customers would be offered the same subscription price for one year; (2) Walmart would promote on its website Netflix's online DVD-rental business; (3) Netflix would pay Walmart a 10% revenue share for each subscriber who transferred, as well as a $36 bounty for each new Netflix subscriber gained from Walmart's referrals; and (4) Netflix would promote Walmart's DVD sales business.

These key terms were incorporated into the Promotion Agreement. The Promotion Agreement did not include a covenant not to compete, did not prohibit Netflix from selling new DVDs, and explicitly permitted Walmart to offer an online DVD-rental service.2 The Promotion Agreement was publically announced May 19, 2005.

Despite the earlier rumors, Amazon did not initiate an online DVD-rental service. Thus, after Walmart's exit in mid-2005, Netflix and Blockbuster remained as the two major competitors in the market. Blockbuster eventually filed for bankruptcy in September 2010, leaving Netflix as the sole major competitor, with over 90% of the online DVD-rental market.

Netflix kept its 3U price at $17.99 from November 2004 to August 2007, when it reduced the price to $16.99, which is where it remained through the end of the class period. During the class period, Netflix began offering video streaming over the Internet, and Netflix has since focused on developing that aspect of its business.

The Subscribers filed several actions, alleging antitrust violations by Netflix, Walmart Stores, and Walmart.com, and seeking to represent a class of Netflix subscribers. The actions were consolidated and an amended complaint filed. The thrust of the Subscribers' complaint is that the Promotion Agreement reflected an illegal allocation of the online DVD-rental market. The Subscribers assert four causes of action: (1) a § 1 Sherman Act violation for unlawful market allocation of the online DVD-rental market (against all defendants); (2) a § 2 Sherman Act claim for monopolization of the online DVD-rental market (against Netflix); (3) a § 2 Sherman Act claim for attempted monopolization of the online DVD-rental market (against Netflix); and (4) a § 2 Sherman Act claim for conspiracy to monopolize the online DVD-rental market (against all defendants).

The district court granted the Subscribers' motion for certification of a litigation class, defining the class as "[a]ny person or entity in the United States that paid a subscription fee to Netflix on or after May 19, 2005 up to and including [December 23, 2010,] the date of class certification." The district court subsequently approved Walmart's settlement with the class.

Netflix moved for summary judgment pursuant to Federal Rule of Civil Procedure 56 as to all claims asserted against it. The district court granted the motion, concluding that there was no per se antitrust violation, and that the Subscribers had failed to raise a triable issue as to antitrust injury-in-fact. The district court also excluded tendered evidence of agreements Netflix had with Amazon, Best Buy, and Musicland, because the agreements raised new theories of liability that were not expressly pleaded in the complaint.

The district court entered final judgment against the Subscribers, after which Netflix filed a bill of costs seeking $744,740.11 in discovery costs. The district court subsequently awarded Netflix $710,194.23 in costs. The Subscribers filed a timely notice of appeal, and Netflix cross-appealed.

We have jurisdiction pursuant to 28 U.S.C. § 1291. "We review de novo the district court's grant of summary judgment." Cascade Health Solutions v. PeaceHealth, 515 F.3d 883, 912 (9th Cir.2008). Summary judgment is appropriate when "there is no genuine dispute as to any material fact and the movant is entitled to judgment as a matter of law." Fed. R.Civ.P. 56(a). A genuine issue of material fact exists when "the evidence is such that a reasonable jury could return a verdict for the nonmoving party." Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 248, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986). Until recently, "[s]ummary judgment [was] disfavored in antitrust cases," High Tech. Careers v. San Jose Mercury News, 996 F.2d 987, 989 (9th Cir.1993), but "any presumption against the granting of summary judgment in complex antitrust cases has now disappeared," In re ATM Fee Antitrust Litig., 554 F.Supp.2d 1003, 1010 (N.D.Cal.2008) (citing Philip E. Areeda & Herbert Hovenkamp, Antitrust Law ¶ 308c2 (3d ed.2007)).

II

The district court properly concluded that the Subscribers failed to raise a triable issue of fact as to antitrust injury-in-fact, and that Netflix was thus entitled to summary judgment. As with all federal claims, a plaintiff must establish Article III standing, which requires proof of (1) injury-in-fact, (2) causation, and (3) redressability. Gerlinger v. Amazon.com Inc., 526 F.3d 1253, 1255 (9th Cir.2008). "For Article III purposes, an antitrust plaintiff establishes injury-in-fact when he has suffered an injury which bears a causal connection to the alleged antitrust violation." Id. (internal quotation marks and citation omitted).

