6

MISCONCEPTION 5: Open Innovation Turbocharges R&D

One of the biggest innovation trends over the past few decades has been “open innovation.” The concept was first introduced in 1983 and later popularized by Henry Chesbrough in Open Innovation: The New Imperative for Creating and Profiting from Technology,¹ as well as related articles in the business press. The basic idea behind open innovation is that the amount of R&D done outside a given company will always dwarf what the company itself can do internally. Intel, for example, the largest U.S. spender, with R&D investment in 2013 of $10.6 billion, only represents 2.3 percent of the $456 billion total U.S. R&D. Accordingly, it is likely that research being done elsewhere is relevant to Intel. Given that, Intel may be able to extend the value of its own R&D by exploiting external research.

In order to define open innovation, it’s useful to understand its antithesis, closed innovation. Chesbrough defines closed innovation as the traditional system in which companies generate, develop, and commercialize their own ideas. To accomplish this, they hire the best and the brightest, and as a result can generate a substantial number of high-quality innovations. Profits from the new innovations are reinvested in R&D, thereby creating a “virtuous cycle of innovation.”

In contrast, Chesbrough defines open innovation as a system in which companies combine internal and external means for generating, developing, and exploiting innovations. This allows companies to use the best path (internal or external) for development and commercialization of their own ideas, as well as use the best ideas (internal or external) to fill their development and commercialization pipelines.

Chesbrough argues that the closed innovation model began to erode toward the end of the twentieth century, in part because increased labor mobility made it more difficult for companies to control their intellectual property. The other factor contributing to erosion of closed innovation was dramatic growth in venture capital (VC). Venture capital facilitated the creation of startup companies as an additional avenue for these mobile employees to exploit the innovations they helped create within the large companies.

In Chesbrough’s view this erosion meant internal R&D was no longer a source of competitive advantage. As evidence of this, Chesbrough points to the fact that startup companies can’t afford to do basic research, yet despite that, they are able to unseat large established companies. As a particular example, he points to the fact that Lucent and Cisco competed directly in the same industry, yet had vastly different innovation strategies.

While it is true that the two companies ultimately competed in the same industry, this occurred only after the convergence of voice and data. Lucent came from the voice side. It was the equipment manufacturing business spun out of AT&T Corporation in 1996, with a goal of achieving scale economies by selling equipment to AT&T’s rivals. As a means to attract former rivals to purchase from Lucent, AT&T included Bell Labs and all its patents as part of the spinout. Not surprisingly, given its legacy, Lucent conformed to the closed innovation model. It sought fundamental discoveries to “fuel future generations of products and services.”

Cisco came from the data side. Cisco was founded in 1984 by two former IT managers at Stanford to commercialize a router and companion software developed at the university. Thus Cisco is a great example of the phenomenon from the last chapter explaining why small companies appear to generate more radical innovation than large companies—the founders took the core invention from their prior employer, Stanford. In fact, they took it from their current employer—the founders continued to work at Stanford for 18 months after startup.

Because Cisco began with technology developed elsewhere rather than from internal R&D, it inherently followed an open innovation strategy. Thus its innovation strategy at least initially was to partner with or invest in promising startups as a way to gain new technology. Ultimately, however, Cisco spent as much on R&D as it did on A&D (acquisition and development). Thus it had moved from exclusively open innovation (using Stanford’s innovation) to the more hybrid form envisioned by Chesbrough.

While the argument that the two strategies (open innovation and closed innovation) competed directly is valid, the argument that open innovation is the means for startups to unseat large established companies is debatable. Cisco was a monopolist in the router market for many years, and thus had no large companies to unseat. Cisco only had to compete with Lucent when voice and data began to merge in the late 1990s. At that point, the two firms had comparable size. Cisco’s market capitalization of roughly $100 billion approached that of Lucent (roughly $150 billion).

The response to Chesbrough’s prescription for open innovation has been so widespread that a recent University of California, Berkeley survey of executives from the largest companies in the United States and Europe found that 78 percent reported applying open innovation for many years. Moreover, they expected use of open innovation to grow even further.

The critical question, however, is how have companies who’ve adopted open innovation fared? What’s the impact of open innovation on companies’ RQ? Before answering that question, we first need to acknowledge that open innovation takes many forms: crowdsourcing ideas, funding research at universities and government labs, licensing technologies from those labs, fostering startup companies through corporate venture capital (CVC), joint ventures with rivals and suppliers, and outsourcing R&D. These are very diverse activities, so it’s unlikely they all have the same impact. Accordingly, we review the record on two broad classes of open innovation, idea sourcing and idea development, separately.

