6
A Problem in the Making: How Firms Formulate Sharable Problems for Open Innovation Contests
Martin W. Wallin, Georg von Krogh, and Jan Henrik Sieg
Abstract
Crowdsourcing in the form of innovation contests stimulates knowledge creation external to the firm by distributing technical, innovation-related problems to external solvers and by proposing a fixed monetary reward for solutions. While prior work demonstrates that innovation contests can generate solutions of value to the firm, little is known about how problems are formulated for such contests. We investigate problem formulation in a multiple exploratory case study of seven firms and inductively develop a theoretical framework that explains the mechanisms of formulating sharable problems for innovation contests. The chapter contributes to the literatures on crowdsourcing and open innovation by providing a rare account of the intra-organizational implications of engaging in innovation contests and by providing initial clues to problem formulation—a critical antecedent to firms’ ability to leverage external sources of innovation.
Introduction
During new product development, firms must solve numerous technical, innovation-related problems such as design flaws, glitches in pilot-production, and process or performance shortcomings in product upgrades. As a result, innovation has often been characterized as a problem-solving process (see e.g. Allen, 1966; Iansiti & Clark, 1994; Pisano, 1996; Thomke, 2003; Terwiesch, 2008; Foss et al., 2016). In the post World War II period, firms relied heavily on their own capabilities to solve innovation-related problems (Chandler, 1977; Langlois, 2003). Today, firms increasingly solve problems through engaging with knowledgeable outsiders, such as customers, suppliers, and technical experts. Indeed, the appeal of crowdsourcing very much rests on the idea of “tap[ping] the latent talent of the crowd” (Howe, 2006). However, in this chapter we will develop an argument that successful crowdsourcing is less about tapping latent talent and more about enticing outside talent to solve a particular problem. In particular, we will demonstrate that problem formulation may be a critical antecedent to successful crowdsourcing.
Crowdsourcing is a means of solving problems by outsourcing a task to a crowd through an open call (Howe, 2006, 2008; Jeppesen & Lakhani, 2010; Afuah & Tucci, 2012). This “open innovation” process (Chesbrough, 2003; Langlois, 2003; von Hippel, 2005; Laursen & Salter, 2006; Chesbrough et al., 2008; Mowery, 2009; Dahlander & Gann, 2010) entails “finding creative ways to exploit internal innovation, incorporating external innovation into internal development, and motivating outsiders to supply an ongoing stream of external innovations” (West & Gallagher, 2006: 319). “Open innovation contests” (hereafter “innovation contests”) is a form of crowdsourcing where firms stimulate problem solving by “broadcasting” (Lakhani et al., 2007; Jeppesen & Lakhani, 2010) a problem statement to a very large pool of self-selecting outside solvers and by proposing a monetary reward for solutions (Terwiesch & Xu, 2008; Jeppesen & Lakhani, 2010; Sieg et al., 2010; Boudreau et al., 2011; Afuah, 2018).
Innovation contests are currently making successful inroads into private as well as public sectors (Tapscott & Williams, 2006; National Research Council, 2007; McKinsey & Company, 2009; White House, 2010). For example, the X Prize Foundation offered $10 million for the first flight to an altitude of 100 kilometers, Netflix awarded $1 million to the team of researchers that managed to substantially improve the accuracy of the company’s movie recommendation algorithm, and the DARPA Grand Challenges encourage the development of autonomous robotic vehicles. Publicly known examples of leading firms using contests to address pressing issues are Eli Lilly, Dow AgroSciences, Audi, Shell, and Pepsi.1 The trend towards outside problem solving has also given rise to specialized intermediaries such as InnoCentive, NineSigma, and Idea Crossing, that provide platforms, support, and consultation for innovation contests.2
While prior research has contributed greatly to the understanding of design and effectiveness of innovation contests (e.g. Taylor, 1995; Che & Gale, 2003; Schottner, 2008; Terwiesch & Xu, 2008; Jeppesen & Lakhani, 2010; Boudreau et al., 2011), surprisingly little is known about how firms formulate the problems that set them off. This is an important gap considering that problem statements are “sine qua non” of innovation contests (Jeppesen & Lakhani, 2010: 1018). Similarly, received literature on how firms leverage external sources of innovation largely omits problem formulation and jumps to the question of how firms search, source, incentivize, and contract for innovation (see West & Bogers, 2014 for a review). As we will show, since neither the existing literature on innovation contests nor the encompassing open innovation literature offer a theoretical basis for approaching problem formulation for innovation contests, we briefly review two literature streams in organization theory that explicitly deal with innovation-related problem formulation and explore to what degree they can account for problem formulation for innovation contests.
Literature Review
Innovation Contests
Innovation contests are a form of crowdsourcing where a firm (the seeker) “broadcasts” (Jeppesen & Lakhani, 2010) a written problem statement to self-selecting external problem solvers and proposes a fixed monetary reward for solutions meeting pre-specified criteria (Taylor, 1995; Che & Gale, 2003; Schottner, 2008; Terwiesch & Xu, 2008). Innovation contests have mostly been used to solve “technical, innovation-related” problems (von Hippel, 1994), such as the need for a novel method to ensure the authenticity of a recycled PET polymer for fabric uses; the need for a computer algorithm to simulate insect behavior in response to a new pesticide used on wheat; and the specification of a new synthesis route for a chemical compound that satisfies specific criteria in terms of yield, purity, and cost. Innovation contests for such problems use “targeted prizes” posted ex ante, as opposed to “blue-sky prizes” which provide ex post rewards to innovation (Scotchmer, 2004); this chapter is only concerned with the former, which requires a written problem statement to run. Another central feature is that seekers either organize innovation contests themselves (as in the case of P&G’s Connect and Develop program, see Huston & Sakkab, 2006), or employ an innovation intermediary that maintains a large network of external problem solvers and consults seeker firms on how to effectively leverage innovation contests (Jeppesen & Lakhani, 2010; Sieg et al., 2010).
