Organisational knowledge and innovation failures: the moderating effect of knowledge spillovers Open Access

The notion that individuals acquire knowledge through experience – whether it results in success or failure – can be traced back at least to Adam Smith (Ferreira et al., 2020). Building on this premise, Garcia Martinez et al. (2019) suggest that learning emerges from a process of trial and error, driven by repeated efforts to address problems as they arise. Within this framework, organisational knowledge (OK) is seen as a dynamic process that unfolds during the development of techno-economic activities. These activities often entail failure, which, as an experiential outcome, can expose problems that activate learning mechanisms (Leoncini, 2016; Costa et al., 2023).

Organisational theory has long recognised learning as a critical factor in explaining performance disparities among firms (Chesbrough, 2010; Jha and Basu, 2025), largely due to shifts in how knowledge is accumulated through experience (Zouaghi et al., 2018). For companies pursuing innovation, the ability to generate, sustain and advance internal knowledge is fundamental. Learning plays a pivotal role in reshaping existing practices by analysing, refining, adapting and restructuring organisational routines and procedures (Leoncini, 2016; Jung et al., 2024).

Nevertheless, failure has become an increasingly common element in the corporate landscape (Danneels and Vestal, 2020; Dabić et al., 2021) and is gaining traction in academic discourse (Khanna et al., 2016; Forsman, 2021). If failure is defined as any action that does not yield the expected outcome, then even high-performing firms encounter failures throughout their operations (Amankwah-Amoah and Wang, 2019).

Despite its growing relevance, research offering insights into the potential benefits of failed innovative projects – such as increased likelihood of future success – remains scarce. Rhaiem and Amara (2021) identify a notable gap in the literature regarding the development of comprehensive theoretical models that explain learning from failure. Surprisingly, most existing studies on innovation failure do not explicitly engage with theoretical frameworks. Only a limited number of works – such as the behavioural approach by Deichmann and Ende (2014), the social learning theory explored by Kim and Miner (2007), the contributions by Magazzini et al. (2012) and the cognitive learning theory advanced by Shepherd and Cardon (2009) – represent exceptions. These studies address the inherent uncertainty and risk of the innovation process (Metcalfe, 1995) or offer responses to systemic innovation failures (Metcalfe, 2005). More recently, Stojcic (2024) has examined how training for innovation and public support – both national and EU-level – differentially impact established and emerging innovators.

To date, much of the literature has focused on empirical efforts to refine concepts and develop measurement tools. Rhaiem and Amara (2021) emphasised the centrality of knowledge in the learning process and advocate for placing it at the core of future research agendas. They propose that future studies should invest in knowledge-based theoretical frameworks that stress the importance of developing and retaining knowledge within organisational settings.

In 2020, Ferreira and colleagues conducted a study to investigate how failures affect innovation within firms (Ferreira et al., 2020). They found that there is a negative correlation between innovation failure and firms’ experience level and their access to external knowledge. Although the study accounted for knowledge spillovers enabled by external knowledge access, it did not analyse their impact on failure occurrence. Upon assessing failed innovations, it became evident that an organisation’s accumulated knowledge undergoes a transformative process influenced by its past experiences. These insights inspire our research question:

Consequently, our objective is to meticulously examine the influence of KS on the relationship between organisational knowledge and innovation failures, and on the relationship between failed innovations and the organisation’s future innovation capability.

Supported by the arguments presented, we find that it is not only the OK that can negatively impact innovation failures, nor is it only the eventual tolerance of failure that improves innovation, but rather the relationship with KS, which will impact future innovative capacity. Thus, our research does not support the sensationalist conclusion that success in future innovations following failures comes exclusively from tolerating these failures. On the contrary, we can state that the future innovative capacity of a firm after failure in innovative projects stems from its relationship with KS, i.e. the access to knowledge that enables it to systematically and deliberately analyse what went wrong to improve in the future. Our results show that regardless of the size and intensity of innovation failures, KS provide the safety net for absorbing and understanding the knowledge needed to decide to innovate again (re-innovating).

Our research makes several contributions. Firstly, it responds to calls from previous studies (e.g. Rhaiem and Amara, 2021) for more research on innovation failures that apply a knowledge-based view (KBV). Our research directly examines two mechanisms by which failure may diminish innovation failures and the mechanisms themselves, and how they do not impede future innovation: OK and KS.

Secondly, our research shows that the OK has a positive impact on companies’ innovative success, reducing innovation failures. Analysing past failures in detail as part of the ongoing effort to extract lessons from them can often be a Herculean task.

Thirdly, our research focuses on the knowledge the firm can gain, examining the effect of KS. Research on learning from innovation failure has primarily centred on learning from “post innovation failure”, leaving the processing of the conclusions drawn from that failure in a black box. As Ferreira et al. (2020) argue, the other side of learning is precisely the reflection, synthesis, abstraction and articulation of the knowledge drawn from the experience that failed. While these considerations are acknowledged, previous research has not studied the impact of KS on future innovative capacity.

