A recent opinion by Judge DuBois in Comcast Cable Comm'ns, LLC v. Sprint Comm'ns Co., LP, Civil Action No. 12-859 (E.D. Pa. Nov. 21, 2016), denies each party's motion to exclude the other's expert witness on damages. (According to the opinion, plaintiff Comcast's expert, Michele Riley, is prepared to testify that a reasonable royalty for the patent in suit is $123,253,162, while defendant Sprint's expert, Dr. Alan Cox, is of the opinion that an appropriate royalty would fall somewhere between $300,000 and $1,500,000. I'll have a little more to say about this range at the end of the post.) The opinion touches on a number of important issues relating to the calculation of reasonable royalties.
The patent in suit, "titled 'Transferring of a Message,' claims a method 'for inquiring about information relating to a [wireless] terminal of a cellular network from the cellular network, from a messaging server external to the cellular network'” (p.4). Comcast sought to exclude Dr. Cox's testimony for several reasons, including (1) his use of forward citation analysis--"a method of estimating the value of a particular patent based on the number of times the patent is cited by later patents"--to "'corroborate' his overall opinion on the value of reasonable royalty to the '870 Patent"; (2) his use of a license agreement between Comcast and Celltrace as a comparable; and (3) his use of three other Comcast licenses as support for the proposition that Comcast prefers lump-sum licenses. On the first issue:
Dr. Cox compiled a pool of patents that are technologically similar to the ‘870 patent, based on International Patent Classification system labels, and published within six months before and after publication of the ‘870 patent. . . . Cox next determined how many times each patent in the resultant pool was cited by later patents. . . . Using this data, Dr. Cox determined the “percentile ranking” of the ‘870 patent—that is, “the percentage of categorically similar patents that had less forward citations” than the patent in question. . . . Dr. Cox performed the same analysis for each of the U.S. granted patents that were part of the Nokia-Comcast Agreement, under which Comcast purchased 36 patents from Nokia, including the ‘870 patent. . . . Utilizing this process, Dr. Cox determined the percentile ranking of each patent in the Nokia-Comcast Agreement, based on which he approximated the value of each patent purchased under that Agreement relative to the other patents covered by the Agreement. . . . This process resulted in Dr. Cox’s opinion that the ‘870 patent “represent[ed] 2.5 percent of the total value of all patents in the Nokia-Comcast [Agreement.” . . . Dr. Cox then multiplied that 2.5 percent by the total purchase price of the Nokia-Comcast Agreement ($600,000) to estimate the value of the ‘870 patent at $15,000 (pp. 8-9).
Sprint argued that forward citation analysis has been "discredited," but the court disagrees:
In advancing this argument, Comcast relies on a recent trial court decision which it claims rejected the forward citation method as unreliable under Daubert. Comcast Mot. to Exclude at 14; Finjan, Inc. v. Blue Coat Sys., Inc., No. 13-CV-03999-BLF, 2015 WL 4272870 (N.D. Cal. July 14, 2015). But Finjan does not reject forward citation analysis outright—rather, that case recognizes that “a qualitative analysis of asserted patents based upon forward citations may be probative of a reasonable royalty in some instances.” Finjan, 2015 WL 4272870 at *8. Comcast also points to a recent paper published by researchers at the University of Pennsylvania in support of its argument. Comcast Mot. to Exclude at 14, Ex. 21 (“Penn Paper”).7/
7/ David S. Abrams, Ufuk Akeigit, and Jillian Popadak, Patent Value and Citations: Creative Destruction or Strategic Disruption?, U. Penn. (2013), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2351809.
The authors of the Penn Paper conclude some of the patents with the highest lifetime revenues have fewer citations than patents with median revenues. Penn Paper at 1. However, the forward citation method of analysis has been recognized in the academic literature as reliable since the 1990s. . . . Indeed, one meta-analysis of published research on forward citation analysis, which included the Penn Paper, found “forward citation intensity is, in fact, correlated with economic value.” . . . In short, courts have not rejected forward citation analysis outright. This Court determines that a single academic paper—the Penn Paper—is not sufficient to rebut decades of literature supporting forward citation analysis (pp. 9-10).