Innovation contests may offer significant rewards for firms (Green & Stokey, 1983; Taylor, 1995; Fullerton & McAfee, 1999; Che & Gale, 2003; Maurer & Scotchmer, 2004; Schottner, 2008; Terwiesch & Xu, 2008; Jeppesen & Lakhani, 2010; Boudreau et al., 2011): (a) The seeker only pays for solutions meeting its performance criteria and thus shifts the cost of failures (e.g. the sunk costs of problem-solving efforts) to external problem solvers; (b) the seeker stimulates external problem solvers, who often outnumber internal staff by far, to exert massively parallel problem solving efforts and thus potentially decreasing the time to find a solution; (c) the seeker stimulates problem solvers with heterogeneous knowledge sets and may thus profit from more diverse solutions than could be generated internally.
Critical to our analysis, innovation contests imply a split of the seeker firm’s innovation process into problem formulation inside the firm and problem solving outside the firm by external solvers. The internal problem formulation process needs to generate problem statements, indicating what problems the seeker wants solved, i.e. written descriptions of the problems and criteria for the desired solutions (Lakhani et al., 2007; Terwiesch & Xu, 2008). The statements are the “sine qua non” (Jeppesen & Lakhani, 2010: 1018) of innovation contests since solvers rely on them when developing solutions and since the statements are usually the only piece of information solvers receive from seekers (see e.g. Terwiesch & Xu, 2008; Jeppesen & Lakhani, 2010; Sieg et al., 2010). Existing research on contest design has largely adopted an economic perspective, addressing issues such as the optimal number of problem solvers (Fullerton & McAfee, 1999; Che & Gale, 2003; Terwiesch & Xu, 2008; Boudreau et al., 2011;), award structure (Taylor, 1995; Terwiesch & Xu, 2008), and the relative benefits of contests as incentive mechanisms (McLaughlin, 1988; Glazer & Hassin, 1988; Maurer & Scotchmer, 2004). Recently, researchers have also investigated innovation contests from an organizational perspective, shedding light on questions such as the problem solving effectiveness of innovation contests (Jeppesen & Lakhani, 2010) and managerial challenges for seeker firms (Sieg et al., 2010). However, to our knowledge, existing theory and research has not yet addressed the issue of problem formulation for innovation contests.
Cognizing Problems or Constructing Problems?
Two streams of literature on organization theory provide competing perspectives on problem formulation for innovation contests: the cognitivist perspective (e.g. March & Simon, 1958/1993; Cyert & March, 1963; Simon, 1996; Nickerson & Zenger, 2004; Felin & Zenger, 2012); and the constructionist perspective (e.g. Nonaka, 1994; von Hippel & Tyre, 1995; Carlile, 2002; Bechky, 2003; Nonaka & von Krogh, 2009). The cognitivist perspective holds that problems are abstract, universal, and objective cognitive representations by problem solvers, and views problem formulation as a by-product of problem solving. Managers deliberately select or stumble upon problems and subsequently organize the search for solutions (Cohen et al., 1972; Nickerson & Zenger, 2004; Hsieh et al., 2007). In contrast, the constructionist perspective views problems as pieces of subjectively held knowledge (Landry, 1995) that are actively formulated by applying, integrating, and creating knowledge.
Cognizing problems is a powerful image when problems have clear solution criteria and a given set of permissible solution moves within a solution space (Newell & Simon, 1972; Kotovsky et al., 1985), or when problem solvers start with an initial problem representation, develop a first solution, generate additional information about the problem by testing the solution, and adapt representations based on new information (Simon, 1977; Kulkarni & Simon, 1988; Fernandes & Simon, 1999; Klahr & Simon, 1999). The problem statement is here a cognitive or symbolic “representation” of a solution space (Fernandes & Simon, 1999). The cognitivist perspective can under certain conditions account for solving both simple puzzles such as “Missionaries and Cannibals” and complex and “ill-structured” problems (Simon, 1973). Indeed, decades of work on problem solving in innovation rests upon this basic perspective (e.g. Nelson & Winter, 1977; Nelson, 1982; Loch et al., 2001; Thomke & Bell, 2001). However, the perspective may be less useful when accounting for problem formulation as separate and precursory to problem solving, as the two processes are assumed to co-evolve.
Constructing problems is an alternative image that does not assume problems to be abstract, universal, and objective representations. The constructionist perspective on problem formulation can be traced back to the knowledge-based view of the firm (Kogut & Zander, 1992; Grant, 1996; Spender, 1996; Tsoukas, 1996), research on adaptive learning (von Hippel & Tyre, 1995; Tyre & von Hippel, 1997), the concept of communities of practice (Brown & Duguid, 1991, 2001), and organizational knowledge creation theory (Nonaka, 1994; Nonaka & von Krogh, 2009). In sum, problem-related knowledge depends on the physical, social, and technical context in which the problem is encountered (Dougherty, 1992; von Hippel & Tyre, 1995; Carlile, 2002; Bechky, 2003). Tyre & von Hippel (1997) stress that different physical contexts provide individuals with clues about the problem and opportunities to use their tacit problem-solving skills. Problem-related knowledge is therefore tacit or “sticky.” What an individual tacitly knows about a problem in one context cannot be shared with individuals in another context without effort; knowledge is less “universal or objective” than often assumed (Dougherty, 1992; von Hippel, 1994; Carlile, 2004). Efforts to share problem-related knowledge include social activities such as discussions, experiments, joint physical demonstrations, and face-to-face conversations (Knorr-Cetina, 1981; Carlile, 2002; Bechky, 2003; Hargadon & Bechky, 2006). Moreover, in order to formulate innovation-related problems, individuals must share tacit problem-related knowledge (Nonaka, 1994; Nonaka & von Krogh, 2009) through close face-to-face dialogue. These dialogues allow teams to create concepts that externalize tacit problem-related knowledge encoded in writing (Nonaka & Konno, 1998). Problem formulation for innovation contests may thus be understood as knowledge externalization. However, the existing literature considers externalization within a team where knowledge can be shared and justified through intense interactions. As explained later, such interactions are severely limited in innovation contests. The constructionist perspective, therefore, highlights two challenges for seeker firms: encoding tacit knowledge in written problem statements and making problem-related knowledge sharable with external solvers in different contexts. Armed with these insights, we use the rest of the chapter to explore how firms formulate sharable problems for innovation contests.