In addition to Article III standing, private antitrust plaintiffs must also demonstrate antitrust injury, which is (1) "injury of the type the antitrust laws were intended to prevent" that also (2) "flows from that which makes defendants' acts unlawful." Brunswick Corp. v. Pueblo Bowl-O-Mat, Inc., 429 U.S. 477, 489, 97 S.Ct. 690, 50 L.Ed.2d 701 (1977). This can be established by showing that consumers paid higher prices for a product due to anticompetitive actions of a defendant, such as a horizontal market allocation scheme. See In re Cardizem CD Antitrust Litigation, 332 F.3d 896, 910-11 (6th Cir.2003).

The Subscribers' injury-in-fact theory is that Netflix subscribers paid supracompetitive prices for their DVD-rental subscriptions once Walmart exited the online DVD-rental market pursuant to the allegedly anticompetitive Promotion Agreement. Specifically, the Subscribers contend that Netflix would have reduced its 3U subscription price to $15.99 per month but for Netflix's allegedly anticompetitive conduct. Free from the competitive threat of Walmart, the Subscribers allege, Netflix was able to maintain this artificially high price point.

Applying the standards applicable to antitrust cases, the district court properly concluded that the Subscribers had not raised a genuine issue of material fact as to antitrust injury-in-fact. The Subscribers failed to adduce evidence raising a triable issue of fact that if Walmart remained in the market, Netflix would have reduced its prices.

The undisputed record belies this assertion. Netflix never lowered its 3U price at any time in response to Walmart. Even though Walmart entered the market with a lower price ($18.76 to Netflix's $19.95) for a comparable 3U plan, Netflix did not alter its 3U plan price for a full year after Walmart entered the market. When Netflix eventually did change the price, a year later, it increased the price to $21.99 per month. Netflix also did not reduce its price when Blockbuster offered a 3U plan for $14.99 (while Netflix's was $17.99), even though Blockbuster had a much greater share of the market than Walmart, and even though Netflix rightfully viewed Blockbuster as a competitive threat. Thus, the district court properly determined that no reasonable juror could conclude that Netflix was going to lower its 3U price to $15.99 in response to Walmart when (1) Netflix had never lowered its prices in response to Walmart at any time and (2) Netflix did not lower its price in the face of the $14.99 price cut by Blockbuster, which was objectively a greater competitive threat.

Walmart never had more than 60,000 subscribers, in contrast to Netflix's two million subscribers in mid-2004. Moreover, Walmart gained an average of just 5,000 subscribers per quarter between when it entered the online DVD-rental market in June 2003 and when the Promotion Agreement was announced in March 2005; during this same time period, Netflix was adding 250,000 subscribers per quarter. Thus, Walmart's subscriber market share peaked at 2.4% in early 2004 and declined from there, hitting 1.4% in February 2005. This subscriber decline was most prevalent during the final quarter of 2004, when Walmart lost 7,000 subscribers. Walmart.com's director of entertainment and photo opined that, in mid-2004, it had become clear that Walmart's venture would fail. Walmart attributed this self-described "increased rate[] of attrition" from an already anemic subscriber base to a host of factors, including Walmart's inability to match Netflix's guaranteed 1-to 2-day delivery and Walmart's confusing, poorly designed website.

Not only was Walmart's online DVD-rental business lagging, it was perceived as such by Netflix, Blockbuster, and Amazon. For example, Blockbuster's senior Vice President testified that Blockbuster was not surprised that Walmart exited the business, in part because Walmart was unable to be "the low-cost provider" and received negative reviews from consumers. He concluded that Walmart's exit was "logical." Amazon held similar views. As one Amazon employee stated, he and his colleagues "spent next to no time thinking about Wal-mart" because Walmart wasn't "taking it seriously; they had one distribution center; they had limited selection; they had faulty systems that didn't really work."

The Subscribers' evidence consists of some internal documents produced by Netflix and Walmart employees that purportedly indicate that Walmart was regarded and treated and as a true competitor. For example, the Subscribers note that in November 2003 Walmart described its expectations of growth in the online DVD-rental market over the coming months, and that Hastings said in October 2002 that Walmart's entrance into the DVD-rental market was "unsettling." However, these communications pre-dated Walmart's entry into the market and subsequent poor performance.

The Subscribers also cite a number of documents in which Walmart claimed that its service was successful, such as a "talking points" memo. However, these documents were meant for promotional and motivational purposes, not performance analysis, and the memos do not contain any hard market data. Rather, the documents contain language best described as puffery. The Subscribers also rely on certain news articles that purportedly show that outside observers also considered Walmart a threat. However, many of the documents pre-date Walmart's entry into the market, and others refer to the market challenges posed by Blockbuster, not Walmart. Indeed, Netflix's internal documents indicate that it was analyzing a potential price cut to $15.99 in response to Blockbuster, not Walmart. Further, as the district court emphasized, any Netflix price decrease was "(1) in response to Blockbuster (not Walmart); (2) always couched in terms of possibility; and (3) never actually occurred." Netflix's internal documents show that by late 2004, Netflix treated Walmart as a negligible threat. Indeed, much of the Subscribers' documentary evidence actually supports Netflix's position and convincingly reveals that Walmart did not view itself and was not viewed by others as a competitive threat in late 2004 and early 2005.

