ORDER GRANTING IN PART AND DENYING IN PART APPLE'S MOTION TO EXCLUDE

Re: Dkt. Nos. 319, 355-14

VINCE CHHABRIA, District Judge.

Apple's motion is granted with respect to the survey data for the `260 patent and all damages estimates dependent on that data. Given the limited explanation Unwired has offered for the assumptions underpinning Dr. Allenby's "equilibrium price premium" model, that data must also be excluded based on the current record. If Unwired believes it can properly explain how the EPP model fits the facts of this case, the Court will reconsider its conclusion after a formal Daubert hearing. Apple's motion is otherwise denied.

I. Dr. Allenby's report

Dr. Allenby's survey for the `260 patent measured the value attributed to manual or automatic "content sharing," which Dr. Allenby described as a "[m]ethod of transferring songs, movies, and applications across devices . . . owned by the user." Dkt. 318-6 at 26.1 This is not consistent with any user-facing benefits conferred by the claims asserted on the `260 patent. Rather, it appears that Dr. Allenby's questions for the `260 patent reflect the claim construction Unwired initially proposed for "provisioning," which would have made the term essentially synonymous with "providing." Dkt. No. 269 at 4; see also 357-4 at 10. Unwired's proposed claim construction was rejected in favor of the narrower "enabling or modifying communications capabilities." Dkt. No. 269 at 7. In light of that construction, "provisioning" features on a user's device bears no reasonable connection to sharing content across devices. Unwired suggests that Dr. Allenby's survey was intended to measure the value attributed to automatically provisioning all of one's devices after manually provisioning a single device. But this would still make the survey inaccurate, as the benefit of the invention was tied to remote provisioning, not automatic provisioning. See id. at 3-4. And this would still leave the survey misleadingly ill-defined, as songs and movies have no effect on communications capabilities and are therefore not "provisioning" content. The combined result is a survey question — and survey responses — targeted at an invention other than the one at issue in this litigation. This is a problem of admissibility rather than weight. See, e.g., Fractus, S.A. v. Samsung, No. 6:09-CV-203-LED-JDL, 2011 WL 7563820, at *1 (E.D. Tex. Apr. 29, 2011). The `260 patent survey results must therefore be excluded, along with all damages estimates hinging on those survey results. Unwired has requested an opportunity to correct its questions and redo its conjoint survey. But where the initial effort misses the mark so badly, it would be inappropriate to incentivize overreaching by allowing a second attempt. Cf. Network Prot. Scis., LLC v. Fortinet, Inc., No. C 12-01106 WHA, 2013 WL 5402089, at *8 (N.D. Cal. Sept. 26, 2013).

The survey language on the `446 and `092 patents is a closer question. In both cases Dr. Allenby described the consumer-facing benefit of the invention in terms that clearly exceeded its incremental improvement over the prior art. The `446 patent doesn't represent the invention of voice transcription. Dkt. No. 318-6 at 27. The `092 patent doesn't represent the invention of automatic location detection. Id. at 26. The patents represent modest improvements in these areas, based on the kinds of minor technical modifications most ordinary consumers likely wouldn't appreciate. But at this stage, the Court's concern is with limiting evidence that is irrelevant or unfairly prejudicial, that lacks an adequate technical or scientific ground for Daubert purposes, and that jurors may find confusing, misleading, or unhelpful. Although these survey questions perhaps skirt the line, the fundamental problem — their high level of abstraction — doesn't render the survey results entirely unhelpful in assigning value to the improvements in the surveyed areas, particularly as Unwired may argue that its improvements over the prior art were instrumental in making previously existing functionalities effective on the accused devices. See Sentius Int'l, LLC v. Microsoft Corp., No. 5:13-CV-00825-PSG, 2015 WL 331939, at *3 (N.D. Cal. Jan. 23, 2015). Any flaws in the survey can be adequately drawn out at cross and mitigated with proper instructions. Nor is this case analogous to VirnetX, where the damages expert "relied on the entire market value of Apple's products" and "did not even attempt to subtract any . . . unpatented features from the [royalty] base" if those features weren't separately sold. VirnetX, Inc. v. Cisco Sys., Inc., 767 F.3d 1308, 1328 (Fed. Cir. 2014). The damages estimate in that case failed to comply with "settled principles of apportionment" because it adopted the entire market value rule without demonstrating that "the patented features [drove] the demand for [the] entire multi-component product." See LaserDynamics, Inc. v. Quanta Computer, Inc., 694 F.3d 51, 67 (Fed. Cir. 2012). A survey that approximates damages for a patented feature at a high level of abstraction is committing "error" of a different order, and doesn't give rise to a per se admissibility problem. See VirnetX, 767 F.3d at 1330 ("[W]e have long acknowledged that any reasonable royalty analysis necessarily involves an element of approximation and uncertainty." (citations and internal quotation marks omitted)). Apple's motion is therefore denied with respect to the survey data for the `446 and `092 patents.