IDEA SOURCING

One of the most prevalent forms of open innovation is idea sourcing—developing and commercializing an idea that originated outside the company. It is so prevalent that a recent innovation survey conducted by Ashish Arora, Wes Cohen, and John Walsh² found that 49 percent of the most important product innovations from manufacturing companies originated from an outside source. These sources included suppliers, customers, rivals, commercial consultants and labs, independent inventors, and university and government labs. Thus, there is no question that idea sourcing is widespread. The important question is one of impact—which sources yield what results?

The most common source of external ideas is customers—accounting for 27 percent of the ideas behind companies’ most important innovations. This finding reinforces an important stream of research on user innovation pioneered by Eric von Hippel at MIT.³ The principle behind user innovation is that users have acutely felt needs often not met by existing products. To satisfy these needs they could generate their own innovations. However, because they typically lack the resources to produce those innovations, the more expedient solution is to convince existing suppliers to develop them. If the user comprises a large share of the manufacturer’s sales, or is representative of a class of users that comprise a large share, then it often makes sense for the manufacturer to develop that idea. A classic context for user innovation is medical devices. Companies such as Johnson & Johnson actively encourage prominent surgeons to invent devices. Not only will the inventing surgeon use the device, but knowledge that a renowned surgeon invented the device (through publications, conferences, and the manufacturer’s own marketing) fuels diffusion of the device to other surgeons.

The next most common source of external ideas is suppliers, comprising 14 percent of companies’ most important new products. A classic example of companies that rely on supplier innovation is Toyota. Suppliers have three advantages as a source of ideas. First, they have strong incentives to generate these ideas as a way to strengthen the buyer’s dependence upon them. Second, they typically have in-depth knowledge of the customer’s needs and capabilities, so can develop ideas that best match them. Finally, they have well-developed relationships with the customer that facilitate the transfer of knowledge.

One concern suppliers should have when they innovate for their buyers, however, is that the buyers provide the innovation to other suppliers. If that occurs, the innovating supplier may fail to profit from its innovation. This happened to a friend of mine, Dan Pulos, who held multiple patents for products he produced for a Fortune 50 company. That company contracted with another supplier to produce some of Dan’s inventions—a clear patent infringement. The problem pursuing that claim, however, is that the Fortune 50 company accounted for approximately 50 percent of Dan’s revenues. Dan couldn’t afford to lose the customer, so he had to tolerate the infringement.

There do, however, appear to be strategies suppliers pursue when faced with similar threats. Jenny Kuan, Dan Snow, and Susan Helper present a rich study characterizing these for the 70 percent of automotive suppliers who contribute design work.⁴ In survey responses from 1,400 suppliers, the authors found that companies employ three different strategies in dealing with these contracting hazards. The most interesting of these is a “high R&D” strategy in which the suppliers offer technologically advanced products to multiple firms. This gives them “seller power,” both because they don’t rely too heavily on any single buyer and because they have cutting-edge technology not available from other suppliers.

Other external idea sources are less prevalent because they lack the strong communications channels and/or the aligned incentives characterizing the customer and supplier relationships. These other sources include rivals, consultants/service providers, and independent inventors. Each comprises about 8 percent of ideas for companies’ most important new products.

The least common idea source (5 percent) is universities. This is somewhat surprising and also unfortunate because university technology transfer offices (TTOs) all maintain markets for university inventions as a way to generate licensing revenue from faculty research. However, as we saw in the last chapter, the record on companies utilizing university technology is weak (university licensing revenue is only 1 percent of grant revenue). The likely reason for this is that the ideas produced by universities (other than those produced under contracts with companies) tend to be earlier stage, and therefore more remote from commercial applications.

So we have a sense of how prevalent various external idea sources are. Accordingly, we also know their relative rank. What’s interesting about the relative rank is that it should reflect the relative returns. In other words, customer ideas are likely most prevalent because they generate the highest returns. This is an important insight. An interesting follow-on question, however, is whether the returns to customer ideas are high because the cost to acquire and/or commercialize them is low, or because customer ideas are better and therefore generate higher revenues.

Ashish, Wes, and John examined this as well by looking at sales revenue from new products emanating from the external idea sources. They found that a new product from customer ideas had very little impact on revenues. On average, it comprised only 17 percent of company sales. In contrast a new product from a specialist source (consultants, independent inventors, or universities) comprised 26 percent of company sales. Thus sales from specialist sources are roughly 60 percent higher than those from customer ideas even though use of customer ideas is more prevalent (27 percent versus 21 percent). This suggests the prevalence of new products based on customer ideas stems from low cost to acquire and commercialize customer ideas rather than because their quality is higher.