The Subscribers also rely on expert testimony. "As a general rule, summary judgment is inappropriate where an expert's testimony supports the nonmoving party's case." Southland Sod Farms v. Stover Seed Co., 108 F.3d 1134, 1144 (9th Cir.1997) (internal quotation marks and citation omitted). See also Dolphin Tours, Inc. v. Pacifico Creative Serv., Inc., 773 F.2d 1506, 1511 (9th Cir.1985) (anticompetitive injury could be inferred from an expert's conclusion that the plaintiff would have attained "a market share of roughly twenty percent" instead of the two percent it did reach). However, the mere proferring of unsupported expert testimony does not create a triable issue as to antitrust injury-in-fact. "In the context of antitrust law, if there are undisputed facts about the structure of the market that render the inference economically unreasonable, the expert opinion is insufficient to support a jury verdict." Rebel Oil Co., Inc. v. Atl. Richfield Co., 51 F.3d 1421, 1435-36 (9th Cir.1995). "Expert testimony is useful as a guide to interpreting market facts, but it is not a substitute for them," and it "has little probative value in comparison with the economic factors that may dictate a particular conclusion." Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209, 242, 113 S.Ct. 2578, 125 L.Ed.2d 168 (1993) (internal quotation marks and citation omitted). We agree with the district court that the Subscribers' experts' testimony is contrary to the undisputed market facts. The opinions are founded on speculation about Walmart's potential to remain in the market based on its general retail strength, untethered to its actual performance in this particular market. The testimony also ignores the fact that the Promotion Agreement allowed Walmart to rent DVDs.3

Gerlinger confirms the conclusion that, as a matter of law, the Subscribers have failed to raise a triable issue of antitrust injury-in-fact. In Gerlinger, Amazon.com and Borders book store entered into an agreement pursuant to which Borders' website directed customers to a site hosted by Amazon.com. 526 F.3d at 1255. In return for referring its online customers to Amazon, Borders received a commission for each book sold and agreed to abandon the online book sales market during the term of the agreement. Id. The plaintiff alleged that the agreement was an unlawful market allocation and that he paid supracompetitive prices as a result of it. Id. Like the Netflix Subscribers, Gerlinger's theory of injury-in-fact was that, if Borders continued competing in the market, book prices would have fallen. Id. But, as here, defendants presented evidence that prices in fact remained the same or even went down following the agreement. Id. This Court concluded that the plaintiff had not raised a triable issue as to whether he would have paid less for a book absent the allegedly anticompetitive agreement. Id. at 1256.

Thus, even considering the facts in the light most favorable to the plaintiffs, the district court properly concluded that the Subscribers did not raise a genuine issue of material fact as to antitrust injury-in-fact. Accordingly, we affirm the district court's summary judgment as to the four Sherman Act claims. Because the Subscribers have not shown injury-in-fact, we need not — and do not — reach the merits of their antitrust claims. See Gerlinger, 526 F.3d at 1256 ("We do not reach the merits, however, because the plaintiff has not shown any injury-in-fact caused by the agreement, and he therefore lacks Article III standing to bring this claim....").

III

Both parties contest the district court's cost award. We review a cost award under the abuse of discretion standard. Arakaki v. Lingle, 477 F.3d 1048, 1069 (9th Cir.2007). "A district court abuses its discretion if it does not apply the correct law or if it rests its decision on a clearly erroneous finding of material fact." Jeff D. v. Otter, 643 F.3d 278, 283 (9th Cir.2011) (internal quotation marks and citation omitted). However, we review the threshold question of whether the district court has the authority to award costs de novo. Russian River Watershed Protection Comm. v. City of Santa Rosa, 142 F.3d 1136, 1144 (9th Cir.1998).

During discovery, the Subscribers sought production of electronically stored information. The Subscribers required that electronic information other than spreadsheets be produced in the static Tagged Image File Format ("TIFF").4 The Subscribers also requested that these images contain searchable text and relevant metadata,5 and that the produced documents be numbered sequentially and include certain identifying information. Netflix enlisted electronic discovery vendors to assist with responding to the Subscribers' discovery requests.

As can be expected from a case of this magnitude, discovery was extensive, and Netflix ultimately produced almost 15 million pages in response to the Subscribers' discovery requests. After summary judgment was granted, Netflix requested $744,740.11 as costs for discovery-related tasks, and was ultimately awarded $710,194.23 by the district court. The Subscribers appeal portions of that award, and Netflix cross-appeals.

The Subscribers argue that the district court (1) erred by broadly construing § 1920(4) in its taxing of e-discovery and data management costs totaling $317,616.69, and (2) abused its discretion in taxing consulting fees, TIFFs, and copying costs totaling $245,471.31. Netflix cross appeals, arguing that the district court abused its discretion in disallowing $21,000 in costs to copy certain PowerPoint files.