Research Design
Research Setting and Case Selection
We adopted a multiple case study design (Yin, 2003; Eisenhardt & Graebner, 2007) involving a method of constant comparison, cycling between data collection, data analysis, and further data collection based on emergent themes (Corbin & Strauss, 2008).
We employed a two-step theoretical sampling strategy. First, we identified “typical cases” (Miles & Huberman, 1994; Yin, 2003) by selecting firms that had worked with InnoCentive (representing a closed crowd, as discussed in Chapter 3 by Viscusi & Tucci, 2018), an innovation intermediary that specializes in organizing innovation contests for its clients (for a discussion about the role of intermediaries in innovation, see Howells, 2006). Second, final case selection followed the logic of replication (Yin, 2003). From the population of typical cases (N ~ 60) we selected six firms constituting “polar types” (Eisenhardt, 1989; Eisenhardt & Graebner, 2007) representing successful and unsuccessful cases. The underlying logic is to select cases that either predict similar results (a “literal replication,” here a successful outcome) or predict contrasting results for reasons that the generated theory can explain (a “theoretical replication,” here a non-successful outcome) (Yin, 2003). Three cases were deemed successful based on the firm’s ability to formulate a stream of problems for innovation contests, meaning that the firm made a long-term commitment to the innovation contest that over time yielded several submitted problems. The remaining three cases were classified as unsuccessful as none of these firms was able to formulate a stream of problems (i.e. either only a few problems submitted in total or several in a few “bursts,” thus not representing a stream of problems) and subsequently quit working with InnoCentive after a few contests. Table 6.1 provides overviews of the case firms. To protect their anonymity, we have labeled the successful cases Sbrinz, Appenzeller, and Schabziger, and the unsuccessful cases Baer, Tomme, and Flada.
Table 6.1. Overview of case companies.
Data Collection
We collected data from four sources, between October 2007 and September 2010. First, we interviewed a total of eighteen (senior) managers and (senior) scientists from the selected six InnoCentive client firms (several informants were interviewed repeatedly). The interviews followed semi-structured, open-ended guidelines and took the form of “guided conversations” (Yin, 2003: 89). We asked managers about their activities in the ongoing collaboration with InnoCentive and the activities of their staff in formulating problems. Second, we were granted deep access to InnoCentive. We conducted formal interviews with eighteen InnoCentive staff (senior managers, scientists, and salespeople). In these interviews, we asked questions about InnoCentive’s relationship with the case firms. Similar to the seeker interviews, we focused both on individual problem formulation processes and on capturing InnoCentive employees’ sense of how problem formulation “works.” Interviews with informants at seeker firms and InnoCentive were partially conducted in person and partially by telephone. The interviews lasted between fifty minutes and two hours; all of them were recorded and subsequently transcribed verbatim (resulting in about 600 pages of text). We also visited the InnoCentive headquarters for a three-day visit. During that time, we were granted free access to all InnoCentive premises. We held numerous informal conversations, observed the daily work in the building, and were invited to internal meetings. We took detailed notes of our observations and wrote them up within twenty-four hours. Third, we collected data from archival sources such as InnoCentive press releases and published problem statements. Data from archival sources totaled about 300 pages of text. Fourth, during the course of the research project we conducted eleven supplementary interviews with managers and engineers/scientists at other seeker firms from various industries, such as engineering products, consumer products, and chemical products. The purpose of these validation interviews was to crosscheck data points and interview statements and to solicit broader feedback on emerging ideas, constructs, and relationships. In all, we conducted forty-seven interviews and transcribed them verbatim. Appendix 6.1 provides examples of informants’ quotes.
Findings
Towards a Framework for Formulating Sharable Problems for Innovation Contests
The framework for formulating sharable problems for innovation contests consists of two dimensions—the problem formulation process and organizational roles within this process–captured as the horizontal and vertical dimensions of Figure 6.1. The framework is developed inductively from literal and theoretical replications of successful and unsuccessful cases (cf. Yin, 2003). In the successful cases, firms were able to formulate a stream of problems for innovation contests, while in the unsuccessful cases they were not able to do so. First, we detail the necessary process elements and roles and provide examples from the successful cases. We then present evidence ex negativo that all the generative process elements and organizational roles are conditions for seekers to formulate sharable problems for innovation contests.
Figure 6.1. A model of problem formulation for innovation contests.
The problem formulation process for innovation contests comprises three generative elements: problem generation, activities to motivate scientists to draft precise written statements of problems suitable for innovation contests; problem separation, activities to disentangle problems from their internal context by incorporating the perspective of solvers in drafting problem statements; and problem publication, activities to manage legal risks and control competitive implications.
The tasks within the three generative process elements are distributed among four organizational roles: developers, scientists/engineers who take ownership of problems and have the lead in generating, separating, and publishing specific problems; advisors, experienced scientists/engineers who assist developers in problem generation and separation; coordinators, R&D managers who motivate scientists to generate problems and are partially responsible for problem publication; and sponsors, senior managers who take the budgetary responsibility for innovation contests and help to generate problems by promoting innovation contests towards scientists. Developer, advisor, coordinator, and sponsor are our own labels for these roles and do not represent the job titles of seeker employees.
