The role of human and financial capital in the business model design-performance relationship: evidence from Austrian start-ups

Previous empirical work investigated the role of business model design in relation to performance, both in entrepreneurial (Zott & Amit, 2007) and general settings (e.g. Aspara et al., 2010; Brettel et al., 2012; Hu & Chen, 2016; Wei et al., 2014). Additionally, business model designs were tested jointly with other factors originating from the resource-based view, such as social capital (Spiegel et al., 2016), marketing (Brettel et al., 2012), innovation (Ma et al., 2018; Wei et al., 2014), or others such as firm size (Aspara et al., 2010), firm age or external environment (Pati et al., 2018; Zott & Amit, 2007).

Numerous studies (e.g. Gerdoçi et al., 2018; Guo et al., 2018; Ma et al., 2018; Pati et al., 2018) built on the business model taxonomy provided by Amit and Zott (2001), who differentiated four potential sources of value creation present, namely efficiency, complementarities, lock-in, and novelty. The efficiency-centred business model design concerns the measures that firms may adopt to realize transaction efficiency by means of their business models, thereby aiming at reducing transaction costs for all transaction participants. This business model reduces the uncertainty or information asymmetry among participants through increased transparency and allows for scaling. Further, differentiating levels of entrepreneurial orientation, Wiklund and Shepherd (2005) see low novelty levels associated with a focus on the marketing of tried-and-true products or services. Distinguishing between innovator and replicator firms, Aspara et al. (2010, p. 43) state that once having discovered and refined a new business model, a replicator firm may create further value by learning about and refining its business model and by maintaining the model in operation once it has been replicated.

In contrast to efficiency-centred business models, the novelty-centred business model design enables new ways to realize economic exchanges among various participants by, for example, linking transaction participants in innovative ways, bringing together previously unconnected parties, or by offering novel incentives to participants (Zott & Amit, 2007). Novelty-centred firms continuously engage in business model innovation, which is the ‘creation of novel value by challenging existing industry-specific business models, roles, and relations in certain geographic market areas’, providing ‘entirely new value to certain people and/or organizations’ customers’ (Aspara et al., 2010, p. 47). Wiklund and Shepherd (2005) see high entrepreneurial orientation associated with a strong emphasis on R&D, technological leadership, and innovations. Novelty-centred firms frequently act as pioneers with their business models, and following Teece (2010), their business models are difficult to imitate. Additionally, they often rely on trade secrets or copyrights to sustain their competitive advantage.

Previous research suggests that both models can exert a positive influence on firm performance. The two designs are neither exhaustive nor orthogonal (for example, novelty-based business models may engender reduced transaction costs), nor are they mutually exclusive (Zott & Amit, 2007). Based on these arguments, we argue that both business models could have a positive impact on the performance of start-ups in a developing start-up ecosystem, such as in the case of Austria:

The study is based on a survey among new ventures in Austria, which was complemented by collecting data about the participating firms by using company material and information. There are no public data available in Austria regarding newly established highly innovative start-ups with growth ambitions. This gap has been addressed by a publicly funded database in Austria (Austrian Startup Monitor), which was initiated in 2017 and aims to identify all newly founded ventures in Austria established since 2008, i.e. not older than 10 years (Leitner et al., 2020). The innovation content, for example in the field of technology, product, service or business model, and the growth potential were considered as essential characteristics of start-up companies (e.g. Saemundsson & Candi, 2017). These factors differentiate start-up companies from traditional company foundations, which are not characterized by the same growth orientation and the need for external equity capital financing. Thus, in a first step, new innovative ventures that were founded between 2008 and 2018 were identified by using various sources, such as information about the location of the new venture (e.g. located in a business incubator), reports in the media about new ventures, or data about private or public funding of new ventures. In addition, information about new ventures from two existing databases that cover Austria, i.e. Crunchbase and Dealroom, were used to identify start-ups. A total of 1,500 start-ups were found in the different sectors, which also confirms the strong growth rate of innovative start-ups since 2008. The companies were invited to participate in an online survey in 2019 (Leitner et al., 2020). The online survey encompassed basic questions about the new ventures such as age, demographics of the founding team, sources of funding, markets served and performance. Using various measures, such as sending out targeted reminders and short phone calls, the research team aimed to achieve a high response rate and a balanced sample concerning the location, sector and age of the company. In total, 194 companies participated in the survey. We checked for a possible non-response bias by comparing the sample with the total population, taking into account the location, the sector and age, which revealed no significant differences between either groups, thus delivering no evidence for a bias.