A

The district court's decision rested upon a "broad construction of section 1920 with respect to electronic discovery production costs." That determination was largely founded on our decision in Taniguchi v. Kan Pacific Saipan, Ltd., 633 F.3d 1218, 1221 (9th Cir.2011), which was subsequently reversed by the Supreme Court, ___ U.S. ___, 132 S.Ct. 1997, 182 L.Ed.2d 903 (2012). In light of the intervening Supreme Court precedent, we vacate the award of some of the costs the Subscribers have challenged as "non-recoverable" and remand for further consideration of some of those costs pursuant to a narrow construction of § 1920(4).

1

In awarding costs, the district court explicitly adhered to a broad interpretation of § 1920(4) pursuant to our decision in Taniguchi. After the Supreme Court's reversal of Taniguchi, we must reassess what electronic discovery costs may be properly taxed as costs of "making copies of any materials where the copies are necessarily obtained for use in the case" under 28 U.S.C. § 1920(4). In conducting this assessment, we find persuasive the reasoning of the Third Circuit in Race Tires, the Fourth Circuit in Country Vintner, and the Federal Circuit in CBT Flint Partners, LLC v. Return Path, Inc., 737 F.3d 1320 (Fed.Cir.2013). We are also guided by the Supreme Court's reiteration, in Taniguchi, of the limited reach of § 1920. In that decision, the Supreme Court reversed our determination that costs of a translator of written documents can constitute the costs of an "interpreter" under § 1920(6). 132 S.Ct. at 2005. In doing so, the Court underscored "the narrow scope of taxable costs" and reminded us that "[t]axable costs are limited to relatively minor, incidental expenses as is evident from § 1920." Id. at 2006.

In establishing the boundaries that must be given to § 1920(4), some historical perspective is useful. "Although the taxation of costs was not allowed at common law, it was the practice of federal courts in the early years to award costs in the same manner as the courts of the relevant forum State." Taniguchi, 132 S.Ct. at 2001 (citing Alyeska Pipeline Serv. Co. v. Wilderness Soc'y, 421 U.S. 240, 247-48, 95 S.Ct. 1612, 44 L.Ed.2d 141 (1975)). This diversity of rules led to a great diversity of awards, which in turn led to some losing litigants being "unfairly saddled with exorbitant fees." Alyeska, 421 U.S. at 251, 95 S.Ct. 1612.

"In 1853, Congress undertook to standardize the costs allowable in federal litigation." Id. In doing so, they sought to "simplify the taxation of fees, by prescribing a limited number of definite items to be allowed." Country Vintner, 718 F.3d at 255 (quoting Cong. Globe, 32nd Cong., 2d Sess.App. 207 (1853) (statement of Sen. Bradbury)). The result was the Fee Act of 1853, ch. 80, 10 Stat. 161, which was a predecessor to § 1920. The Fee Act "depart[ed] from the English practice of attempting to provide the successful litigant with total reimbursement." Race Tires, 674 F.3d at 164 (quoting 10 Charles Alan Wright, Arthur R. Miller, Mary Kay Kane & Richard L. Marcus, Federal Practice and Procedure § 2665 (3d ed.1998)). The Supreme Court has since held that Congress intended with the Fee Act to "impose rigid controls on cost-shifting in federal courts." Crawford Fitting Co. v. J.T. Gibbons, Inc., 482 U.S. 437, 444, 107 S.Ct. 2494, 96 L.Ed.2d 385 (1987). Thus, § 1920 "define[s] the full extent of a federal court's power to shift litigation costs absent express statutory authority to go further." W. Va. Univ. Hosps., Inc. v. Casey, 499 U.S. 83, 86, 111 S.Ct. 1138, 113 L.Ed.2d 68 (1991).

The language of § 1920(4) first appeared in § 3 of the Fee Act, stating that "lawful fees for exemplifications and copies of papers necessarily obtained for use on trial... shall be taxed by a judge or clerk of the court." 10 Stat. at 168. The language was altered over the years to apply to "cases," not just trials, and courts were later given discretion to award the costs, rather than being mandated to do so. Race Tires, 674 F.3d at 165. The statute originally applied to making only "copies of paper," but now applies to the "costs of making copies of any materials." Judicial Administration and Technical Amendments Act of 2008, Pub.L. No. 110-406, § 6, 122 Stat. 4291, 4292; see also In re Ricoh Co., Ltd. Patent Litig., 661 F.3d 1361, 1365 (Fed.Cir.2011) ("[E]lectronic production of documents can constitute ... `making copies' under section 1920(4).").