A newer source of external ideas that is generating considerable attention is crowdsourcing. Crowdsourcing is a means to generate ideas or services by soliciting them from a large group, typically via an online platform like Kaggle (a predictive analytics crowdsourcing site) or Topcoder (a computer programming crowdsourcing site). To create a crowdsourcing competition, the sponsoring organization characterizes the real-world problem it is trying to solve, offers a cash prize, and broadcasts an invitation to submit solutions. One such example from Karim Lakhani and Kevin Boudreau, two of the leading researchers on crowdsourced innovation, is Merck’s competition on Kaggle to generate ideas for streamlining its drug discovery process. Merck hosted an eight-week contest with a $40,000 prize that generated 2,500 proposals from 238 teams. The winning solution utilized a machine-learning approach and relied upon expertise (computer science) not housed within Merck (at least at the time).⁵

This illustrates one of the advantages of crowdsourcing—it draws on a larger and more diverse set of skills, experience, and perspectives than those available within any given company. In addition, the sponsor obtains free labor. In the case of Merck, this was a sizable amount of free labor (the 237 teams and 2,499 proposals that didn’t win the $40,000 prize). If we assume that each team consists of three people working one-fourth time (10 hours per week) for the eight-week period, and we further assume the national average salary of roughly $100,000 (before benefits) per developer, Merck obtained $28.6 million of labor for $40,000!

Finally, the sponsor obtains free experimental results from the teams and proposals that didn’t win. These nonwinning proposals provided Merck a much deeper sense of what the solution space looked like—including what solutions didn’t work and why. Moreover, while the knowledge is crowdsourced, the sponsor maintains all the intellectual property for both the successful and unsuccessful solutions.

One question that remains underexamined with respect to crowdsourcing ideas is the extent to which they are implemented by the sponsoring organization. While the crowdsourced solution to the screening problem was ultimately implemented at Merck, there is evidence suggesting this is the exception rather than the norm. Implementing such solutions seems to fall victim to classic knowledge transfer problems.

Successful transfer of external knowledge requires a number of stars to align. In particular, the external source and internal receiver both have to be motivated to make the exchange, the knowledge has to be inherently transferable, and the recipient has to have sufficient existing knowledge to absorb the new knowledge. Finally, the receiver needs to feel the knowledge is valid. Often, however, external knowledge is automatically dismissed as being inferior—the classic Not Invented Here (NIH) syndrome. This may be particularly acute in the case of crowdsourcing since the source is anonymous. Indeed, a field study of a startup company⁶ (with 50 scientists) ran four crowd contests over a three- to six-month period. The study found that while some prizes were awarded, none of the ideas were used in subsequent R&D efforts. Interviews with the companies’ employees indicated that the failure to implement the solutions stemmed from perceived quality. The company’s scientists felt the solutions were not relevant (did not meet all the contest criteria), and for those that were relevant, they felt the solutions tended to resemble ones the company had already considered. Note of course this could be fact, or could merely reflect NIH syndrome. Regardless, it suggests crowdsourced ideas may pose implementation problems.

Thus the two big questions about the long-run viability of crowdsourcing are (1) whether highly skilled participants will continue to offer free labor, and (2) the extent to which the crowdsourced ideas are implemented by the sponsoring company.

IDEA DEVELOPMENT

While idea sourcing considers open versus closed innovation at the front end of the R&D pipeline, idea development pertains to open versus closed innovation at the middle of the pipeline. Open innovation at the development stage can take many forms such as joint ventures, collaborations, and so on. However, we focus attention on the most extreme form of open innovation at this stage: outsourced R&D. This focus stems from the fact that the NSF has data on outsourced R&D over a longer period.

What is most striking about the NSF data is that it indicates outsourced R&D increased by a factor of 20.5 (2,050 percent) during a period in which R&D itself (measured as the number of scientists and engineers working within companies) increased by roughly one-eighth that amount (250 percent). Of course, the real question regarding outsourcing R&D is not the prevalence, but rather the effectiveness.

Not surprisingly companies believe their shift toward outsourced R&D has benefited them. In fact, a survey of CIOs and CEOs revealed that 70 percent of them believe outsourced innovation improved their financial performance.⁷ However, we know from Chapter 1 that companies have had no good way to gauge whether this is true. So to truly understand the effectiveness of outsourced R&D we need to examine its impact on RQ.

To do that, I separated companies’ R&D into two buckets—R&D conducted internally, and outsourced R&D—and treated them as separate inputs for producing company sales. Doing this allowed me to determine how productive each form of R&D was—how much a 10 percent increase in each of internal and outsourced R&D increased company revenues. When I looked at internal R&D, I found that on average, a 10 percent increase in R&D spending increased revenues 1.3 percent (while this sounds like the company is losing ground, revenues are typically 50 times larger than R&D expenditures).

