2. Background

The new product development process is inherently ambiguous, and key facets of the process, such as the project scope (Stacey and Eckert, Reference Stacey and Eckert2003), are often unclear. Designers and product development teams experience sudden changes in problem constraints (Duclos et al., Reference Duclos, Otto and Konitzer2010; McComb et al., Reference McComb, Cagan and Kotovsky2015) or shifts in project requirements (Clarkson et al., Reference Clarkson, Simons and Eckert2004) and are tasked with generating well-defined solutions to ill-defined problems (Dorst, Reference Dorst2006). Part of the uncertainty in new product development stems from a gap between customers’ expectations and the resources available to designers to meet these expectations. This gap induces a resource scarcity, or the inability to cope with resource deficits (Zhao and Tomm, Reference Zhao and Tomm2018) such as time (Zika-Viktorsson et al., Reference Zika-Viktorsson, Sundström and Engwall2006), money (Koshe and Jha, Reference Koshe and Jha2016), human capital (Brown et al., Reference Brown, Adams and Amjad2007) and physical resources such as building materials or tooling capabilities (Jaselskis and Ashley, Reference Jaselskis and Ashley1991). Dealing with a resource scarcity changes how people approach problems and causes them to deviate from rational and logical decision-making (Shah et al., Reference Shah, Mullainathan and Shafir2012). For instance, decisions made in the presence of scarcity are often based on a fixation that causes individuals to neglect important tasks to accommodate urgent, but less important tasks (Kazmer and Haythornthwaite, Reference Kazmer and Haythornthwaite2001) that ultimately worsen the state of scarcity (Zhao and Tomm, Reference Zhao and Tomm2018). Dealing with these scarcities can hinder designers’ cognitive function, depleting cognitive resources such as attention and working memory (Zhao and Tomm, Reference Zhao and Tomm2018), inducing cognitive overload (Shah et al., Reference Shah, Mullainathan and Shafir2012) and decreasing their ability to make effective design decisions. Marston et al. (Reference Marston, Allen and Mistree2000) claimed that design decisions are an irrevocable allocation of resources and thus, a designer’s ability to utilize resources determines their ability to create successful products (Hazelrigg, Reference Hazelrigg1996).

The availability of resources, such as time, money or materials, can change throughout the new product development process and can affect a team’s ability to solve problems (Schrader et al., Reference Schrader, Riggs and Smith1993; Zhao and Tomm, Reference Zhao and Tomm2018). For instance, resource management skills such as resource planning, allocation and scheduling can decrease project duration (Schrader et al., Reference Schrader, Riggs and Smith1993) and are critical for success, especially in competitive environments such as new product development (Chua et al., Reference Chua, Kog and Loh1999; Xin Chen et al., Reference Xin Chen, Moullec, Ball and Clarkson2016). Proper resource utilization requires both efficiency, or how well something is done, and effectiveness, or how useful something is. To use resources efficiently and effectively, a team must understand the problem they are trying to solve and identify the appropriate resources to solve it. To quantify this ability, Letting et al. (Reference Letting, Calpin, Soria Zurita and Menold2023) developed a resource accuracy metric that defines an individual’s resource accuracy as their ability to allocate relevant resources to critical obstacles they face. The allocation of relevant resources was studied because not all resources are inherently valuable or useful (Ketchen et al., Reference Ketchen, Hult and and Slater2007); rather, the value of a resource is derived from the ability of a team to use the resource to overcome a particular obstacle. In an example from the prior work (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023), a participant was struggling to educate their customer about their product; they indicated that “whenever [I] bring up the technology, it just goes over people’s heads.” The participant then utilized educational resources on LinkedIn to help them better communicate the product to their customer. In this instance, LinkedIn is considered to be a relevant resource for educating customers about obstacles because it directly helps the participant to overcome the obstacle. In contrast, another participant identified funding and building a business model as their main obstacles, but when asked about resources they were using, they only identified social networking. While networking may be generally helpful to the participant or their team, it will not directly solve their lack of funding nor will it teach them how to build a business model.