The court also rejects the argument that it Dr. Cox’s use of forward citation analysis was insufficiently reliable because he "only counted citations to the fully published patent," and not "citations to the patent application or the patent’s international counterparts," stating that this method "is supported by academic literature" and that two cases Comcast cited in its support were distinguishable:
. . . In Finjan, the expert was excluded because he failed to “tie the methodology to the facts of the case” and failed to consider “potential problems” with his method of forward-citation analysis. Finjan, 2015 WL 4272870 at *8. For example, “many of the [p]laintiff’s patents reference one another,” and the court observed that patent value should not be based on “the number of times an inventor cites himself in prosecuting related patents.” Id. In contrast, Dr. Cox tied his analysis to the facts in this case by adjusting the forward citation method to account for the age and category of the ‘870 patent and the other patents covered by the Nokia-Comcast Agreement. . . . And in Oracle, the District Court excluded expert testimony based on forward citation analysis because the expert failed to count citations to a patent’s two predecessor patents. Oracle at *2. But on this issue Comcast does not argue that Dr. Cox ignored any predecessors of the ‘870 patent. . . (pp. 10-11).
Second, the court rejected Comcast's arguments regarding Dr. Cox's use of an agreement between Sprint and another company as a comparable:
To corroborate his damages opinion, Dr. Cox considered a 2010 agreement between Sprint and Celltrace (hereinafter “Sprint-Celltrace Agreement”), granting Sprint a perpetual license to use the technologies encompassed in three patents owned by Celltrace in exchange for a lump sum payment of $1.5 million. . . One of these patents, U.S Patent No. 6,011,976 (“the ‘976 patent”), shares an IPC code with the ‘870 patent, meaning it is technologically similar. . . . According to Dr. Cox, that similar patent could therefore account for an equal portion of the total price, $500,000, or, though less likely, the full $1.5 million price. . . . Dr. Cox conducted a forward citation analysis of these three patents, and using their comparative value, he approximated the portion of the Sprint-Celltrace Agreement purchase price attributable to the ‘976 patent at $1,000,000. . . . Dr. Cox states that this corroborates his overall opinion that a hypothetical negotiation for a license for the ‘870 patent would be in the range of $300,000 to $1,500,000. . . .
Comcast argues that Dr. Cox’s use of the Sprint-Celltrace Agreement is not reliable because the circumstances of the Sprint-Celltrace Agreement are not comparable to those of the ‘870 patent. . . . In support of this argument, Comcast states that the United States Court of Appeals for the Federal Circuit has held that experts may consider settlement agreements only when they constitute “the most reliable license in the record.” Id. (citing LaserDynamics, Inc. v. Quanta Comput. Inc., 694 F.3d 51, 77 (Fed. Cir. 2012)). But Comcast’s interpretation of LaserDynamics strains the Federal Circuit’s decision. In LaserDynamics, the Federal Circuit did not limit experts to consideration of “the most reliable license in the record.” LaserDynamics at 77–78. Rather, that Court reversed the trial court because it admitted into evidence “the least reliable license [in the record] by a wide margin” where there were twenty-nine other licenses involving the patent-in-suit, many of which were “far more reliable.” Id.
Most importantly, the decision in LaserDynamics concerned the admissibility at trial of the settlement agreement under Federal Rule of Evidence 403. . . . Thus, even if the Sprint-Celltrace Agreement were itself inadmissible under LaserDynamics, Federal Rule of Evidence 703 permits an expert such as Dr. Cox to rely on inadmissible evidence if experts in the field would reasonably rely on such evidence.
In contrast to the excluded license in LaserDynamics, the Sprint-Celltrace Agreement is not the “least reliable license” in the record, and the record is not replete with “far more reliable” licenses. LaserDynamics at 77–78. Rather, there are many similarities between the Sprint-Celltrace Agreement and the hypothetical negotiation—for example, the patents are technologically comparable; that Agreement involved a lump-sum payment, and Sprint has a preference for such payments; and both involve a non-exclusive license agreement. . . And although the comparison is not perfect, Dr. Cox acknowledges and “account[s] for” the differences. . . .