However, when I looked next at outsourced R&D, I obtained the astounding result that outsourced R&D had an RQ of zero! This means a 10 percent increase in outsourced R&D yields no increase in company revenues. As a result, the company’s profits actually decrease by the amount of spending on outsourced R&D.

This result was almost implausible, so I did additional analyses to understand what might be going on. The first analysis I conducted was interviewing companies about when and why they outsource their R&D. These interviews indicated that companies outsource R&D for a number of reasons, and that those reasons vary with top-level R&D strategy rather than industry mandate (in less than 5 percent of industries is it the case that all companies outsource).

At one end of the outsourcing spectrum, companies outsource only to universities and government labs. They do this to gain access to basic research as well as to identify potential employees. In the middle of the spectrum, companies outsource under special circumstances. For example, they use outsourcing as a flexible substitute for internal hiring when future demand is uncertain; they outsource activities where they lack capability and don’t intend to develop it internally (because they would operate below efficient scale); or they outsource testing (particularly in the case of pharmaceutical trials by contract research organizations, CROs). At the extreme end, companies outsource all “noncore” R&D activities.

Unfortunately, the NSF Survey of Industrial Research and Development (SIRD) did not collect data on the destination for outsourced R&D; however, these data were collected in its successor, the Business R&D and Innovation Survey (BRDIS). A report of BRDIS data⁸ indicates 3.4 percent of outsourcing is to universities, 81.3 percent is to companies, and the remaining 15.2 percent is to government agencies and other organizations. Since the vast majority of outsourcing is to companies, outsourcing seems to conform to rationale in the middle and extreme ends of the spectrum.

The second analysis investigated how reliable the result of unproductive outsourced R&D actually was. To examine that, I looked at what was happening to the productivity of internal versus outsourced R&D over time. I found that there had been no change in the productivity of either internal R&D or outsourced R&D. This meant outsourced R&D had never been productive. This was a reassuring result, because it meant I didn’t have to worry about how outsourced R&D might have been changing.

The third analysis examined whether outsourcing was truly unproductive, or whether it just appears that way because poor quality companies are the ones that outsource their R&D. This turned out not to be the case. The internal RQ of outsourcing companies is the same as that for companies who don’t outsource at all.

The fourth analysis checked whether outsourced R&D just appears to be nonproductive because companies outsource their worst projects. This idea came from discussions with a colleague who works with Merck. He said that Merck outsources the projects it thinks it might want to kill because it is easier to kill outsourced projects than it is to kill internal projects. To kill an outsourced project, you merely fail to renew the follow-on contract. In contrast, killing an internal project requires fighting political battles that may alienate the companies’ researchers.

To check whether companies were outsourcing their worst projects, I looked at what happens when they first start outsourcing their R&D. If they truly outsource their worst projects, you would expect their internal RQ to go up at that point. This is because getting rid of the bad projects leaves the company with a stronger portfolio of projects. This stronger portfolio should generate higher returns (RQ). That’s not what happens. In fact, internal RQ is slightly higher before outsourcing, but by an amount so small it’s not worth considering. Thus there is no evidence project quality is driving the lower productivity of outsourcing.

In short, after checking which companies outsource, who they outsource to, and what they outsource, it appears outsourced R&D is inherently less productive than internal R&D. Why might that be? While the data don’t provide insight, one possible explanation is that R&D produces internal spillovers—knowledge that gets generated by one project but can be utilized in other current or subsequent projects as well. To understand the intuition for internal spillovers, remember that companies carry portfolios of projects. As we know from Chapter 2, the most likely outcome from these projects is termination (the 123 of 125 projects that are never launched commercially). This means the value of R&D projects likely lies outside project outcomes themselves—possibly in the ability to recycle knowledge gleaned from the failures into subsequent projects.

I provided one example of technology reuse in Chapter 2—the redeployment of ion beam technology at Hughes Aircraft Company from satellites to semiconductors. Had Hughes outsourced the ion beam R&D, the outsourced company would have derived the benefit from any other application (to the extent that they too were broad-based). While the Hughes case is anecdotal, there is some quantitative evidence that companies fail to capture spillovers when they outsource. A recent study by Carmen Weigelt of Internet banking adoption finds that banks that outsource their initial IT integration are less able to develop new applications and accordingly have lower revenues from their Internet operations.⁹

An alternative explanation for the lower productivity of outsourced R&D came from Nick Heinz, chief financial officer of Sears Home Services, who used to share my St. Louis to Los Angeles commute. On one of those flights, I told Nick about the finding that outsourced R&D was unproductive and said I was trying to understand why. He said, “Oh, I know why. It’s the ‘consultant effect.’” He explained that only 30 percent of consulting recommendations are adopted, because all the knowledge to implement them resides with the consulting company, since it did all the investigation. As a result, the funding company would have to replicate much of the work it had already paid for to have sufficient expertise to implement the consultant’s recommendation. This explanation is reminiscent of the problems discussed earlier regarding implementation of crowdsourced ideas.

