Introduction
“I have not failed. I’ve just found ten thousand ways that won’t work.” — Thomas Edison
Failed ideas and inventions, also known as “negative information,” can be economically valuable because anyone starting with knowledge of the dead ends can avoid spending time and resources on attempts that are ultimately going to fail. Clearly, negative information is valuable, but how does an economist value negative information? In this article, we explore techniques to overcome the challenges of determining the value of negative information, and show how negative information can inform allegations of trade secret theft in IP disputes.
Edison’s quote implies that failed ideas and inventions are valuable because they embody information on what does not work when creating something new. Unlike successful ideas and inventions, however, negative information is not priced by the market and is unlikely to directly generate any income. And yet a need to value negative information can arise in trade-secret-related disputes. In fact, the Uniform Trade Secrets Act from 1979 explicitly recognizes negative information as “information that has commercial value from a negative viewpoint, for example the results of lengthy and expensive research which proves that a certain process will not work could be of great value to a competitor.”[1]
Waymo’s February 23, 2017, lawsuit[2] against Uber for alleged theft of its trade secrets illustrates both the existence of and the need for valuing negative information. According to Waymo, Uber tried and failed to develop its own technology for a self-driving car, so it acquired Otto, a company started by a former Waymo employee who allegedly misappropriated Waymo’s technology. Waymo alleged that its employee’s theft of its intellectual property, including information about “dead-end designs,” allowed Uber (through Otto) to save a substantial amount of development time and cost. Waymo also claimed that its “extensive experience with ‘dead-end’ designs” and research findings that were unsuited for the market continued to be critical for its ongoing development[3] and likely underpinned Otto’s $680 million valuation at acquisition despite having been on the market for only six months. In this example, negative information consists of Waymo’s dead-end designs and know-how of what does not work when developing technology for a self-driving car.
Cases similar to Waymo are common in the technology space, where cutting-edge research can often be unfruitful or unmarketable and yet informative for product development.[4] In addition to ventures that are successful but only after many failed attempts (as in the Waymo example), negative information is often present in ventures that are ultimately deemed unsuccessful. Negative information can also be present in the valuation of ventures with significant research investments but prior to the development of any income-generating intellectual property.
The rationale for why negative information can be valuable is fairly intuitive; however, quantifying the value of negative information in general and particularly in the context of a trade secret dispute can be less straightforward. In the absence of comparables in the market and any attributable income, the only viable method for valuing negative information is a “cost to replicate” analysis.
A “cost to replicate” analysis attempts to answer a key “but-for” question: absent the alleged infringement of negative information, how much would it have cost the defendants to replicate the at-issue trade secret on their own? Importantly, the value of negative information is not how much it cost the plaintiffs in “creating” the negative information—a relatively simple accounting exercise—but what it would have cost the defendants to replicate it—an unknown counterfactual.
We begin in the next section with a brief note on recent trends in trade secret cases that underscore the importance of appropriately measuring the value of negative information.
Recent Trends in Trade Secret Cases in the US
Trade secret disputes in general—and particularly those in the technology space—have become more frequent due to a confluence of economic, legal, and regulatory factors. First, as the US workforce becomes increasingly more mobile—as has been the trend since the Great Recession[5]—trade secret disputes are expected to increase. This is particularly true in the technology sector, where increasing demand and compensation have resulted in the largest churn in employees relative to all other job sectors.[6] Negative information is at issue when an employee moves to a competitor and uses their knowledge of mistakes from their prior work to avoid making the same mistakes at their new place of employment. For example, in Novell, Inc. v. Timpanogos Research Group, Inc.,[7] following a settlement between the parties, the defendant recognized that he “mistakenly believed that, because [he and others] had developed certain technology while employed by Novell, [they] could take elements of that technology with [them] when [they] left.”[8]
Second, recent court decisions invalidated the patentability of certain types of subject matter (including patents related to software), thereby increasing reliance on trade secrets as an alternative for protecting intellectual property.[9] All else being equal, a general increase in trade secrets implies a concurrent increase in trade secret disputes, including ones involving negative information.
Third, the Defend Trade Secrets Act of 2016 (DTSA) enabled business to sue and seek remedies for theft of trade secrets in either state or federal courts, thereby further increasing the likelihood of trade secret cases.[10] And finally, trial outcomes for trade secret cases appear to be heavily skewed in favor of plaintiffs—almost 70% of the time in favor of plaintiffs—thus incentivizing plaintiffs to continue bringing their cases to court.[11]
Typical Approaches to IP Valuation
As trade secrets disputes increase in volume and prominence, analysts need guidance for appropriately valuing the negative information components of these disputes. Valuation methods for intangible assets like intellectual property include income approaches, market approaches, or cost approaches.[12] In the context of valuing negative information, the income and market approaches are typically inapplicable.