For example, Comcast notes that in negotiating the Sprint-Celltrace Agreement, Sprint argued invalidity and noninfringement, whereas the hypothetical negotiation requires Dr. Cox to assume the negotiating parties consider the patent at issue to be valid and infringed. . . . Comcast argues that the Sprint-Celltrace Agreement is therefore a “non-comparable agreement.” Comcast Mot. at 17. But by concluding that the hypothetical negotiation would result in “the upper range” of the Sprint-Celltrace Agreement, Dr. Cox has, in effect, accounted for this difference. . . . This conclusion is based on the fact that, because Sprint argued invalidity and noninfringement in negotiating the Sprint-Celltrace Agreement, Sprint was likely motivated to seek a lower license price than a reasonable licensee who, as in the hypothetical negotiation, admitted validity and infringement. Therefore, a reasonable licensee in the hypothetical negotiation would be less motivated to negotiate a lower price than Sprint was in the Sprint-Celltrace Agreement. Because a reasonable licensee would be less motivated to seek a lower license price, Dr. Cox concludes that such a licensee “would have negotiated a price in the upper range of the $500,000 - $1,500,000 spread” of the Sprint-Celltrace Agreement. . . .
Whether the Sprint-Celltrace Agreement is “sufficiently comparable such that [Dr. Cox’s] calculation is a reasonable royalty goes to the weight of the [expert testimony], not its admissibility” under Daubert and Federal Rule of Evidence 702. . . (pp. 12-14).
Third, the court accepts the use of three other Comcast licenses as evidence that Comcast prefers lump-sum licenses to running royalties (pp. 14-15).
Sprint's motion to exclude Ms. Riley's testimony centered on the issue of apportionment:
Because of the complexity of the ‘870 patent and Sprint’s alleged infringing process, Ms. Riley uses apportionment as part of her reasonable royalty determination. Comcast alleges that several of Sprint’s SMS and MMS systems infringe various claims of the ‘870 patent. Therefore, to calculate a reasonable royalty, Ms. Riley first calculated the profitability of Sprint’s SMS and MMS systems. . . . Ms. Riley then referred to [technical expert] Dr. Akl’s report in which he “separated the overall process of each specific messaging call flow into steps” based on infringing and non-infringing steps. Sprint Mot. to Exclude at 7–8. Finally, using Dr. Akl’s steps, Ms. Riley apportioned the share of Comcast’s SMS and MMS profitability that can be attributed to the alleged infringement of the ‘870 patent. . . .
Sprint argues that the Court should prohibit Ms. Riley from using apportionment based on counting steps as unreliable under Daubert because courts have rejected similar methods. . . . For example, Sprint notes that the Federal Circuit has observed that apportionment “cannot be reduced to a mere counting of lines of code.” Lucent Tech., Inc. v. Gateway, Inc., 580 F.3d 1301, 1333 (Fed. Cir. 2009). But that observation is taken out of context—the Lucent Tech. court goes on to observe that “the glaring imbalance between infringing and non-infringing features must impact the analysis of how much profit can properly be attributed” to the infringing technology. Id. The Lucent court does not state that counting lines of code is per se unreliable; rather, it recognizes that such simplistic apportionment alone is not probative of value.
Sprint cites three other cases for the proposition that an expert’s opinion is not reliable when the expert apportions damages based on the features in the patent at issue, as Ms. Riley did in this case. Sprint Mot. to Exclude at 12–13; Stragent, LLC v. Intel Corp., 2014 WL 1389304, at *3 (E.D. Tex. Mar. 6, 2014); Good Tech. Corp. v. MobileIron, Inc., 2015 WL 4090431 (N.D. Cal. July 5, 2015); Computer Assocs. Int’l, Inc. v. Altai, Inc., 775 F. Supp. 544, 571–72 (E.D.N.Y. 1991). These cases are no more persuasive than Lucent. In Stragent, the court focused on the fact that the expert’s apportionment relied on “arbitrary assumptions that have no basis in the facts of [the] case.” Stragent at *4. In Good Tech, the court excluded the expert’s opinion because the expert had conducted “no investigation into whether any of the criteria is more important than others, or how strongly each criterion is tied to the patents.” Good Tech at *7. And Computer Associates mentions the counting of lines of code only in passing. None of the cases support Sprint’s assertion that counting components or features is per se unreliable, and this Court declines to so hold (pp. 16-18).