Generative Process Elements in Formulating Sharable Problems
Problem generation: Activating scientists and drafting problem statements. Problem generation for innovation contests involves activating scientists who hold problems, selecting suitable problems, and drafting problem statements. Since innovation contests represent a “new way of working,” as the research director of Sbrinz said, many scientists in the three case firms tended to focus on the risks entailed in innovation contests and were, therefore, not immediately prone to using innovation contests in order to solve one of their own problems. At Sbrinz, the research director therefore sought to promote a “dare-to-try attitude” by making scientists aware that innovation contests were a “calculated risk” (and thus acceptable) for the firm. In parallel, a mid-level R&D manager at Sbrinz was engaged in “internal marketing,” i.e. she held formal and informal meetings in which she promoted innovation contests as an opportunity for scientists to leverage external knowledge for solving their problems. These meetings also sought to make scientists feel comfortable with the legal and competitive risks inherent in innovation contests. Similarly, at Schabziger, several R&D managers activated scientists by “walking around” and asking about specific problems that could benefit from external resources.
Scientists as well as R&D managers in the three case companies were well aware of the costs of an innovation contest—the posting fee as well as the proposed reward. They sought a “rational” approach, by only attempting to generate “suitable” problems, i.e. only selecting those problems where an innovation contest would be “worth it” or cost effective. To assess the cost and benefits of innovation contests for certain problems, seeker employees engaged in several activities. First, multiple informants stressed that they tried to “make sure that you reached all the internal brain capacity before you go out” (quote), for instance, by tapping into personal networks and presenting the problem in different meetings within the R&D function. Second, R&D managers prioritized problems for submission by discussing the potential of problem candidates to advance various research projects. Third, seeker scientists also tried to ensure that the seeker organization would have a “real need” for the solution.
When seeker scientists become motivated to submit a suitable problem to an innovation contest, they need to draft a problem statement. However, as one informant put it, “…when you have a problem normally you don’t define the problem; normally the problem arises in the lab.” Seeker scientists, therefore, do not have a problem statement available; they rather need to draft problem statements by “put[ting] the problem on paper.” The writing process itself resembles a “mental simulation” in which the scientist tries to imagine a situation when she/he receives the solution. An informant said, “…we do try and think about what information we need to have to know if this really is a solution.” Scientists, thus, try to think upfront about technical requirements and often “end up working with the list of requirements” (scientist at Sbrinz). Sometimes, these requirements are framed negatively in terms of exclusion criteria and seeker scientists “think about what kind of solutions we don’t want to have,” as another informant put it.
In many cases, a crucial step to defining precise, scientifically valid solution criteria is to determine a problem’s causes, i.e. to trace observed problem manifestations back to a more fundamental scientific phenomenon. For example, if the absorption capacity of a diaper needs to be improved, this problem can be tracked back to the absorption capacity of different polymers in the diaper and ultimately to certain aspects of ion condensation that need to be controlled to improve the absorption capacity. In many instances, seeker and intermediary scientists facilitated the search for problem causes by asking questions, as described by one informant from InnoCentive:
The first three questions are “What’s the problem?” and then “What’s the underlying problem?” and then the third one says, “No, really, what’s the problem?” And…most people know what they want, right, they say we need this or we need that. And that’s not the problem, that’s a solution. […] So what I usually ask them is “Why don’t you just do it?” They say, “Well, we need this, you know, this part, you know, or we need this piece of equipment that does that, analyzes this.” And I will say, “Why don’t you just make one?” And they say, “We can’t” and so then I keep beating them going “Why not?” Because eventually you get to why we can’t because we don’t have this or we don’t know how to do that and then you start getting to what the actual problem is, right. And it’s probably only a small piece.
Other informants referred to this questioning process as “finding the endpoint,” getting to the “bottom of the problem,” “finding the underlying problem,” or “getting to the root cause.”
Problem separation: Giving and taking the solver’s perspective. Problems are entangled in their specific context through taken-for-granted assumptions held by scientists and others. At Sbrinz, for instance, “requirements that are kind of part of the culture everybody knows” comprised an environmental policy excluding a long list of chemical compounds from possible solutions. Separating problems from the internal seeker context requires the seeker to uncover such contextual assumptions. To this end, seeker scientists either cognitively take the perspective of external problem solvers or encourage others to give their perspective, e.g. by interacting with internal colleagues and intermediary scientists who attempt to simulate the perspective of a potential solver.
Our informants told us that taking the perspective of solvers or “putting yourself in the shoes of the solver” enables them to better understand how the problem is entangled in the specific seeker context. For instance, seeker scientists often took technical equipment for granted but recognized it as an important part of the problem’s context when they imagined that they were in the position of a solver. Imagining being a university student or retired engineer, seeker scientists realized that these solvers would “of course” have little equipment available to conduct experiments. Such mental simulations also allow seeker scientists to surface solvers’ time constraints and limitations regarding solvers’ ability to test potential solutions, as an informant’s reflection on the outcome of such a simulation makes clear:
You can’t say that it [the solution] can’t be something we know because then they [the solvers] don’t know if they are wasting their time. They can’t make a judgment. And you can’t say something like it shouldn’t be patented in the patent literature at large because that’s, it’s too much, it takes too long to do that search.
Putting oneself in the solver’s shoes does not only reveal solver constraints but also their problem-solving capabilities. For instance, one informant said that when writing a problem statement “…I assume that most of the people who are reading it are scientists or engineers.” Seeker scientists thus (sometimes) assume that problem solvers have had similar scientific training and understand the same concepts as seeker scientists. Using these concepts to describe a problem enables seeker scientists to disentangle the problem from the seeker’s specific context and to make it understandable for external solvers.