As a general rule, costs and fees should be awarded to the prevailing party. Fed. R.Civ.P. 54(d)(1) ("Unless a federal statute, these rules, or court order provides otherwise, costs — other than attorney's fees — should be allowed to the prevailing party."). However, a district court's discretion to award costs is limited to particular types of costs enumerated in 28 U.S.C. § 1920. See Crawford Fitting, 482 U.S. at 441, 107 S.Ct. 2494 ("[Section] 1920 defines the term `costs' as used in Rule 54(d).").

As with all statutory construction questions, we "begin with the language employed by Congress and the assumption that the ordinary meaning of that language accurately expresses the legislative purpose." FMC Corp. v. Holliday, 498 U.S. 52, 57, 111 S.Ct. 403, 112 L.Ed.2d 356 (1990) (internal quotation marks omitted). Section 1920(4) provides that a judge or clerk may tax "the costs of making copies of any materials where the copies are necessarily obtained for use in the case." We first focus on the phrase "making copies." 28 U.S.C. § 1920(4). Because this language is not defined in the statute, we apply "its ordinary meaning." Taniguchi, 132 S.Ct. at 2002.

As the Fourth Circuit explained in Country Vintner, "`[c]opies' has appeared in the taxation statute since its enactment in 1853, when `copy' meant a `transcript,' a `writing like another writing,' or an `imitation.'" 718 F.3d at 258 (footnotes omitted). The noun retains an almost identical meaning today.6 The definitions for the verb "make" relevant to the context of § 1920(4) include "to cause to exist, occur, or appear" and "to bring (a material thing) into being by forming, shaping, or altering material."7 These definitions are consistent with our previous observation that "[s]ection 1920(4) speaks narrowly of `[f]ees for exemplification and copies of papers,' suggesting that fees are permitted only for the physical preparation and duplication of documents, not the intellectual effort involved in their production." Romero v. City of Pomona, 883 F.2d 1418, 1428 (9th Cir.1989), abrogated in part on other grounds by Townsend v. Holman Consulting Corp., 929 F.2d 1358, 1363 (9th Cir.1991) (en banc); see also Zuill v. Shanahan, 80 F.3d 1366, 1371 (9th Cir. 1996).

Section 1920(4) further defines the awarding of costs for making copies, requiring that recoverable costs be restricted to the making of copies "necessarily obtained for use in the case." Applying the statutory construction maxim expressio unius est exclusio alterius (the express mention of a thing implicitly excludes others in its class), we presume that Congress recognized that costs can also be incurred in litigation for making copies that are not necessarily obtained for use in the case, and that such costs are not taxable.

Nevertheless, the statute is not so restrictive as to "specifically require that the copied document be introduced into the record to be an allowable cost." Haagen-Dazs Co., Inc. v. Double Rainbow Gourmet Ice Creams, Inc., 920 F.2d 587, 588 (9th Cir.1990). Section 1920(4) allows for the recovery of costs where the copies were obtained to be produced pursuant to Rule 34 or other discovery rules. See Country Vintner, 718 F.3d at 257 n. 9 (noting decisions of the Federal Circuit — applying the law of this circuit — as well as the Fifth, Seventh, and Eleventh Circuits recognizing that costs are recoverable under § 1920(4) for copying costs for document production).

The faithful production of electronically stored information may require processes such as optical character recognition (which renders material text-searchable), preservation of metadata, and conversion to a non-editable file format. Parties might agree to employ a particular file format or methodology for electronically stored information production, or the court might order them to produce electronically stored information with certain characteristics. See In re Ricoh, 661 F.3d at 1365 (parties agreed that a third party vendor would process and store e-mails in a secure document review database). The Federal Circuit held in CBT Flint Partners that

737 F.3d at 1328. See also Country Vintner, 718 F.3d at 260 n. 19 ("If, for instance, a case directly or indirectly required production of [electronically stored] unique information such as metadata, we assume, without deciding, that taxable costs would include any technical processes necessary to copy [electronically stored information] in a format that includes such information."). When copies are made in a fashion necessary to comply with obligations such as these, costs are taxable so long as the copies are also "necessarily obtained for use in the case."

That § 1920(4) restricts the award of costs to those incurred for copies necessarily obtained for use in the case is not, in itself, a justification permitting the award of costs for any task necessary to the prosecution or defense of a case. As noted by the Third Circuit in Race Tires, "[s]ection 1920(4) does not state that all steps that lead up to the production of copies of materials are taxable." 674 F.3d at 169; see also CBT Flint Partners, 737 F.3d at 1328 ("[O]nly the costs of creating the produced duplicates are included [as recoverable], not a number of preparatory or ancillary costs commonly incurred leading up to, in conjunction with, or after duplication."). The Third Circuit further explained, again in the context of producing electronically stored information, that "[i]t may be that extensive `processing' of [electronically stored information] is essential.... But that does not mean that the services leading up to the actual production constitute `making copies.'" 674 F.3d at 169.