We argue that an interesting case of new product development teams are technology-based startups. Similar to design teams, startup founders employ resources to generate ideas (Kanchana et al., Reference Kanchana, Divya and Beegom2013; Yilmaz et al., Reference Yilmaz, Daly, Seifert and Gonzalez2015; Deo et al., Reference Deo, Blej, Kirjavainen and Hölttä-Otto2021), explore the diffusibility of these ideas through prototypes (Gill et al., Reference Gill, Sanders and Shim2011; Camburn et al., Reference Camburn, Viswanathan, Linsey, Anderson, Jensen, Crawford, Otto and Wood2017; Nelson et al., Reference Nelson, Mahan, McComb and Menold2020), and bring ideas to market just like any product development team (Stone et al., Reference Stone, Tumer and Van Wie2005). Due to these parallels, we argue that startup founders represent a unique subset of new product development teams and are ideal to study the effects of resource constraints on authentic problem-solvers. Employing over 1.5 million people in the United States and accounting for approximately 2.8 percent of US businesses (Wu and Atkinson, Reference Wu and Atkinson2017), startups have significant economic impacts both nationally and globally and are critical for innovation and knowledge transfer (Colombo and Piva, Reference Colombo and Piva2008). Despite their impact, only 10 percent succeed, many of which come close to dissolution (Marmer et al., Reference Marmer, Herrmann, Dogrultan, Berman, Eesley and Blank2011). These teams tend to fail because of the scarcity induced by dealing with resource constraints (Wymer and Regan, Reference Wymer and Regan2005). Often times, when a founder is experiencing a scarcity of financial resources, for example, they tend to fixate on obtaining more funding and neglect other obstacles or problems that they need to solve. In these cases, the founder could find novel ways to allocate resources that may be more abundant to solve their problems, but they are unable to do so because of the scarcity-induced fixation. Marcon and Ribeiro (Reference Marcon and Ribeiro2021) identified six categories of resources that foster success and competitive advantage for startup firms: financial, human, social, organizational, physical and innovation. Financial resources are often used to acquire other types or resources (Ireland et al., Reference Ireland, Hitt and Sirmon2003) and are commonly obtained from personal investments, venture capitalists’ investments and accelerators (Pauwels et al., Reference Pauwels, Clarysse, Wright and Van Hove2016). Human resources, or human capital, can include hiring and training team members to improve their knowledge and necessary skills (Barney, Reference Barney1991), whereas social resources include intrafirm and interfirm relationships (Fukugawa, Reference Fukugawa2018) from which resources can be shared (Adner, Reference Adner2006). Business resources such as the structure of the team and how it conducts work comprise the organizational resource category (Löfsten, Reference Löfsten2016). Physical resources include materials or equipment related to the necessary technology or product that is being developed (Barney, Reference Barney1991) and innovation resources are resources that lead to product development and commercialization (Ireland et al., Reference Ireland, Hitt and Sirmon2003; Löfsten, Reference Löfsten2016).

The field of psychology has identified individual traits that affect the ability of individuals to solve problems and manage resources. While there are many individual traits that affect individuals’ behaviors, such as creativity (Glăveanu, Reference Glăveanu2018), perseverance (Scherer and Gustafsson, Reference Scherer and Gustafsson2015) and cognitive flexibility (Krems, Reference Krems2014), this work focuses on four specific traits that are particularly relevant within the context of resource-scarce environments: risk propensity, self-efficacy, bricolage and perceptions of psychological safety. These traits are particularly interesting within the context of new product development teams because not only have they been linked with increased problem-solving skills (Edmondson, Reference Edmondson1999; Baker and Nelson, Reference Baker and Nelson2005; McGee et al., Reference McGee, Peterson, Mueller and Sequeira2009; Zheng et al., Reference Zheng, McAlack, Wilmes, Kohler-Evans and Williamson2009) but they have also been shown to improve team performance (Edmondson et al., Reference Edmondson, Kramer and Cook2004; Hmieleski and Baron, Reference Hmieleski and Baron2008; Kariv and Coleman, Reference Kariv and Coleman2015; Raveendra et al., Reference Raveendra, Rizwana, Singh, Satish and Kumar2018). Here, it is important to distinguish between risk tolerance and risk propensity. Risk tolerance represents the level of uncertainty, or risk, that an individual is willing to accept in decision-making (Grable, Reference Grable2000). In contrast, risk propensity represents an individual’s inherent tendency, or willingness, to take risks (Sitkin and Weingart, Reference Sitkin and Weingart1995). In the context of problem-solving, dealing with resource scarcities has been shown to exaggerate risk-averse behaviors (Zhao and Tomm, Reference Zhao and Tomm2018), and prior work has shown that risk-averse designers tend to choose less creative concepts (Toh and Miller, Reference Toh and Miller2016) and abandon novel ideas (Starkey et al., Reference Starkey, Toh and Miller2016), ultimately decreasing team performance. In fact, risk propensity has long been considered a defining characteristic of founders (Mill, Reference Mill1871; Kilby, Reference Kilby1971; Al-Mamary et al., Reference Al-Mamary, Abdulrab, Alwaheeb and Alshammari2020), and recent work has found risk propensity to have a positive impact on entrepreneurial intention (Moraes et al., Reference Moraes, Iizuka and Pedro2018; Al-Mamary and Alshallaqi, Reference Al-Mamary and Alshallaqi2022). Further, founders with high levels of self-efficacy are more comfortable taking risks than those with low self-efficacy (Densberger, Reference Densberger2014).