As described above, in order to operationalize the business model variable, we built on the methodology of Zott and Amit (2007), who identified two latent variables that characterize the design theme of a business model (efficiency and novelty). During December 2019 and March 2020, one of the authors evaluated the business models of these start-ups (Ruthensteiner, 2020). During that period, the author took one measurement of both design themes (efficiency and novelty) respectively for each of the 194 firms in our sample, collecting cross-sectional data on our independent variables. On average, it took the author about three hours to collect data on a given business model, to analyse the model and to complete the survey. Sources of data included websites, newspaper articles, annual reports as well as data gathered within the scope of the survey (for a similar methodological approach see Zott & Amit, 2007). For our estimations, we removed eight firms with a firm age of one year as well as four outliers from the sample, resulting in a final sample size of 182 start-ups. To cross-validate the assessment of the business models, an independent researcher not involved in the research project reviewed the results of the assessment resulting in a final assessment of the business model. By using both information from the survey and data collected via secondary sources we also aimed to reduce a possible common methods bias.

In order to operationalize the independent variable business model design (‘BM design’), we relied on a measurement scale originally developed by Zott and Amit (2007), which we enriched with additional items based on our literature review on business model taxonomies (Aspara et al., 2010; Teece, 2010; Wiklund & Shepherd, 2005). For the efficiency-centred BM design, after drawing on five items proposed by Zott and Amit (2007) we enriched the scale with two additional items suggested by Wiklund and Shepherd (2005) and Aspara et al. (2010). For the novelty-centred BM design, after relying on seven items suggested by Zott and Amit (2007), we added two additional items that were applied by Wiklund and Shepherd (2005) and Teece (2010).

Given the difficulty of obtaining objective measures of BM design, we deemed it appropriate to use perceptual measures obtained from one of the authors (Covin & Slevin, 1988; Dess & Robinson, 1984). The strength of each of these items in a given BM design was measured using Likert-type scales (items ranked 0 – ‘I do not agree’ to 4 – ‘I totally agree’) and coded into a standardized score by aggregating all item scores for each design theme into an overall score for the composite scale using equal weights (Mendelson, 2000). This process yielded quantitative measures of the extent to which each business model in the sample leveraged efficiency and novelty as design themes.

Since we enriched the scales with additional items, it is reasonable to perform convergent and discriminant validity, reliability and overall model fit checks (Brettel et al., 2012). We thus performed exploratory factor analysis to test for convergent validity in a first step. Generally, the thresholds for the loadings should exceed the acceptable standards of 0.32 (Tabachnick et al., 2007). For the efficiency-centred BM design, we deleted two items that did not have a reasonably high loading (Routinize_knowledge, Replication), whereas we deleted four items for novelty-centred BM design with low loadings (RD_technology_innovation, Trade_secrets_copyrights, Difficult_imitate, Pioneer_business_model). By performing these adjustments and eliminating a total of six items (2 for efficiency-centred BM design, 4 for novelty-centred BM design), the measurements are acceptable in terms of convergent validity. The particular items are reflected in Appendix Table 6. Re-estimating the model without the outlined items, all remaining items load above 0.60 and are significant at p⩽0.05.

We assessed discriminant validity by analysing the square roots of the average variance extracted (AVE) (Fornell & Larcker, 1981). The square roots of the AVE values for each construct are greater (0.81 & 0.73, respectively, for BM efficiency and BM novelty) than the correlation (0.33) between both factors, thus proving satisfactory discriminant validity (Fornell & Larcker, 1981; Hu & Chen, 2016).