The court also rejects the argument that Ms. Riley shouldn't have relied on Dr. Akl's opinion because (according to Sprint) Akl "'manipulated' Sprint’s call flow diagrams by combining certain steps into a single step and omitting certain steps so as to reduce the number of non-infringing steps; and simultaneously adding steps to increase the number of infringing steps," and because he "improperly . . . gave equal weight, and therefore equal value, to each step that he identified" (p.18). According to the court, Dr. Akl "adequately supported and explained in detail the reasoning behind each part of his step-counting process . . . " (p.19).
Four matters I'd like to note before concluding this post. First, it's interesting to see that the purchase price for the portfolio of 36 patents of which the patent in suit was part was only $600,000. As Ted Sichleman notes in the paper of his that I blogged about last week:
. . . Some defendants have argued that the cost of acquiring a patent or portfolio should set an upper bound to the amount of damages collectible in litigation. For example, in a recent case an accused infringer argued to bar the request of a large patent aggregator, Intellectual Ventures, for over $300 million in licensing fees on the ground that it acquired the asserted patent for only $750,000. See Dan Levine & Tom Hals, Exclusive: Intellectual Ventures Faces Novel Attack on Patent Business, Reuters, Oct. 29, 2013, http://www.reuters.com/article/us-intellectual-ventures-lawsuit-idUSBRE99S05120131029. However, these arguments differ from the ones here in that the appropriate level of R & D, commercialization, and opportunity costs should be determined ex ante, from the perspective of an innovator deciding whether to make an investment, rather than ex post, once the fate of the investment is known. Otherwise, incentives will be misaligned. Relatedly, as Tom Cotter has aptly explained, Patent Assertion Entities (PAEs) are a “type of intermediary or broker, providing a service and the spread between the price they buy and the price they sell is their compensation for that service. Plus, there's always some risk they won't get anything.” Id. (p.46 n.277).
Second, regarding the second issue surrounding Dr. Cox's opinion, I'd note that whether courts should more readily consider settlements as comparables, and how if at all courts should adjust comparables given that the hypothetical negotiation assumes that the parties bargained assuming both infringement and validity, are among the issues discussed in Jonathan Masur's very interesting paper The Use and Misuse of Patent Licenses, 110 Nw. U. L. Rev. 115 (2015). (His more recent paper with Erik Hovenkamp, noted here, argues that, given all of the potential problems with comparables, courts should stop using them altogether.)
Third, go back and look at that range in valuation between the two experts: $123 million versus (at most) $1.5 million. Without casting aspersions at either expert's opinion, it's hard not to conclude that something has gone terribly wrong in the law of patent damages when two competent experts can reach such widely different opinions, differing by a factor of more than eighty. As I note in my Patent Damages Heuristics paper, as the range of possible outcomes increases, so does the risk of disadvantaging the more risk-averse party and of rendering settlement marginally more difficult. "The range of possible outcomes may be a very pressing problem in contemporary patent damages law. See, e.g., John C. Jarosz & Michael J. Chapman, The Hypothetical Negotiation and Reasonable Royalty Damages, 16 Stan. Tech. L. Rev. 769, 809 (2013) (stating that “for opinions issued since 1978 in which a suggested royalty rate was reported for both the patent holder and the infringer, the range has been as high as three hundred to one,” and that “[i]n many cases, the difference has been more than twenty to one. And the range has not declined over time.”) (p.10 & n.35). I'm not sure what the solution to this problem is. Perhaps predictability and simplicity should be more important considerations that they currently are, though the question then arises whether efforts to improve predictability and reduce costs would risk more inaccuracy.
Fourth, here's a link to the "Penn paper" the court refers to, and here is the abstract to that paper (I haven't read the paper yet myself):
Prior work suggests that more valuable patents are cited more and this view has become standard in the empirical innovation literature. Using an NPE-derived dataset with patent-specific revenues we find that the relationship of citations to value in fact forms an inverted-U, with fewer citations at the high end of value than in the middle. Since the value of patents is concentrated in those at the high end, this is a challenge to both the empirical literature and the intuition behind it. We attempt to explain this relationship with a simple model of innovation, allowing for both productive and strategic patents. We find evidence of greater use of strategic patents where it would be most expected: among corporations, in fields of rapid development, in more recent patents and where divisional and continuation applications are employed. These findings have important implications for our basic understanding of growth, innovation, and intellectual property policy.
Thanks to Nadia Soboleva of NERA (the firm with which Dr. Cox is affiliated) for bringing this very interesting opinion to my attention.