The proper application of a narrowly construed § 1920(4) requires that the tasks and services for which an award of costs is being considered must be described and established with sufficient specificity, particularity, and clarity as to permit a determination that costs are awarded for making copies. "`Document production' and other similarly generic statements on the invoices are unhelpful in determining whether those costs are taxable." In re Ricoh, 661 F.3d at 1368. Further, a description of a task is useful only to the extent it accurately reflects the task for which copying costs are sought.

2

In support of its opposition to the Subscribers' motion for the district court to review costs, Netflix submitted the declaration of Vivian Liu-Somers, a project manager for Esquire Litigation Solutions, a vendor that provided litigation support services to Netflix. In her declaration, Liu-Somers identified 44 different charges appearing on Esquire Litigation's invoices, and combines those charges into 21 groups. For each of the 21 groups, Liu-Somers provided a brief description of the services to which the charges refer.

The Subscribers challenge certain of these groups of charges as not taxable under § 1920(4), further combining them into four broader task categories: (1) "data upload," (2) "endorsing," (3) "keyword," and (4) "professional services." The Subscribers also challenge a charge invoiced by SFL Data, a different litigation support vendor providing services to Netflix, for "Electronic Data Discovery" tasks. We review each of the five challenged categories of charges asserted by the Subscribers in turn.

a

As challenged by the Subscribers, the "data upload" category of charges refers to two different groups of charges described by Liu-Somers in her declaration. She grouped the charges of "Data Upload," and "Catalyst Data Upload," and indicated that these charges referred "to the reproduction of documents for potential production into a database where they could be viewed." Separately, Liu-Somers described the charge of "Upload Production Documents" as referring "to the process of reproducing the collection of the documents actually being produced for viewing after all the processes necessary to prepare the documents in the required formats and with the required labels have been completed."

The Subscribers' challenge to these costs rests on the use of the word "upload" in the description of the charge. They note that an "upload" of electronically stored information is the movement of data from one location to another, and indicates the data is being transmitted.8 The Subscribers argue that the costs incurred for such a task are not recoverable under § 1920(4) because the task is "akin to the costs of moving boxes of information from client site to law firm to the room where reviewers would review them." Focusing on the word "reproduction" in both of Liu-Somers' descriptions of the services performed, Netflix responds that the task of uploading data was not merely for moving data, but necessarily involved "making copies." Neither argument fully responds to the question of whether the charges are taxable under § 1920(4).

Consideration of whether certain tasks are taxable pursuant to § 1920(4) "calls for some common-sense judgments guided by a comparison with the paper-document analogue." CBT Flint Partners, 737 F.3d at 1331. Instead of moving boxes of information, an upload of data is more akin to the paper-document analogue of faxing a document from the client site to the law firm, a process which involves the transmitting of data from location to location and which also results in a facsimile copy of the original document.

The cost of making the copy is not rendered non-taxable merely because it was created by using fax machines rather than a photo-copier. Conversely, § 1920(4) does not award costs merely because a process resulted in the creation of a copy. That the fax process creates a copy does not, by itself, establish that the cost is taxable under § 1920(4). Rather, a further determination is required: whether the copy was necessarily obtained for use in the case. Like any other copy, if the faxed copy was created to be produced in discovery, the cost of making the fax would be taxable under § 1920(4). However, if the faxed copy was created solely for the convenience of counsel, the cost of making the copy would not be taxable.

Liu-Somers' description of the "data upload" and "catalyst data upload" charges indicates the uploading process created new copies of documents inside a database. Assuming, without deciding, that the specific uploading task constituted "making copies," the further determination is required whether the copies were necessarily obtained for use in the litigation. Liu-Somers declared that "the reproduction of documents was a necessary step in the document production process because it facilitated a selection of documents for production from the set of documents for potential production." This description establishes only that the copies were essential to the document production process, and fails to establish the copies were necessarily obtained for use in the litigation.

A narrow construction of § 1920(4) requires recognition that the circumstances in which a copy will be deemed "necessarily obtained" for use in a case will be extremely limited. That a chosen "document production process" requires the creation of a copy does not establish that the copy is necessarily obtained for use in the case. A lawyer may review electronically stored information for privilege either by viewing the original documents on the client's computer or, alternatively, by viewing copies uploaded to the lawyer's computer. Although the latter method of review requires the creation of a copy, the ability to conduct the review by looking at the original document establishes that the uploaded copy was not necessarily obtained for use in the case.