Internal consistency as a form of reliability of our measures was validated using standardized Cronbach alpha coefficients, yielding 0.91 for the BM efficiency measure and 0.94 for the BM novelty measure, respectively. Our measures satisfy Nunnally’s (1978) guidelines, which suggest 0.7 as a benchmark for internal consistency. The results of a confirmatory factor analysis applying the maximum likelihood procedure to assess overall model fit indicated that the measurement model moderately fits the data (χ2 = 200.52; df = 53; χ2/df = 3.78; CFI = 0.93; TLI = 0.91; RMSEA = 0.12; SRMR = 0.07) and thereby confirmed the unidimensionality of each construct (Gerbing & Anderson, 1988). Table 1 below provides an overview of the characteristics of both factors, while Table 6 in Appendix provides a more detailed picture on the survey, including final item loadings of each factor.

Two moderating variables were used in the study. Human capital was operationalized in the form of start-up experience as a binary variable. Previous start-up experience describes whether at least one of the company’s founders had previous start-up experience; it was coded as a binary variable and defined as 1 if a member had previous experience and 0 otherwise (Patzelt et al., 2008). Financial capital was also measured as a binary variable. A critical threshold within our sample concerns the amount of funding received. Considering the specifics of the Austrian market with its rather small number of substantial investments, we defined a threshold of 300,000 euros acquired external capital, which accounted for about 40% from our set of start-ups that received more than or equal to EUR 300,000. This level is considered a milestone in a company’s development and proof that it has convinced a major investor or several investors (e.g. business angels) in the Austrian investment landscape (AAIA et al., 2023). According to the Austrian Investing Report 2022 (AAIA et al., 2023), the majority of business angels invest less than EUR 300,000 in a typical investment; going beyond this level can be seen as a first success in a company’s development. We coded the variable as 1 if the venture was above this threshold and 0 if it was below.

For our dependent variable, we followed previous studies and used two performance measures related to sales and employment growth (e.g. Buergel et al., 2000; Murphy et al., 1996). Sales was measured on the basis of figures provided by the respondents, who indicated the volume of sales realized within the past year. Due to the skewed distribution of this variable, it was decided to normalize the absolute sales by converting them to a common log scale (e.g. Westhead & Cowling, 1995). We then divided that amount by the firm’s age to create a proxy for average yearly sales growth (Buergel et al., 2000). Due to the different firm ages of the sample firms, employment growth was measured based on average yearly growth rates. The growth rate of the performance variable (sales) between two time intervals was calculated as the difference at these two points in time after taking logs, and then the difference divided by the length of the interval, in our case this being the firm’s age (Buergel et al., 2000).

Based on our literature review we suggest that additional factors such as the size of the founding team, firm age and industry sector affiliation be included as control variables. Initial constraints in terms of human capital as well as financing are considered to be probably less pronounced if the start-up is founded by more individuals. We therefore include the number of founders as control variable (Buergel et al., 2000). Moreover, we integrated two sectorial dummy variables to differentiate between firms that are primarily attributed to the IT and software development sector or the life science sector (Brettel et al., 2012).

We applied ordinary least squares (OLS) regressions with the two dependent variables sales growth and employment growth. We tested the robustness and validity of both model specifications in several distinct ways.

First, in order to allow for a meaningful comparison of the variables measured along different scales and to reduce potential collinearity, we standardized all independent variables before entering them into the regression models (see Rothaermel & Alexandre, 2009). Second, to account for the skewed distribution of our employee performance measure as well as firm age, we performed a logarithmic transformation of both variables. Third, we assessed the threat of multicollinearity by estimating variance inflation factors (VIFs) across all regression models and found that the average VIFs for all variables and interaction terms were well below the recommended ceiling of 10 (Cohen et al., 2003). In a fourth step, we checked for linearity. The R function ‘Residualplots’, plots the residuals versus each predictor and the fitted values and computes the lack-of-fit tests for the individual variables, which indicated that linearity is given in both model specifications. Fifth, performing a Shapiro–Wilk test we found that non-normality was not an issue in either model. Sixth, being concerned with heteroscedasticity in a cross-sectional study (Bowen & Wiersema, 1999), we graphically checked for heteroscedasticity within both model specifications, revealing that heteroscedasticity is not present when estimating through a standard OLS model. Seventh, we checked for potential outliers, influential points and leverage values, which did not raise any concerns. Eighth, we controlled for the risk that external financing might be endogenous to our performance variables. We thus tested for potential endogeneity (i.e. the concern that external financing could be correlated with the error term) by running 2-stage least squares regressions with instrumental variables and by using the Hausmann test, which indicated that the OLS regressions are more appropriate (see Greene, 2003). Overall, therefore, we concluded that our model specifications proved robust and valid.