Social cognitive theory posited by Bandura (Reference Bandura1989) suggests that self-efficacy, or the belief that an individual has in their ability to achieve a certain outcome (Bandura, Reference Bandura1977), determines their motivation and perseverance through challenges and obstacles. Self-efficacy, however, is considered to be domain-specific. Stemming from Bandura’s self-efficacy, entrepreneurial self-efficacy (ESE) represents an individual’s belief in their own ability to successfully execute entrepreneurial tasks. ESE has been linked with persistence and success in entrepreneurial endeavors (Cardon and Kirk, Reference Cardon and Kirk2015; Santoro et al., Reference Santoro, Ferraris, Del Giudice and Schiavone2020) and high levels of ESE have been shown to increase firm performance (Hmieleski and Baron, Reference Hmieleski and Baron2008) and resource management skills (McGee et al., Reference McGee, Peterson, Mueller and Sequeira2009). Chen et al. (Reference Chen, Greene and Crick1998) suggested the use of ESE to assess the potential of an individual’s ability to effectively manage resources. More specifically, they assert that individuals assess their ESE in reference to available resources and obstacles they face. While multiple studies have demonstrated the importance of ESE, very few studies have investigated its relationship with other critical traits such as bricolage.

The ability to make use of limited resources to solve emerging problems, or bricolage (Baker and Nelson, Reference Baker and Nelson2005), has also been individually linked with increased performance (Kariv and Coleman, Reference Kariv and Coleman2015). In a study on startup firms, Baker and Nelson (Reference Baker and Nelson2005) found that successful teams possessed the ability to consistently adapt, recycle or reimagine resources in novel ways. This ability is also correlated with new venture adaptability and firm growth (Yu et al., Reference Yu, Li, Su, Tao, Nguyen and Xia2020). Even for nascent, resource-constrained firms, Senyard et al. (Reference Senyard, Baker, Steffens and Davidsson2014) identified bricolage as a pathway to innovation. Extending this work, Stenholm and Renko (Reference Stenholm and Renko2016) found that successful bricoleurs exhibited greater entrepreneurial passion and were less likely to quit entrepreneurial endeavors. For teams that operate under resource constraints, such as those that persist in startups and new product development, this ability to be resourceful and leverage combinations of resources is critical for success.

Team success is largely dependent on team cohesion as well. Edmondson’s theory of psychological safety (Edmondson, Reference Edmondson1999), or the shared belief that a team is safe for interpersonal risk-taking, is a team-level construct that can be used to approximate team climate. Psychological safety is particularly critical for team performance in knowledge-intensive and complex tasks that require creativity and sense-making (Sanner and Bunderson, Reference Sanner and Bunderson2015). A team’s level of psychological safety affects their willingness to share knowledge and communicate openly (Edmondson et al., Reference Edmondson, Kramer and Cook2004), and when team members are willing to communicate openly, they contribute more efficiently to the team (Edmondson and Lei, Reference Edmondson and Lei2014) and leverage the wisdom of the collective (Salas et al., Reference Salas, Rosen, Burke and Goodwin2009), where teams that successfully interact yield greater success. Psychological safety has been shown to mediate the relationship between inclusive leadership and innovative work behavior in management (Javed et al., Reference Javed, Naqvi, Khan, Arjoon and Tayyeb2019) and is positively related to innovation performance and innovation capabilities when studied as an organizational-level construct (Andersson et al., Reference Andersson, Moen and Brett2020). In the context of design, psychological safety has been identified as a dynamic and reliable measure of team climate over time, meaning that a team’s psychological safety can be fostered and measured throughout the design process (Cole et al., Reference Cole, O’Connell, Gong, Jablokow, Mohammad, Ritter, Heininger, Marhefka and Miller2022a). Cole et al. (Reference Cole, Marhefka, Jablokow, Mohammed, Ritter and Miller2022b) found that teams with greater levels of psychological safety generated higher-quality ideas than those with lower levels of psychological safety. Cauwelier et al. (Reference Cauwelier, Ribiere and Bennet2019) studied French and American engineering teams and discovered that increased team psychological safety leads to increased learning and knowledge creation.