While taking the solver’s perspective was often an important first step for seeker scientists in problem separation, seeker scientists also found it helpful to discuss with internal colleagues or intermediary scientists to surface additional contextual assumptions. In these discussions, seeker colleagues and intermediary scientists “took” the perspective of problem solvers and “gave” it to seeker scientists. In other words, seeker colleagues and intermediary scientists represent the voice and interests of solvers in the discourse around drafting problem statements. Seeker colleagues, for instance, ask “naïve questions” that external problem solvers may pose such as: “Would this be an acceptable solution for you?” or “Is this parameter measurable, does a solver know when they’re done?” A seeker scientist who had run several innovation contests said that he often reminded colleagues that solvers typically work alone and, thus, have difficulty solving problems that require multiple skill sets.
Intermediary scientists attempt to make the perspective of solvers explicit by voicing concerns that the solvers may have, or describing their likely reaction to information in the problem statements. Through these interactions, seeker scientists often uncover hidden assumptions. For instance, one intermediary scientist described a typical interaction with a seeker scientist who had implicitly assumed that solvers’ technical capabilities could be ignored when defining solution parameters:
And then, if there are things that you [seeker scientist] would prefer, you say, you know, we’d like to be able to measure that […] at, let’s say, 100 ppm or something, but it would be really nice if we can measure it down at 10 ppm. And then solvers know, “Hey, if I can go even lower, mine has a better chance of winning.” […] They’re [solvers] not stupid, right, they all know to go lower. But what happens is, if you set it at 1 ppm even though you’d take 25 ppm, probably the smartest guys, the two guys that know the most about the problem look at one and go, “That’s impossible. I’m not even going to waste my time. I can get it to, let’s say, twenty but there’s no way I can get it to one” and they may just walk away.
Such “walking away” from problems that cannot be solved with existing skills is an important market feature, ensuring that those who stay are also those who are the most competent to solve the problem. However, by literally raising the “voice of solvers” and describing how the most competent solvers would react to the seeker scientist’s proposed solution parameters, the intermediary scientist demonstrated that the seeker scientist had implicitly focused on the internally most desirable solution parameters and, thereby, ignored many sufficiently competent solvers with feasible solutions to hand.
Several informants from seekers and the intermediary similarly reported that seeker scientists often (wrongly) assumed that solvers had the same access to data that they themselves enjoyed. For instance, an intermediary scientist reported that when discussing a modeling problem that required extensive time-series data, he told seeker scientists, “You know what, do me a favor, you go get it [data] and we’ll put it in the challenge. I don’t want the solvers having to do it.” By advocating solvers’ interests, the intermediary scientist made explicit for seeker scientists how much their problem statement relied on access to their specific data source. In several cases, seeker scientists also made an incorrect assumption that solvers could take on the same amount of risk as could seeker employees. To correct this flawed assumption, one intermediary scientist described solvers’ risk-taking propensity to a seeker scientist as follows:
If I have to spend $1,000 to try to make $20,000 and I don’t know if I’m going to get it or not I’m probably not going to do that, right? There’s, no guarantee. […] actually they’ll put in a tremendous number of hours, but if they have to, you know, pull $500 of their own money out then they’re like “maybe this isn’t worth the effort”.
Exposed to a large network of solvers, the intermediary scientist here performs a valuable function. By voicing solvers’ potential concerns, the intermediary scientist has made explicit that seeker scientists had assumed that solvers could spend the same resources on experiments that they could themselves. In hindsight, the seeker scientist described this interaction with the intermediary scientist: “We wanted the solvers to do some experiments, but then [after discussing with the intermediary scientist] we understood that probably it was too risky for the solvers.”
Problem publication: Managing legal risks and controlling competitive implications. Problem publication involves managing legal risks and controlling competitive implications of information to be published in the problem statement. For instance, according to an informant, publishing a problem might “alert competitors” and “lead them to our ground” which “can block us” if competitors were developing intellectual property rights (IPR-) protected solutions to the problem. To evaluate the severity of such competitive implications, seeker employees tried to guess competitors’ current state of knowledge. Seekers sometimes developed a work-around, for instance, by not asking for a novel synthesis route to the desired chemical compound, but to a structurally similar one.
Making problem-related information public can also jeopardize future patenting activities, as publicized information can no longer be part of a patent application. The seeker organizations, thus, gauged the sensitivity of information by first discussing it internally. Both Appenzeller and Schabziger, for instance, defined a contact person within the legal/IP department who was responsible for assessing the legal and competitive risks and then clearing each problem statement. At Sbrinz, seeker lawyers were involved in negotiating a master services agreement, but individual scientists were entrusted with evaluating whether a specific piece of information was publishable. The seeker organization then either decided to accept the risk that some sensitive information would be published, on the grounds that any innovation contest involves some level of legal risks or disguised sensitive information. For instance, the seeker organizations had been working on problems to be submitted to innovation contests internally and, thus, knew about many partial solutions. Seekers did not want to pay for these known partial solutions but faced the risk that solvers might sue them if these solutions were later part of a patent application. In light of this risk, seekers either excluded whole solution approaches (e.g. any synthesis route involving a particular compound) or defined a “filter.” In the latter case, seeker employees set up criteria according to which the innovation intermediary would filter out submitted solutions before forwarding solutions to the seeker. Seekers also prepared documentation of internal work on the problem to be able to prove that these partial solutions had actually been generated internally.
Organizational roles in problem formulation for innovation contests. The developer is a technical/scientific role that is usually filled by bench scientists or engineers within the seeker organization. Developers are essential to the problem formulation process: they carry out most operational tasks, i.e. they write and rewrite the problem statement and, thus, spend more time and resources on formulating a specific problem than any other role. In other words, while all roles are necessary, most work in innovation contest problem formulation falls onto the developer. Because of their investment in time and resources, developers usually have a strong and genuine interest in getting the problem resolved.