Income Approaches. Income approaches value intellectual property based on its potential to generate revenue or decrease future costs. For example, intellectual property that adds value to a product and increases its selling price can be valued by a discounted cash flow analysis of incremental profits. Negative information, however, does not manifest as any value-adding feature in a product or service and therefore lacks any attributable income. Market approaches value intellectual property based on arm’s-length transactions. For example, the price for licensing a patent can be informed by the licensor’s past or existing licenses for the same patent. Trade secrets are valuable in part because they are held in secret, and so it is rare for trade secrets in general to be licensed at arm’s length—and anyway it is difficult to imagine a traded market could exist for negative information.
Cost Approaches. Cost approaches are the valuation methods that can be applied to valuing negative information. A cost approach values intellectual property based on development costs. For negative information, cost approaches consider the costs avoided from not having to have invested in dead-end designs or failed procedures. For example, the costs incurred to develop intellectual property can inform on the cost savings proffered to a potential user of such intellectual property. Cost approaches don’t require consideration of future profits resulting from the use of intellectual property, which is useful because negative information is generally not associated with generating profit streams.
Academic studies of cost approaches distinguish between creation costs and re-creation costs.[13] The former refers to the original development costs of intellectual property, while the latter are what it would cost to develop intellectual property at a later point in time, which may be different for multiple reasons, for instance changes in labor or materials prices, or advances in public knowledge.[14]
A cost to replicate analysis aims to estimate re-creation costs, because it considers the cost the defendant would have incurred to replicate the intellectual property had they developed it themselves. In other words, the defendant saved itself development costs by misappropriating intellectual property. This can be considered unjust enrichment, and it can be calculated. In the Waymo-Uber matter, Waymo’s expert witness opined on Uber’s savings on development time by analyzing “the time periods that relate to the value of each of the trade secrets” and “the work that was foregone by virtue of… acquisition of the trade secrets.”[15]
Cost to Replicate Analyses Can Be Complicated by Negative Information
Negative information can be valued with a cost to replicate analysis; however, there are complications unique to negative information. Academic studies have noted “the difficulty in proving that the nonuse of a mistake or dead-end experiment damaged the plaintiff or unjustly enriched the defendant.”[16] When assessing economic harm, economic experts attempt to model the “but-for” world absent the alleged infringement. But it may be unreasonable to assume that defendants would have necessarily made the same mistakes or followed the same dead-ends as the plaintiffs did—in fact, defendants might have committed more errors, or made fewer mistakes. To illustrate using Waymo, let’s assume that Waymo undertook ten failed attempts prior to the successful development of its self-driving technology. Further, let’s assume that Uber acquired knowledge of these ten failed attempts, and with this negative information in hand, proceeded directly to developing its own successful technology for a self-driving car.
In this scenario, the value of Waymo’s negative information depends on how many of the ten failed attempts Uber would have undertaken in the but-for world. In part, that depends on the degree of interconnectedness of the failed attempts. Some efforts may necessitate sequential and iterative attempts, each new attempt learning from the prior ones; for other efforts the order may be random, making it imprudent to assume that the defendant in the but-for world would have undertaken the same number of attempts and in the same order as the plaintiff. So, while we can estimate the cost to replicate each of the attempts, we also have to have a choice framework to determine which of these failed attempts defendants would have pursued in the but-for world.
As a starting point for such a choice framework, we can examine plaintiffs’ failed attempts to identify which are relevant negative information. Waymo may have undertaken many failed experiments, not all of them necessarily linked to the development of its self-driving technology. Therefore, even before determining the potential mistakes the defendants would have likely made, experts must assess which failed experiments were undertaken in pursuit of the relevant goal (i.e., the product being replicated by the defendant).
Among plaintiffs’ relevant failed experiments, the damages expert must assess which ones would have been undertaken by the defendants in the but-for world, and which ones would have been undertaken in parallel or in sequence. In the Waymo case, Waymo’s expert analyzed the specific time periods for each trade secret before opining that some of the trade secrets, but-for the alleged misappropriation, “might have been done in a series,” while other trade secrets “probably would have gone in parallel.”[17] If it appears that the plaintiff chose from several approaches at random until arriving at last at the successful one, it may be appropriate to assume that the defendant would have followed a different order of approaches, and so would not have made all the same mistakes before finally finding the successful approach. Damages experts should account for the defendants’ existing knowledge—for example, a more advanced incumbent competitor, already equipped with relevant knowledge, may perform fewer failed experiments than a new entrant in the market.