Similarly, Liu-Somers' description of the charges for "Upload Production Documents" indicates that the copies were created as a "necessary step in the document production process in order to view the documents as they appeared in the actual production being made." As with Netflix's description of the other "uploading" charges, this description establishes only that the copy is essential to the document production process that Netflix (or its litigation support vendor) elected to employ, and fails to establish the copies were necessarily obtained for use in the case. Accordingly, these charges are non-taxable under § 1920(4).

b

The Subscribers also challenge what they describe as "endorsing" activities. However, the only indication in the Subscribers' opening brief that they have even challenged the award costs for "endorsing" activities is the cursory mention, in their Statement of Facts, that "Netflix incurred $21,134.82 for `Endorsing' tasks, which included the `branding of image files with unique sequential production numbers and confidentiality designations.'" Because these matters were not "specifically and distinctly argued" in the open briefing, we will not consider them. Christian Legal Soc'y Chapter of Univ. of Cal. v. Wu, 626 F.3d 483, 487 (9th Cir.2010).

c

The Subscribers challenge the cost award for what they term "keywording" activities. As described by Liu-Somers in her declaration, Esquire Litigation used a variety of terms to charge Netflix "for the use of automated software processes to reproduce the set of documents for potential production into a reduced set of documents that did not include certain types of documents that did not need to be produced." Netflix attempts to shoehorn the filtering process into the ordinary meaning of "making copies" by arguing that filtering is "simply a mechanical process of making a copy of all documents that fit the supplied criteria." The argument reveals its own flaw, disclosing that the charge was incurred for two separate tasks: (a) identifying the documents that fit the supplied criteria, and then (b) making copies of those documents. The former task is akin to a person (lawyer, paralegal, or otherwise) mechanically reviewing a stack of documents and (based upon criteria supplied by a lawyer) separating them into two piles: one consisting of documents that might be potentially be produced, and the other consisting of documents that will not be produced. Netflix argues that its vendor "simply performed the mechanical process of applying the criteria it received from Netflix's attorneys to the documents being reproduced for production in the required formats" (emphasis added). The argument is contradicted by Netflix's implicit acknowledgment that the filtering process was also applied to documents that were not copied. As such, the application of automated software filtering processes to identify which documents to copy and which documents to not copy is not taxable.9

d

The Subscribers challenge the cost award for "professional services." Absent from Liu-Somers' declaration are descriptions for any charges titled "Professional Services," indicating the Subscribers' label for this category amounts to a generic statement "unhelpful in determining whether those costs are taxable." In re Ricoh, 661 F.3d at 1368. In their opening brief, the Subscribers further describe this category as including "activities like processing, native review, data analysis, project management, and production services." Liu-Somers' declaration demonstrates that the Subscribers have grouped an extremely broad range of activities into a single category. The declaration includes descriptions for a variety of narrower categories, ranging from "the imaging of the documents to create an electronic `page' ready for bates and confidentiality branding and redactions" through "prepar[ing] the documents for production in the required formats." Netflix's response is equally generic and unhelpful, asserting that "these are all costs required in order to make copies of the information being produced" and "are akin to the costs of copy center employees, photocopy technicians, copier repairmen, and other overhead costs included with the flat per page rate charged by traditional outside photocopy vendors — costs which have always been deemed recoverable." The record before us leaves us unable to resolve whether any of the large variety of specific charges that the Subscribers broadly challenge as "professional services" are taxable under a narrow construction of § 1920(4). Thus, we must remand the issue to the district court for its determination in the first instance.

e

Finally, the Subscribers challenge a specific item on one invoice from SFL Data, asserting the charge was for "native review processing." Netflix counters that the invoice actually states that the "cost involved the copying of nearly 80GB of data for production constituting 167,311 documents." The $10,000 cost appears to have been incurred (at a flat rate per agreement with Netflix) for a variety of different tasks, including native review processing, optical character recognition, exporting documents, converting documents to TIFF, populating custom fields, and prepping for further processing. Although the cost incurred for some of these tasks appears to be taxable, the present record does not permit a conclusion that all of the tasks for which SFL charged Netflix a flat rate of $10,000 are taxable. To the extent the invoiced tasks exceeded optical character recognition, conversion to TIFF, and other activities essential to the making of copies necessary to the case, they are not taxable.

3

In summary, we conclude that of the $317,616.69 challenged by the Subscribers as non-taxable under § 1920(4), only those costs attributable to optical character recognition, converting documents to TIFF, and "endorsing" activities — all of which were explicitly required by Subscribers — are recoverable on the record before us. We therefore affirm the cost award in part, and vacate it in part. We remand for taxing of costs in accordance with this opinion.

B

The district court did not abuse its discretion in awarding $245,471.31 in consulting fees, TIFF images, and copying costs.

1

The Subscribers first complain that the district court failed to explain its findings adequately. However, "a district court need not give affirmative reasons for awarding costs; instead, it need only find that the reasons for denying costs are not sufficiently persuasive to overcome the presumption in favor of an award." Save Our Valley v. Sound Transit, 335 F.3d 932, 945 (9th Cir.2003).