Table 2 provides an overview of our data, revealing the entrepreneurial nature of our sample of start-ups. Sample firms are characterized by a higher median business model efficiency level (3.2) than a median novelty level (2.2). In 55% of the sample companies, the start-ups’ founders did not have any previous start-up experience. Less than half (43%) of the companies had previously received external financing exceeding EUR 300,000. In 2019, the median age of a sample start-up amounted to four years, the median number of founders was three. Whereas 38% of the sample could be attributed to the IT and software development sector, 6% of the start-ups were classified as belonging to the life science sector. The median level of average sales per year (log) amounted to 3.08 out of a possible range of 0.92 to 8.34, the median level of the number of employees was 6.5, and the growth rate calculated as described above accounts for 0.5. Table 3 below presents the zero-order correlations among all variables used in the regression analyses. While some correlations among the explanatory variables are significant, they do not pose a multicollinearity problem as their variance inflation factors (VIFs) indicate a low value (Kleinbaum et al., 1998).

Hierarchical multilinear regression was performed to test the stated hypotheses, which is a suitable approach when assessing multiplicative terms in regression analysis (Cohen et al., 2003). Table 4 reports the results of the hierarchical regression analyses for the dependent variable sales growth, Table 5 presents the results for the dependent variable growth of employees.

To test the hypotheses, we first added the control variables (Model 1 and Model 5), then the independent variables (Model 2 and Model 6), then the respective two-way interaction terms (Model 3 and Model 7), and finally the three-way interaction terms (Model 4 and Model 8) each start-up performance variable, respectively. Among the control variables, the age of the firm has, as expected, a strong impact on the growth of the firm. Start-ups in the life science sector show a slower growth rate compared to other firms. In addition, the number of founders has a positive impact on the employment growth.

Looking at the BM novelty measure, the variable is positively and significantly related to both performance measures (Model 2: β = 0.08, p⩽0.1, Model 6: β = 0.16 p⩽0.05) thus indicating support for Hypothesis 1a. With regard to our BM efficiency measure, we find a slightly negative but insignificant effect. We therefore reject Hypothesis 1a and confirm Hypothesis 1b.

Model 3 and Model 7 were created by adding interaction terms. A positive and significant moderating role of previous founding experience could be found for novelty-centred BM design in both model specifications (Model 3: β = 0.17, p⩽0.1, Model 7: β = 0.40, p⩽0.01), with the same effect being insignificant for start-ups with an efficiency-centred BM design for both performance measures, thus indicating support for Hypothesis 2.

Within the sales model, we find a positive but insignificant interaction effect between external capital and efficiency-centred BM design, with the same effect being positive within the employment model specification. However, the interaction term between external financing and BM novelty is even negative (not significant) in the employment model. Thus, we have to reject Hypothesis 3.

Next, we included the three-way interaction terms in Model 4 and 8. Contrary to our expectations, a significant negative moderating role of both previous start-up experience and external capital could be found for novelty-centred BM design (Model 4: β=−0.31, p⩽0.1, Model 8: β=−0.69, p⩽0.05). The corresponding three-way effects for efficiency-centred BM design were negative but statistically insignificant in both model specifications. Hypotheses 4 is thus not supported. To examine this relationship in more detail, we conducted a post hoc analysis to determine the extent to which firm age might have an influence. We split the sample based on median firm age (= 4 years) and ran two separate models with the employment growth variable (results not disclosed here). These models show that the three-way interaction term is stronger for the group of younger start-ups and significant only there. Thus, there is no evidence that older start-ups that follow a novelty-oriented BM have no start-up experience and have raised little or no external equity capital grow significantly slower.

In order to better understand the significant interactions, we followed the methodology suggested by Aiken et al. (1991) and plotted the significant effects on an x-axis of independent variables and a y-axis of performance for the employment-related performance measure based on the unstandardized regression coefficients (see Figs. 2 and 3). For higher-order interactions, all lower-order interactions and main effects were taken into consideration. Values of all predictor variables were set at one standard deviation above and below the mean.