In deriving the choice framework, the expert should also be informed by the age of the negative information. Mistakes made in research many years ago may be less relevant to current technology, and therefore less likely to be repeated. Academic studies support such consideration of obsolescence when valuing intellectual property.[18]
Similarly, changes in public knowledge will affect whether it is reasonable to assume a failed approach would have been tried in the but-for world. Since public knowledge by definition cannot be a trade secret, if any dead ends become publicly known before the alleged misappropriation by a defendant, then they should not be included in the analysis. This is especially relevant when plaintiffs release products into the market that make public certain results of their research. An example of this is when a company releases a product which incorporates a specific method, then the public may conclude that alternative methods are less desirable, making negative information about those alternative methods less valuable.
Determining which of the defendant’s projects were undertaken in pursuit of the relevant intellectual property requires understanding the defendant’s intent at the time of development. To this end, case-specific information in the form of documents produced in the usual course of business and testimony from those involved in the projects can be instrumental. Such information can guide the expert when modeling her choice framework. For instance, internal documents and testimony could indicate that defendants considered and were most likely to pursue a particular subset of the plaintiff’s full list of failed experiments. In such situations, the expert’s cost to replicate analysis would be deterministic and include the cost for the specific subset of experiments. Alternatively, when the testimony and documents do not identify a clear subset, the expert can resort to a probabilistic model. For example, if the defendants considered two experiments but did not indicate which one they would undertake, the expert could assign a 50 percent probability to both and calculate an expected value for her cost to replicate analysis.
Concluding Remarks
A well-supported valuation of negative information cannot be approached as a one-size-fits-all solution. Factors discussed in this article—including the number of potential failed ventures, the ordering of these ventures, and age of the technology in question—can vary from one case to another, which in turn impacts the expert’s modeling choices and final valuation. Nonetheless, identifying the relevant set of factors for consideration and appropriate means to account for them provides a useful framework for conducting the necessary individualized valuation of negative information.
Ideas and inventions that fail have economic value. We may associate market success with value, but the existence of negative information raises the possibility that some failed ideas and inventions can be valued in the right context. When Sir Isaac Newton said “if I have seen further, it is by standing on the shoulders of giants,” he was likely referring to those responsible for giant insightful failures as much as successful ones.
The opinions expressed are those of the authors and do not necessarily reflect the views of the firm or its clients. This article is for general information purposes and is not intended to be and should not be taken as legal advice.
Unif. Trade Secrets Act § 1 cmt., 14 U.L.A. 637–38 (1985). ↑
Waymo LLC v. Uber Techs. Inc. No. C 17-00939 WHA (N.D. Cal. Jan. 29, 2018). ↑
Waymo LLC v. Uber Techs., Inc., Complaint, at 10.
Waymo also created a vast amount of confidential and proprietary intellectual property via its exploration of design concepts that ultimately proved too complex or too expensive for the mass market; Waymo’s extensive experience with “dead-end” designs continues to inform the ongoing development of Waymo’s LiDAR systems today. The details actually used in Waymo’s LiDAR designs as well as the lessons learned from Waymo’s years of research and development constitute trade secrets that are highly valuable to Waymo and would be highly valuable to any competitor in the autonomous vehicle space. ↑
See, e.g., Fitbit v. Jawbone (three separate lawsuits: D. Del. 15-cv-775 and D. Del. 15-cv-0990, and N.D. Cal 15-cv-4073); Genentech, Inc. v. JHL Biotech, Inc., No. C 18-06582 WHA (N.D. Cal. Mar. 1, 2019). ↑
Neil Eisgruber, “Trends in Trade Secret Litigation Report 2020,” Stout, Apr. 2, 2020, at 21. ↑
Paul Petrone, “See The Industries With the Highest Turnover (And Why It’s So High),” LinkedIn Learning Blog, Mar. 19, 2018. ↑
46 U.S.P.Q.2d 1197, 1217 (Utah Dist. Ct. 1998). ↑
“Novell settles suit against Timpanogos,” Deseret News, Aug. 29, 1998. ↑
Eisgruber, supra note 6, at 16–18. ↑
Id. at 21. ↑
Id. at 17–18. ↑
Robert F. Reilly & Robert P. Schweihs, Valuing Intangible Assets 96 (1998). ↑
Id. at 97–98. ↑
Id. at 97–98. ↑
Deposition of Lambertus Hesselink (September 26, 2017) at 264:11–265:11, Waymo LLC v. Uber Techs. Inc., No. C 17-00939 WHA (N.D. Cal. Jan. 29, 2018) (“Hesselink Deposition”). ↑
Charles Tait Graves, The Law of Negative Knowledge: A Critique, 15 Tex. Intell. Prop. L.J. 387, 412. ↑
Hesselink Deposition, supra note 16, at 264:11–265:11. ↑
Reilly & Schweihs, supra note 13, at 99–100. ↑