The advisor is also a technical/scientific role often filled by senior and well-respected scientists/engineers from the seeker organization and the innovation intermediary. Compared with developers, advisors remain more passive during problem generation. Advisors typically wait to act until called in by developers or coordinators to help to draft a problem statement. Advisors are also involved in problem separation. During problem separation, advisors support developers by drawing on experience from the problem context, and from having participated in previous innovation contests where they have familiarized themselves with the specific problem formulation requirements as well as knowledge of the problem itself.
In contrast with developers and advisors, the coordinator is a management role involved in both problem generation and problem publication. In problem generation, coordinators serve as a “spokesperson” or “ambassador…who carries the flag” of innovation contests, according to one informant. Coordinators also play an important role in problem publication. They do so by providing managerial approval to select a particular problem (over other problems) and a specific problem statement (over alternative statements), thereby removing much uncertainty and anxiety from developers regarding priorities and the release of potentially sensitive information.
The sponsor is a senior management role filled by a person within the seeker organization who is ultimately responsible for innovation contests. Sponsors are crucial for problem generation since they have the internal clout to build awareness for and legitimacy of innovation contests among scientists and mid-level R&D managers.
Theoretical Replications of Unsuccessful Cases
In this section, findings from three theoretical replications of unsuccessful cases (Yin, 2003) demonstrate how deviations from the three generative process elements and four organizational roles hinder seekers from formulating streams of problems for innovation contests.
Deviations in problem generation and the coordinator and sponsor roles. Innovation contests were introduced “top-down” in all three cases. After senior R&D managers who assumed the sponsor roles had decided to use innovation contests, scientists were given the opportunity to submit problems. However, many scientists were worried about shouldering the work load of a problem formulation process, alerting competitors by publishing “good” problems, and running into patenting problems if solvers submitted solutions touching on current patenting activities. As sponsors did not anticipate these concerns, they focused on promoting the economic rationale behind innovation contests and did not activate scientists with problems. At Baer and Flada, sponsors thus “extracted” problems by selecting the most “visible” problems and assigning individual scientists or research groups to submit these problems to innovation contests. At Tomme, the sponsor remained rather passive and waited for research groups to approach him with problems. Given these introductions of innovation contests to the firm, there were two similar implications in all three cases. First, at none of the firms did the sponsors perceive the need for an R&D manager to assume a coordinator role, working with scientists to select suitable problems that would have a good chance of being solved. Second, seekers instead selected “holy grail” or “crapshoot” problems—long-standing, relatively complex problems with low solution chances.
In all three cases, innovation contests met with negative feedback from seeker scientists and yielded an “objectively low” solution rate. These were two major factors that contributed to the reported perception of failure and the decision to quit running innovation contests. The findings show that the formulation of problems for innovation contests requires careful attention to problem generation. The findings also underscore that the construct of “problem generation” does not correspond to the identification or extraction of the most visible problems (as in the case of Baer, Tomme, and Flada), but to “activation” of scientists with problems and thereafter selection of suitable problems. Sponsors and coordinators play a vital role in activating scientists and selecting problems by discussing with developers which problems they are motivated to formulate and which they perceive as having a good chance of being solved.
Deviations in problem separation and the advisor role. At Baer, Tomme, and Flada, we observe that sponsors did not associate problem formulation for innovation contests with problem separation. Instead, sponsors approached problem formulation as the technical task of writing down the problem. Sponsors at these three case firms expected the innovation intermediary to perform this technical task in return for the intermediary’s fee. However, developers gradually learned about the challenges of problem separation as they received feedback from the intermediary scientists and then realized that the problem formulation process would require significant allocation of time and effort. Since sponsors did not approach the challenge of problem separation, they also did not designate internal advisors, who would have assisted developers in problem separation. In retrospect, as Flada’s CTO remarked, this created a difficult situation in which “the scientists were very much exposed to the InnoCentive people, so they had to do a lot on their own with little support within Flada.” A lack of internal support meant that problem separation was less effective, since internal advisors did not explain to developers how to turn a problem description into a document that fits with InnoCentive guidelines and would be understandable for external solvers. Intermediary scientists did assist developers with these challenges but, as outsiders, they were less capable of explaining how a separated problem statement would differ from standard internal problem descriptions. Developers also lacked informal “sparring partners” who were well acquainted with their specific research problems. In sum, the lack of an internal advisor role made problem separation more cumbersome for internal scientists and thus undermined the formulation of problems for innovation contests. The three cases provide evidence that problem separation should involve an internal advisor role.
Deviations in problem publication and the coordinator role. At Baer, Tomme, and Flada, scientists drafting problem statements reported that controlling sensitive information in the problem statements and evaluating the commercial effectiveness of innovation contests fell under their individual responsibilities. Scientists, in other words, did not perceive the competitive and legal risks of innovation contests to be of a magnitude that would affect the whole organization and that would need its overall commitment, but rather as risks facing them individually. Scientists sought to minimize their personal risks to the detriment of innovation contest effectiveness in the seeker organizations. For instance, rather than controlling information and appraising the sensitivity of the problem, scientists reported minimizing sensitive information in problem statements. At Baer, Tomme, and Flada, we observed that many problems came from areas where the firms had long research histories and, thus, knew about many solution approaches. Since the firms were patenting some of these approaches, developers had to define a filter, i.e. exclude some solution approaches in the problem statement, to avoid solvers submitting solutions the firms already knew about. Such known but not IP-protected solutions could have caused the firm to pay royalties for IP in a known solution. According to the CTO of Flada, “The pressure on scientists was such that they felt that if they don’t define that filter right they will look stupid because sooner or later we might run into some difficulties [in patenting processes].” Scientists, thus, defined very narrow filters that excluded many solution approaches. Since there was no coordinator role at Flada, no one counteracted the scientists’ tendency to be overly conservative in avoiding patenting risks. Moreover, senior R&D management was too far removed from individual scientists’ work on technical problem formulation to sense and address these concerns quickly enough. In sum, the three cases provide evidence that problem publication needs to involve control of sensitive information and appraisal of the value of external knowledge. Coordinators should support scientists in these activities because scientists will otherwise resort to minimizing instead of controlling sensitive information, thus appraising external knowledge extremely conservatively and undermining the chances of obtaining solutions from external solvers.