Apple overstates its argument for excluding Dr. Allenby's EPP model. The EPP isn't entirely Dr. Allenby's "own creation," but an adaptation of existing methods. See Dkt. No. 353-4 at 7; Greg M. Allenby et al., Valuation of Patented Product Features, 57 J.L. & Econ. 629, 632-33, 655-56 (2014). To the extent Apple is challenging the validity of the assumptions behind game theory itself, the Court suspects those challenges could be overcome, as game-theoretic models have longstanding and widespread acceptance in economics and other disciplines. But it is the proponent of the expert testimony who bears the burden of demonstrating admissibility by a preponderance of the evidence. Daubert v. Merrell Dow Pharm., Inc., 509 U.S. 579, 593 n.10 (1993). At this stage, Unwired has not adequately explained — and, in large part, has not even attempted to explain — the basis for and importance of the assumptions supporting Dr. Allenby's approach. See, e.g., Dkt. No. 357-4 at 12 n.7; Dkt. No. 323-12 at 49-52. Without this detail, the Court has no way of determining how and to what extent Dr. Allenby's assumptions affect the model's results. This makes it impossible to determine the model's appropriateness for the facts of this case and leaves the Court no reason to believe jurors will be capable of meaningfully weighing the data they're confronted with. Cf. Uniloc USA, Inc. v. Microsoft Corp., 632 F.3d 1292, 1313 (Fed. Cir. 2011). That some of the assumptions underpinning the EPP model were published in a peer-reviewed paper does not itself justify their application here. See Dkt. No. 357-4 at at 13. Nor does the fact that the EPP model yields relatively conservative results offer assurance that those figures are useful or reliable, even as a basis for comparison. See id. at 12. Instead, by shrouding the EPP results in an air of legitimacy, these factors raise the risk of juror confusion, making it all the more essential that the Court exercise its gatekeeping function until admission is adequately supported. See Kumho Tire Co. v. Carmichael, 526 U.S. 137, 154 (1999). The Court will reconsider its conclusion after a Daubert hearing with expert testimony if Unwired wishes to pursue that option.

II. Mills's damages report

Apple objects to Mills's reliance on the Lenovo license in calculating a "willingness to accept." Apple suggests the Microsoft license as a more comparable alternative. But for Daubert purposes, Unwired is not required to prove that all alternatives are demonstrably worse; it only has to prove that the license in question offers a reasonable enough basis for comparison that a jury can then be expected to weigh the damages evidence it's presented with. Lucent Techs., Inc. v. Gateway, Inc., 580 F.3d 1301, 1326 (Fed. Cir. 2009). Here, the Lenovo license doesn't suffer from significantly different or more numerous flaws than the Microsoft license Apple proposes. To the extent there are concerns about the degree of comparability, these are factual rather than methodological concerns — issues well within the province of the jury. ActiveVideo Networks, Inc. v. Verizon Commc'ns, Inc., 694 F.3d 1312, 1333 (Fed. Cir. 2012). And in any event, the adjustments necessary to correct for differences in the Lenovo license are largely the same as those necessary to correct for differences in the Apple license. Apple suggests that the back-of-the-envelope calculations borrowed from Robbins amount to an impermissible "rule of thumb." But it isn't imprecision that makes for a rule of thumb; it is arbitrariness, or lack of grounding in the facts of the case. VirnetX, 767 F.3d at 1331. Robbins provides sufficient case-specific justification for the license adjustments to make the figures admissible. What weight to assign to the figures in light of that justification is a question for the jury.

Apple's request to exclude Robbins's testimony is denied. See Dkt. No. 357-4 at 19.

IT IS SO ORDERED.