Looking at the two-way effects, previous start-up experience strengthens the performance of start-ups with a novelty-centred BM design. The three-way effects (Fig. 3) show that the impact of a business model design varies in line with the underlying resource situation. The nature of the interaction indicates that start-ups with a low degree of BM novelty perform less well regardless of start-up experience and the amount of external financial capital. However, the picture looks different in the case of a high degree of BM novelty. In the case of a novelty-centred BM, having no start-up experience and little or no external funding is the worst possible combination. Start-ups run by founders with previous start-up experience and less external financing yield the relatively best results.

There are several limitations to this study which might serve as a starting point for further research. First, empirical results could be affected by measurement problems regarding our latent variable ‘BM design’. For example, some firms may pursue several business models simultaneously. We might not have captured all underlying business models of a start-up’s business, and therefore might not explain all the variation in the dependent variable due to business model design themes. Future researchers may hence consider capturing the impact of operating several business model designs simultaneously. A second limitation relates to start-ups undergoing business model changes. Based on the data, it is not possible to derive how long and how consistently the start-up had pursued a certain business model. Start-ups are known to undergo frequent business model changes, thus the potentially important influence of a change of business model design on performance was not taken into account – nor the potential delay between adopting/maintaining a certain business model and its outcomes. Future research may thus consider assessing performance implications for firms that have undergone or are envisioning business model changes. In the context of a longitudinal study, we could also take into account the fact that we only studied surviving companies. A third limitation relates to external capital. No differentiation was made between the different types of investors when an investment was provided for a start-up. Next to the provision of funds, venture capital firms and angel investors, for example, usually support start-ups they invest in from a strategic point of view. Their investment might be more beneficial than governmental subsidies, which do not go hand in hand with a coaching function. Future research could differentiate among the different types of external capital with more granularity. A final main limitation relates to the generalizability of the findings of this exploratory study. The sample was drawn from a set of start-ups in Austria, which has a small but growing start-up scene compared to other European countries (Ernst & Young, 2019; Friedl et al., 2019). The number of start-ups and the investment volume have increased strongly in the observation period between 2008 and 2018, which is also characterized by rising macroeconomic indicators such as productivity and economic growth (AAIA et al., 2023). This specific context needs to be taken into account in the interpretation of the results. We thus cannot assume complete generalization in the context of other countries or established firms (Wu et al., 2024). Future research may consider replicating this study in other countries with a more mature financial capital market and a more established entrepreneurial culture. In addition, studies that examine whether the results can be replicated for established firms are also of interest. This paper opens up other avenues for future research. For example, other elements such as social capital within the resource-based view that could potentially moderate the business model–performance relationship could be explored. Moreover, it would also be of interest to study hybrid business models, and the effect of different business models over the life course (e.g. Leppanen et al., 2023; Spieth et al., 2025). The general economic environment, an economic crisis and high interest rates may also affect the relationship between resource bundles and the business model, which should be considered in future studies.

The findings of our study suggest that the understanding of start-up performance can be strengthened by assessing the business model jointly with its underlying resource base. While previous studies have examined the direct effect of human and financial capital, we also consider the role of the business model that can leverage these resources, broadening our understanding of the strategic development of entrepreneurial firms.

Specifically, our results indicate that differences with regard to performance implications in the case of start-ups with a novelty-centred business model are better explained when considering the previous founding experience of the company’s founders and the provision of external financing. Although our findings provide a first step, the understanding of these relationships needs to be broadened, since the business model is of strategic importance to start-up performance. A better comprehension of how business models are interlinked with their underlying resource base in terms of performance will benefit both academia and practitioners.

With regard to managerial implications, our research generally confirms the necessity of paying greater attention to both joint and interdependent effects of performance predictors rather than relying solely on main effects. The results demonstrate that the underlying resource structure should be explicitly considered when designing a start-up’s business model, especially when it leverages a novelty-centred design theme. This is relevant for entrepreneurs, but also for investors and private and public financing institutions when evaluating a business model. Business model-specific elements such as resources may account for some hitherto unexplained variance in the performance of firms.