While psychological safety is a team construct, the effects of individual perceptions of psychological safety have also been studied in prior work. Individual perceptions of psychological safety represent a single team member’s own perceptions of their team’s climate. By studying individual perceptions of psychological safety, we can draw insights on how members of the same team experience varying levels of psychological safety. For instance, in a study on design teams, Cole et al. (Reference Cole, Jablokow, Mohammed and Miller2023) identified a discrepancy in perceptions of psychological safety between men and women. Specifically, women exhibited greater levels of perceived psychological safety when working with other women when compared to working with men. In contrast, men did not indicate any difference in perceived psychological safety with either gender. Startup teams, like design teams, operate in resource-scarce environments and work to solve complex problems. As such, team psychological safety has also been investigated in the field of new product development; it has been linked with greater levels of bricolage (Cunha, Reference Cunha2005; Faia-Correia and Cunha, Reference Faia-Correia and Cunha2007), or the ability to “create something from nothing” (Baker and Nelson, Reference Baker and Nelson2005). In preliminary work, the authors studied founders and identified individual perceptions of psychological safety to be a predictor of entrepreneurial bricolage (Letting and Menold, Reference Letting and Menold2023). This finding suggests that members of problem-solving teams who perceive their team as psychologically safe may be better equipped to combine resources in novel ways.

Significant prior literature has shown these factors – self-efficacy, bricolage, risk propensity and psychological safety – to be critical to the success of problem-solving teams, and more specifically, new product development teams. No work has studied these traits together, especially within the context of resource-constrained, problem-solving teams. However, prior work in the fields of design and entrepreneurship have employed various techniques to identify distinct groups, or archetypes, of individuals based on a combination of relevant traits. Archetypes allow researchers to study patterns of behaviors or traits across groups of individuals. For instance in design, Begnum et al. (Reference Begnum, Pettersen and Sørum2019) employed qualitative methods to identify different archetypes of interaction designers and described the strengths of each archetype. In entrepreneurship, Ambos and Birkinshaw (Reference Ambos and Birkinshaw2010) identified three distinct archetypes of ventures to study new venture evolution. However, little work has explored how salient individual traits might combine to uncover archetypes of founders, particularly with respect to new product development.

To address this gap, we are interested in investigating how self-efficacy, bricolage, risk propensity and psychological safety might combine to affect resource use. These traits represent how an individual perceives their team environment, their own abilities, their resources and their behaviors. Psychological safety fosters an environment where individuals feel safe taking interpersonal risks, which may enhance their willingness to engage in innovative behaviors. This environment may positively influence ESE, where individuals with higher self-efficacy are more confident in themselves. Entrepreneurial bricolage compliments this relationship by enabling individuals to be resourceful and innovative in resource-scarce environments. Finally, individuals with a greater risk propensity by definition are more likely to take action in uncertain situations. Thus, these constructs may combine to uncover distinct patterns of behavior.

3. Research objectives

In this study, we followed an exploratory QUANT $ \to $ explanatory qual approach where qualitative data were collected and analyzed to further explain the quantitative results (Giddings and Grant, Reference Giddings and Grant2006). This work aimed to explore how individual traits of founders affect their resource management abilities. We argue that technology-based startup founders represent a unique subset of new product developers that can provide significant insights into design behaviors. To explore these topics, two research questions drive this work:

Prior work has shown that individual traits of founders such as self-efficacy, bricolage, risk propensity and psychological safety are important factors of team and startup success. In pursuit of our first research question, we investigate if archetypes of participants emerged based on these individual traits. In this work, we refer to archetypes as distinct groups of founders that are characterized by shared traits, behaviors and decision-making patterns. These archetypes represent generalized profiles that capture similar traits within each group while highlighting differences from other archetypes, which could help to identify common behaviors and challenges faced by different types of founders. We first hypothesize that unique archetypes will emerge, with participants in each archetype exhibiting similar characteristics to one another. Next, we further investigate if the archetypes that emerge based on individual traits will exhibit differences in firm-level characteristics such as the stage of development and firm performance. Here, we hypothesize that based on archetype, there will be differences in how developed the firms are as well as their financial performance. An important distinction in this work is that the results from RQ1 were used to develop RQ2 to further investigate the nuances between the archetypes that were identified in RQ1.