Discussion, Conclusions, and Future Research
Implications for Crowdsourcing
The study demonstrates how crowdsourcing in the form of innovation contests stirs up “encrusted routines” of innovation, how problem formulation enters as a critical challenge, and how managerial actions facilitate effective problem formulation. First, innovation contests tend to—primarily—disrupt the work of scientists and engineers. People who were previously working within insulated functions and roles were suddenly exposed to new challenges and risks. These findings are in line with Alexy et al.’s (2013) findings that the adoption of open source software development influenced different job roles quite differently. Here, the importance of desk scientists is underscored by the fact that they are involved in all elements of problem formulation.
Second, our research demonstrated that problem formulation consists of three elements: problem generation, problem separation, and problem formulation. Failure to recognize any of these elements will invariably lead to failure. The challenge to formulate problems is underscored by the fact that in the practice of engineers and bench scientists, continuous streams of problems, solutions, decisions, and actors make innovation more similar to a garbage-can process with stochastic properties than a linear and rational process of separable problem formulation and solving. This image is mirrored in the literature as well (e.g. March & Simon, 1958/1993; Cyert & March, 1963; Cohen et al., 1972; Simon, 1996; Nickerson & Zenger, 2004; Felin & Zenger, 2012; von Hippel & von Krogh, 2016). However, innovation contests turn this image of innovation on its head. Now problems are exposed to radical incisions; formulation and solution are partitioned in content, time, and space as they are allocated to external and often unknown solvers. To mitigate the challenge of problem generation, management could encourage, support, and activate engineers and scientists who already have possible issues and problems in mind—rather than just telling them to draft problem statements. Our findings on problem separation complement those in the recent literature on innovation contests. For example, Boudreau et al. (2011) found that problems requiring solutions from multiple knowledge domains benefited from large numbers of solvers, whereas restricting entry to the contest is better for single-domain problems. Their result dovetails our finding to build solver representations through perspective giving and perspective taking. Seeker firm personnel need to understand the nature of the problem, how solvers perceive it, and who potential solvers may be, for example, whether they are “marginal” solvers or not (Jeppesen & Lakhani, 2010). In this regard, innovation intermediaries may play a crucial role.
Implications for Open Innovation
The current study adds to the fast-growing literature on open innovation. First, our research demonstrated that problem formulation is a critical antecedent to successfully solving technical problems through innovation contests. Yet, problem formulation is largely inconspicuous in received open innovation literature. For example, in their extensive review West & Bogers (2014) develop a process model of how firms leverage external sources of innovation. Their four-phase linear model includes (1) obtaining, (2) integrating, (3) commercializing, and (4) interactions between the firm and its collaborators. Our study suggests a step zero (0) that precedes obtaining solutions: formulating the problem. Indeed, West & Bogers write that is was impossible to separate search from acquisition—a testament to how intertwined problem solving and problem formulation are in theory as well as in practice. However, as we have demonstrated, under some circumstances they need to be separated.
Second, we expand and explain how problem formulation may impact other forms and aspects of open innovation. For example, recent work has explored how seeker firms initiate collaboration through selective revealing (Alexy et al., 2011). The seeker firm moves beyond building absorptive capacity to exploit external knowledge (Cohen & Levinthal, 1990) to stimulating external problem solvers who next create knowledge. Our model adds that the seeker firm’s capacity to stimulate solvers relies on specific mechanisms of problem formulation. For example, open innovation that stimulates knowledge creation through contract research (e.g. Howells, 1999) or R&D alliances (e.g. Oxley & Sampson, 2004) requires a formulation of the problem being solved. Accordingly, problem formulation research provides an entrée to analyzing how firms adapt and change when they innovate with external partners. Analogous to building and maintaining absorptive capacity for open innovation (Harison & Koski, 2010; Robertson et al., 2012), firms may need to build a “problem formulation capacity” by overcoming Laursen & Salter’s (2014) “paradox of openness.” The capacity for formulating problems needs to mitigate perceived risks, such as the limits to IPR protection (as most knowledge does not meet the bar for patentability). It may also be that problem formulation capacity is more fundamental to the boundary of the firm than problem solving capability. Clearly, firms can draw on outside expertise to solve a problem, when the problem is well defined—but formulating the problem properly is a necessary condition for solving it. We are reminded that problems are partly solved when they are well defined. Indeed, when the problem is well defined, finding a solution becomes a task that can be broken down into assignments, automated to some degree, and managed through project management principles. Problem formulation for innovation contests is a different story altogether.
Third, our work also shows the limitations of innovation contests as an effective “organizational design” for innovation (Felin & Zenger, 2012). The lack of seeker–solver interaction excludes certain problems from being effectively defined. This is the case with a rapid increase in the cost of building solver representation, for example when solvers, their resources, and their capabilities are entirely unknown at the outset, forcing the firm to gather information about potential solvers (e.g. a problem akin to market research for a new product) (see Chapter 7, by Ranade & Vershney, 2007, on the advantages of crowdsourcing when the seeker is unsure about solver strength). Alternatively, the contest method can only solve problems when the scale of solver communities offsets the negative effect of a problem that lacks important information. By broadcasting the problem to a very large group of potential solvers, the seeker can increase the probability that some solvers will find the problem rewarding—and therefore that a desired solution can be found. Yet broadcasting problem statements to solvers is costly (and risky), such as having to evaluate many solution candidates or setting up communication infrastructures. When building solver representation is too costly (or even impossible) or when scale advantages cannot be reached at a reasonable cost, other organizational forms that allow for greater interaction between seekers and solvers may be more effective.