Perhaps the simplest explanation of why outsourced R&D is unproductive is that the resources at the supplier company are inferior. This would explain both why companies outsource R&D and why outsourced R&D is less productive. Quite simply, for the supplier company to offer a contract that provides it a profit, while still being less expensive than conducting the project internally, the supplier has to have lower cost inputs. This point is made vividly in Pierre Azoulay’s study on pharmaceutical manufacturers outsourcing their clinical trials to clinical research organizations (CROs). One outsourcing manager in the study complained, “CROs keep giving us bad people to choose from, and there is nothing I can do about it,” and another complained that the CRO implied that an “A-Team” would be conducting the trial, but “rookies” had been substituted at the last minute.¹⁰

Beyond the questions of why outsourced R&D is unproductive, a final question I’m typically asked when presenting these results is why companies persist with R&D outsourcing, given its lack of productivity. Again the NSF data provide no insights. However, I believe it’s because companies don’t yet realize that outsourcing is unproductive. This would be an obvious extension of the problem introduced in Chapter 1 that companies lack good measures of R&D productivity.

Given this ambiguity about the productivity of R&D, companies are vulnerable to “information cascades.” An information cascade occurs when companies discount their own instincts and follow the behavior of other companies or recommendations of consultants who they believe have superior information.¹¹ Given the attention to open innovation, and the 20.5-fold increase in outsourced R&D, it’s likely companies believe outsourced R&D is productive, even though there is no evidence other than widespread adoption.

THE BIGGER PROBLEM

Unfortunately, the profit decrease from outsourced R&D is only the tip of the iceberg. Outsourced R&D is a slippery slope. Once a company begins to outsource R&D it previously conducted internally, it closes labs and reduces technical staff. As a result, the next time a company has a similar project that it considers conducting internally versus through outsourcing, the deck is stacked in favor of outsourcing. That’s because executing the project internally requires rebuilding the labs and hiring new technical staff, in addition to the marginal cost (labor, equipment, and materials) of the project itself. In contrast, outsourcing just requires the marginal cost of the project. Because each new outsource decision has this flavor, ultimately the company’s R&D capability unravels and its RQ decays.

This unraveling of R&D capability appears to have happened at a major division of a Fortune 50 company. Initially the division began outsourcing drafting to India as a way to satisfy its trading offset requirements. Offsets are a common condition of international trade contracts that requires the foreign exporter to purchase goods or services from the importing country. Drafting seemed to be an innocuous form of outsourcing because it doesn’t involve scientists or engineers. However, the next step in the unraveling was that the division outsourced design, rationalizing that the heart of the design was the standards (which it retained in-house). Ultimately, however, if you don’t do design, you lose the requisite knowledge to determine whether designs meet standards, so you can’t develop standards either. In the end, the division was forced to outsource standards writing as well. Thus outsourcing of R&D not only has zero RQ itself, it leads to further outsourcing, and ultimately to the inability to conduct R&D internally. This unraveling of R&D capability as a consequence of outsourcing R&D is related to the unraveling of innovative capability as a consequence of outsourcing manufacturing that Gary Pisano and Willy Shih document in Producing Prosperity.¹²

SUMMARY

There is a widely held belief that open innovation increases companies’ financial performance. Accordingly, open innovation has been adopted by the vast majority of companies engaged in R&D. More telling, there has been a 2,050 percent increase in the amount of outsourced R&D.

While there is some evidence that open innovation in the form of idea sourcing may improve companies’ financial performance, the record on idea development indicates that R&D outsourcing not only fails to improve financial performance, it actually degrades it! This occurs because outsourced R&D incurs R&D expenditures without increasing revenues. Thus it decreases profits. Worse, however, it appears that outsourcing R&D is a slippery slope wherein company innovative capability decays, so the company increasingly outsources, and capability decays even further.

Overall, these mechanics are so powerful that outsourced R&D accounts for much of the 65 percent decline in RQ that we saw in Chapter 1. The RQ for internal R&D has remained relatively constant, which is great news! It means if companies are willing to undertake the investments to recreate labs and rebuild their technical staffs, over time they should be able to restore RQ to prior levels.