Resource accuracy represents an individual’s ability to identify relevant resources to mitigate current obstacles. It is important to note here that resource accuracy differs from resource acquisition capability which represents an individual’s ability to obtain the resources necessary for firm survival (Pulka et al., Reference Pulka, Ramli and Bakar2018). In this work, we are interested specifically in individuals’ ability to identify necessary, or relevant, resources. We added the qualifier “relevant” to resources to make an important distinction: not all resources are valuable or useful, but they depend on the obstacles a participant faces. For example, if a participant is struggling with obtaining funding to hire another team member, a physical resource such as new laboratory equipment is not relevant to this specific obstacle. However, new laboratory equipment is a relevant resource to a participant struggling with broken equipment. As such, we are specifically interested in these relevant resources, or the resources that directly mitigate obstacles the team currently faces. We hypothesize that participants with higher levels of these traits will exhibit greater resource accuracy.

5. Analysis

To first understand if distinct archetypes exist, a k-means cluster analysis was conducted using R CRAN v. 4.2.0. A k-means cluster analysis is an unsupervised machine learning algorithm that partitions a dataset into distinct groups, or clusters, of participants with similar characteristics (Hair et al., Reference Hair, Anderson, Tatham and Black1984). The algorithm computes the number of clusters and the cluster size by iteratively minimizing the within-cluster sum of squares, which is a significant strength of the method. Emergent clusters should exhibit high within-group homogeneity and high between-group heterogeneity (Morissette and Chartier, Reference Morissette and Chartier2013). More generally, this method measures similarity between participants as the Euclidean distance, or the distance between vectors representing each participant’s traits. Ultimately, k-means clustering creates a smaller number of clusters with meaningful groupings of participants and this method was chosen because it is a simple, yet robust method for exploring patterns and analyzing high-dimensional data sets. In this work, we were interested in investigating if distinct archetypes of participants exist based on salient entrepreneurial traits. In our analysis, the k-means clustering method groups participants based on if they exhibit the following traits: ESE, entrepreneurial bricolage, risk propensity and perceptions of psychological safety. This method allows us to investigate how these individual factors interact, or hang together, and if these interactions result in the emergence of distinct archetypes of participants. Finally, the specific characteristics of each archetype can be analyzed by comparing these traits across the resulting clusters.

K-means clustering was performed iteratively with 25 random initializations until convergence was achieved with a maximum of 10 iterations. The algorithm varied the number of clusters to determine the best number of clusters to represent the true data groupings. Each iteration used a different number of clusters and computed the mean point-to-centroid distance, and any iterations resulting in a single-member cluster were to be omitted, although this was not observed in these data. We determined that our sample size, n = 241, was adequate for this analysis based on prior work that suggests a minimum sample size of 80 based on the number of clustering variables in this study (Dolnicar and Grün, Reference Dolnicar and Grün2008). Using the silhouette method, which plots the average silhouette width for different numbers of clusters, k = 2 clusters minimized the point-to-point centroid distance, yielding two clusters, or archetypes, of similar size with a discernible rationale behind the clustering based on visual inspection and statistical comparisons of the responses. The elbow method, which plots the within-cluster sum of squares against the number of clusters, was also implemented to validate that having k = 2 clusters was optimal. Thus, two archetypes were identified for further analysis and comparison: archetype 1 (n = 134) and archetype 2 (n = 107).

To investigate participants’ resource accuracy, we qualitatively analyzed the thirty-two semistructured interview transcripts. Deductive coding was employed, using a previously developed coding schema by Marcon and Ribeiro (Reference Marcon and Ribeiro2021) that categorizes firm-critical obstacles and resources into the following categories: financial, human, social, organizational, physical and innovation. The codebook we used to identify obstacles, as shown in Table 1, contains each category of obstacles, along with example statements for each code from our data set. The codebook we used to identify resources, as shown in Table 2, contains each category of resources, along with example statements for each code from our data set. Throughout the coding process, the appropriateness of the coding schema was discussed and the process itself was reviewed. The interview data were coded to agreement by two coders.

Table 2. Codebook for resources (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023)

Figures 4 and 5 show examples of how obstacles and resources that were identified for participants A and B, respectively. Red highlight indicates an identified obstacle and green highlight indicates an identified resource. For the purpose of this example, excerpts from the interview were sliced together and extraneous information was omitted.

Figure 4. Example of participant A’s coded interview (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023).