Implications for Innovation Intermediation
Our chapter also contributes to the literature on innovation intermediation (Howells, 2006; Sieg et al., 2010). First, the study underscores the importance of innovation intermediary alignment. The innovation intermediary needs to align the interests of the seekers and solvers, and to do so it must have better access to and knowledge of the solver community than the seeker firm. By serving multiple seeker firm clients, the intermediary maintains a greater diversity in its solver community than any seeker firm is likely to. Such diversity is good for innovation contest methods (Jeppesen & Lakhani, 2010) and is likely to require scale in the number of contests to be upheld. The innovation intermediary achieves alignment by helping the seeker firm’s management to mitigate perceived risks in the contests and by building solver representations. We found that InnoCentive’s deep understanding of its solver community guided the seeker firm’s management to strike a fine balance between protecting proprietary knowledge and releasing effective problem statements to the solvers. The innovation intermediary often represented the “voice of the solver” and reminded seekers firm’s personnel of solvers’ resources and capabilities.
Second, drawing on these observations, we contend that for seeker firms, not knowing the “potential” solver could jeopardize the outcome of innovation contests. Firms that want competitively or legally sensitive problems to be solved should not simply release statements into the public. Accessing a community understood and supported by an intermediary is a better approach, because it allows protection and enables the seeker firm to formulate a problem attuned to expert solvers. Innovation intermediation enables the seeker firm to remain anonymous, if needed. While sometimes beneficial, anonymity also creates new challenges for the intermediary. The name and industry of seeker firms could provide important cues to solvers and when they are not published, the intermediary needs to find ways of extracting other relevant information from the firm and channeling this to the community. Not surprisingly, our interviews with managers indicated that innovation intermediation is considered a safe and cost-effective way of gaining access to a large pool of outside talent.
Limitations and Future Research
Our study faces limitations. The sample includes firms of different size and from different industries, and it tilts towards relatively large chemical, pharmaceutical, and consumer products firms. The sample firms primarily submitted chemical and biological problems to innovation contests. Future research should explore effective problem formulation for innovation contests in other industries and contexts. Our work focused on seeker firms, and the intermediary’s relationship with seeker firms and solvers. Henry Chesbrough has remarked that a challenge for innovation intermediaries is to scale operations to effectively assist a large number of firms in innovation contests.3 The research issue deserves extensive attention in future work. The current work suggests that managers consider the innovation contest a successful open innovation method when it solves a technical problem that has been effectively formulated with the help of an intermediary. An intriguing theme for future work is how the intermediary’s assistance in problem formulation across seeker firms can be standardized and scaled.
Conclusion
Drawing on rich empirical cases, we inductively develop a theoretical framework outlining mechanisms for seeker firms’ formulation of sharable problems in innovation contests. The framework consists of three generative process elements (“problem generation,” “problem separation,” and “problem publication”) and four organizational roles (“developer,” “advisor,” “coordinator,” and “sponsor”). The chapter responds to recent calls to investigate the intra-organizational implications of open innovation by illuminating the organizational processes and roles necessary to stimulate external knowledge creation in successful innovation contests. The study shows that open innovation in general, and innovation contests in particular, hold strong implications for organization theory by questioning the orthodoxy of problem solving.
Acknowledgments
The authors would like to thank Joel West and Linus Dahlander, conference attendees at Imperial College London, and seminar attendees at Bocconi University for their valuable comments. We appreciate the advice and encouragement from the editors and gratefully acknowledge the support from respondents in numerous case companies.
Appendix A6.1
Examples of informants quotes.
| Process elements | First-order concept | Second-order theme |
|---|---|---|
| Process elements | ||
| Example: “The way we chose was to involve the scientists from the start very much to let them come up with the problems. So we don’t say you should have a challenge, we more ask them and that’s the way we have been working all along and I think that’s one thing that’s important. You couldn’t order somebody to make a challenge.” Sbrinz manager | Asking employees for potentially suitable problems | Activating scientists with problems |
| Example: “You have to know what you’re going to do with the answers and what things you’re going to be looking for to know which one [solutions] you like.” Sbrinz scientist (I) | Determining the need for a solution | Selecting suitable problems |
| Example: “We do spend quite a bit of time in determining what the strict criteria are that we would use to judge our solutions in terms of precision, if it’s in terms of compounds, purity, the amount, things like that.” Appenzeller manager (I) | Determining the parameters necessary to evaluate solutions | Drafting problem statements |
| Problem separation | ||
| Example: “I think it’s an attempt really to help the solvers not waste their time either, because if there is a situation where we have tried it and it doesn’t worked, we would tell the solver that.” Appenzeller manager (II) | Estimating the requirements for solvers | Taking the solvers’ perspective |
| Example: “I thought a lot about this and talked to a lot of people in-house, how to overcome the problem that, I mean, this is our main business really. It’s our main competence. We know so much about this and how to produce a question that gets new insights in an area that we already know so much about.” Sbrinz scientist (II) | Discussing the problem with internal colleagues | Giving the solvers’ perspective |
| Problem publication | ||
| Example: “It could be that we are working on some technique that hasn’t been used for that purpose before, and if we then say, and perhaps we don’t even want to mention that because we see that we want to patent that and then we need also to keep it as confidential as possible to be able to patent it. I can’t give you really a very good example, but there are cases that we don’t want to reveal things that we are, because we are in the process of patenting or so.” Sbrinz manager | Disguising legally sensitive information in the problem statement | Managing legal risks |
| Example: “The hard part is knowing what exactly is it that I want and how much of that I can tell on this site and if I can’t tell enough it doesn’t make any sense what I’m asking you all and as an example there’s this order challenge and I’m sure you’ll hear about that because I haven’t helped to write it, but I heard about it, I’ve been participating in the discussions because it was an early one and they had a lot of discussions that they don’t want to say it’s a [name of product category] because it’s a little too much.” Sbrinz scientist (I) | Deleting competitively sensitive information in problem statements | Controlling competitive implications |