Here, the district court explained in its order that it "read the parties' papers and carefully considered their arguments and the relevant legal authority." Further, the order specifically identified the Subscribers' arguments concerning "TIFF conversion costs; copying/'blowback' costs ... documents productions purportedly not delivered; professional fees re visual aids." Under our deferential standard of review, the district court's explanation was sufficient. See Save Our Valley, 335 F.3d at 945 ("[W]e have never held that a district court must specify reasons for its decision to abide the presumption and tax costs to the losing party." (emphasis in original)).

2

The Subscribers next contend that the costs claimed for preparing visual aids were actually unrecoverable consulting fees, relying on invoice descriptions of the title of the person performing the work. To be sure, "[f]ees for exemplification and copying are permitted only for the physical preparation and duplication of documents, not the intellectual effort involved in their production." Zuill v. Shanahan, 80 F.3d 1366, 1371 (9th Cir.1996) (emphasis added) (internal quotation marks and citation omitted). However, there is no authority for the proposition that the title of the person doing the work is relevant to classifying the type of work actually done. In fact, courts have held otherwise. See Race Tires, 674 F.3d at 169 ("Neither the degree of expertise necessary to perform the work nor the identity of the party performing the work of `making copies' is a factor that can be gleaned from § 1920(4)."). In addition, evidence was presented by Netflix that the vendors were only paid to perform the tasks associated with production and not for creating the substantive content of the visual aids. Thus, the record supports the district court's conclusion that the tasks done by the consultants were not "intellectual effort," regardless of the job title listed on the invoices. The district court did not abuse its discretion in awarding $14,355.50 in costs for preparation of visual aids.

3

The district court did not abuse its discretion in awarding costs for TIFF conversions. The Subscribers argue that they were taxed $167,399.70 in excessive TIFFing costs because Netflix paid an unreasonable amount per TIFF page. The Subscribers' expert concluded that the rate of seven cents per page of TIFF conversion was unreasonable because it was above market and because Netflix's first invoice from its e-discovery vendor charged only two cents per page. Netflix's expert testified that the costs Netflix charged were reasonable. He testified that Netflix reviewed proposals from six electronic discovery vendors and that the firm that Netflix eventually went with offered the lowest prices to convert documents to TIFF images out of the six solicited firms. Netflix's expert further explained that the two-cent rate was charged only for a certain type of conversion — PDF to TIFF — and that the total bill for this batch was a mere $54.08.

The district court fully considered the two expert opinions and decided in favor of Netflix. Given our deferential standard of review, there is no basis to disturb that conclusion.

The Subscribers also contend that Netflix was awarded costs for documents unnecessarily produced, resulting in an overcharge of $46,773.71. However, the record does not support that conclusion, and the district court did not abuse its discretion in awarding these TIFF-related costs.

4

The district court did not abuse its discretion in awarding $16,942.40 for copying paper documents. The Subscribers contend that the documentation for these costs was inadequate. However, the invoices at issue indicate the purpose of the charge, such as printing exhibits and copying deposition transcripts, as well as the dates the work was done. Moreover, Netflix supported its bill of costs with a declaration from one of Netflix's attorneys, who clarified that certain documents were produced as "exhibits to depositions" and "disclosure or formal discovery documents." Netflix provided sufficient information for the district court to identify the documents being reproduced and, ultimately, to determine which costs were taxable. The district court did not abuse its discretion in awarding these costs.

C

The district court also did not abuse its discretion in declining to award Netflix $21,000 for producing certain black and white PowerPoint documents. The Subscribers had requested "single-page Group IV TIFF files," which is a black and white version, and separately requested that all documents "be produced in the same order as they are kept or maintained by you in the ordinary course of your business." In the usual course of business, Netflix maintained the PowerPoint in color. However, throughout discovery, Netflix produced black and white PowerPoint presentations. At some point during the litigation, Netflix submitted to the court a color version of a PowerPoint presentation. When the Subscribers learned that were colored slides available, they requested them, arguing that Netflix's failure to do so previously violated the instruction to produce documents as they were normally kept by Netflix. Because Netflix maintained colored slides in the ordinary course of business, the district court concluded that the Subscribers were entitled to production of the documents under their previous discovery requests. Although production of color slides was in some sense duplicative of the previously-produced black and white slides, the district court did not abuse its discretion in denying the cost award given the separate discovery requests, and the fact that Netflix likely could have avoided the duplicative production by first producing the color documents maintained in the usual course of business. The district court was not confused about the issue, as Netflix claims. To the contrary, the record reflects that the district court understood the issue and decided under the circumstances that a cost award was appropriately denied. The district court did not abuse its discretion in doing so.

IV

In sum, we affirm the district court's grant of summary judgment on the antitrust claims, for lack of antitrust injury-in-fact. We affirm the cost award in part and reverse it in part. We need not, and do not, reach any other issue urged by the parties.

AFFIRMED IN PART; VACATED IN PART; REMANDED.

Each party shall bear its own costs on appeal.