Figure 5. Example of participant B’s coded interview (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023).

Participants’ resource accuracy was quantified by summing the weighted frequency of orphaned obstacles, obstacle and resource matches, and extra resources, based on prior work (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023). Here, orphaned obstacles were given a weight of −1, obstacle and resource matches were given a weight of +1, and extra resources were given a weight of +0.5. A weight of −1 was chosen for orphaned obstacles because having an obstacle with no resources to help mitigate it negatively impacts the team. However, when a resource is identified to mitigate an obstacle, this negative impact is turned positive. Thus, a score of +1 was chosen for an obstacle and resource match. Extra resources were weighted +0.5 because while they are helpful, they do not completely mitigate the burden of an unsolved obstacle. The total score for each participant was calculated using the following equation:

(1) $$ ResourceAccuracy=-O+M+0.5R, $$

where O is the number of orphaned obstacles, M is the number of obstacle and resource matches, and R is the number of extra resources. The excerpts shown in Figures 4 and 5 were analyzed as shown in Tables 3 and 4 and received scores of +2.5 and −2.5, respectively.

Table 3. Example of participant A’s obstacle and resource identification (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023)

Table 4. Example of participant B’s obstacle and resource identification (Letting et al., Reference Letting, Calpin, Soria Zurita and Menold2023)

8. Limitations and future work

Although we found significant results in this study, as with any work, there are limitations that should be considered when interpreting our findings. One limitation of the current study is that our participants were recruited exclusively from the NSF I-Corps program. Because of this recruitment strategy, there is the possibility that self-selection bias may be affecting the results. For example, participants from the NSF I-Corps program may represent a different subset of founders than those who have not participated because they may have more time to participate in these training programs or they may now have more resources available to them. The participants in this study were also geographically clustered in the Mid-Atlantic region of the United States, and prior work has found that geographic location can affect cultural norms and resource availability (Peng et al., Reference Peng, Menold and Miller2022). As a result of these limitations, we acknowledge that the findings from this study may not be generalizable to broader geographic locations or founders that have not participated in the NSF I-Corps program. Future work will focus on recruiting participants across the United States both within and outside of the NSF I-Corps program to compare these findings.

Another limitation of the current work is a reliance on self-report data through surveys. While a commonly employed data collection technique across human-subjects research (Blösch-Paidosh and Shea, Reference Blösch-Paidosh and Shea2022; Singh et al., Reference Singh, Cascini and McComb2022; Nolte et al., Reference Nolte, Zurita, Starkey and McComb2023; Budinoff et al., Reference Budinoff, McMains and Shonkwiler2024), self-report surveys may fall victim to certain biases such as social desirability bias. As such, in future work, research should attempt to triangulate self-report data with secondary ratings or other avenues of empirical evidence which may provide a more complete picture of phenomenon of interest. Our survey also included questionnaires on multiple constructs that were chosen based on prior literature. While these were carefully chosen, their length may have induced survey fatigue in our participants and may have affected how participants responded to the questions that were asked. It is also difficult to determine if participants responded truthfully to the survey questions. We do note, however, that the survey responses were reviewed before analysis and we concluded that to the best of our knowledge, the responses seemed to be completed truthfully and thoughtfully. It should also be noted that individual traits and characteristics can change significantly over time, particularly during the early stages of a startup. Thus, it is possible that an individual’s archetype may change and that the results in this study represent an instantaneous measurement of each participant’s self-reported traits. Future work will focus on collecting more longitudinal data to determine how these traits and archetypes might evolve with time.

Regarding the analytical methods used in this work, k-means clustering was employed to explore patterns of behaviors and archetypes within the data without imposing assumptions about predictive relationships. As an unsupervised method of machine learning, k-means clustering allowed us to investigate naturally occurring patterns in these traits without relying on explanatory or predictive models. However, we recognize that this approach has limitations in terms of establishing causal or predictive relationships. Thus, a valuable area for future work could involve supervised methods to test predictive or explanatory hypotheses. Such analyses could assess specific relationships between identified archetypes and performance-based metrics such as resource accuracy or financial performance.

This work studied startup founders that are in the early stages of the new product development process, and as a result, the financial performance data may be affected. A significant area of future work will focus on studying these firms longitudinally as they become more developed. By doing so, we will be able to track the firm over time and gain important insights into their financial performance. Studying the success of these teams longitudinally will provide meaningful insights to the field of design by uncovering the complex relationship between individual traits of founders and startup success.