Grant (1996) brought us the KBV as an extension of the resource-based view. The link between KBV and RBV arises from the fact that KBV, on the one hand, interprets knowledge as a resource (Ferreira et al., 2023) and, on the other hand, like RBV, considers organisations to be heterogeneous entities that possess knowledge (Hoskisson et al., 1999). The basic principle of the RBV is based on the unique characteristics of resources (valuable, rare, inimitable and organisation) and is fundamental to being able to explain the strategic nature of an organisation’s resources and capabilities (Priem and Butler, 2001) and provide competitive advantages to the firm (Wernerfelt, 1995; Fernandes et al., 2022).

KBV is also consistent with the approach that argues that organisations are cultural artefacts (Balogun and Jenkins, 2003) that learn through various activities and can adapt over time. We thus arrive at organisational learning, which enables the firm to acquire, develop and sustain its organisational and knowledge capabilities (Martínez-Sanchez et al., 2020). Given that the organisation’s resources increasingly consist of knowledge-based assets, these assets increasingly assume an important role in ensuring that a competitive advantage is sustained, precisely because they are difficult to imitate, thereby providing the basis for sustained differentiation (Pereira and Bamel, 2021). Therefore, if knowledge is the key strategic resource that enables companies to compete in a dynamic environment, managers must value knowledge and simultaneously create and sustain practices for sharing it, both internally and externally (Durst et al., 2019).

Thus, KBV offers arguments for companies to consider knowledge a strategic resource (Pereira and Bamel, 2021). Thus, two issues must be considered: the temporality or opportunity to access knowledge and the mechanisms of access to this knowledge (Fernandes et al., 2022). Knowledge absorptive capacity (Martínez-Sanchez et al., 2020), knowledge transfer (Su et al., 2020) and knowledge reuse (Lee et al., 2021) are important elements in understanding firms’ competitive advantage. Consequently, the efficiency and effectiveness of knowledge transfer should also be considered (Xie and Wang, 2020). Grant and Baden-Fuller (2004) show that organisations use various instruments to develop knowledge internally or to acquire/access knowledge externally; strategic alliances are one such instrument. By leveraging human resources knowledge, KBV offers companies strategies not only to achieve better organisational results but, above all, to gain a competitive advantage.

In an economic context, where intangible assets are increasingly valued and considered crucial intellectual capital assets, KBV assumes increased importance (Grant, 2002; Mathews, 2003; Fernandes et al., 2022) since knowledge enables the company to achieve the desired organisational results (Grant, 1996; Ferreira et al., 2023).

OK is increasingly recognised as a vital asset for firms navigating volatile, uncertain and highly competitive environments. At its core, OK involves processes of transferring, interpreting and assimilating knowledge, which lead to behavioural and strategic changes necessary for enhancing organisational performance (Zagoršek et al., 2009; Vashdi et al., 2019). These changes are not just operational adjustments but also involve deeper shifts in routines, cognitive frameworks and decision-making approaches that enable firms to respond more effectively to environmental shifts.

A substantial body of literature identifies OK as closely related to the concept of absorptive capacity – the firm’s ability to identify, assimilate, transform and exploit external knowledge – which has been widely studied since the foundational work of Cohen and Levinthal (1990) and later expanded by Zahra and George (2002). This capacity plays a critical role in enabling firms to prevent innovation failures by equipping them with the tools to process information, integrate it into internal operations and apply it to new contexts (Garcia Martinez et al., 2017). Through the development of OK, firms are better positioned to build and sustain dynamic capabilities, particularly those that enable them to detect, interpret and assess knowledge from diverse sources – such as market trends, technological developments, customer feedback or competitors’ actions (Garcia Martinez et al., 2019).

Researchers have demonstrated that enhancing OK significantly reduces the risks associated with innovation and, by extension, lowers the likelihood of entrepreneurial failure (Cefis and Marsili, 2012; Howell, 2015). Firms with well-developed OK are more resilient, as they can learn not only from their successes but also from past failures, adjusting strategies and reallocating resources accordingly. This learning capacity contributes to long-term sustainability, improving both survival rates and overall business performance (Li et al., 2010; Ugur et al., 2016).

In this context, OK can be regarded as a firm-specific strategic capability – an intangible resource that supports innovation, competitiveness and adaptability. It enables ongoing knowledge generation and facilitates the recombination of existing knowledge with new insights, which is vital for developing new products, services and processes (Amara et al., 2016). Through this combinatorial function, firms can connect past experiences with emerging opportunities, thereby strengthening their ability to innovate while reducing uncertainty.

Moreover, scholars such as Fontana and Nesta (2009) and Kim and Lee (2016) argue that OK is not just supportive of innovation but fundamental to a firm’s long-term survival. In rapidly changing industries, where technology cycles are short and customer preferences shift quickly, OK enables firms to stay ahead by continually updating their knowledge and integrating external knowledge into internal processes. This not only improves strategic flexibility but also boosts the firm’s capacity to anticipate change and respond proactively.

Overall, OK is both an enabler and a result of organisational learning processes, contributing directly to firms’ adaptive capacity, innovation performance and strategic renewal. Therefore, its development should be prioritised within organisational strategies, especially in environments where rapid change and complex competition require high levels of learning agility and innovation readiness. Our first research hypothesis thus emerges:

Companies, even within the same industry, exhibit highly complex and unique learning processes shaped by a broad range of internal and external factors (Leoncini, 2016). These organisational learning processes are inherently connected to the outcomes of established routines and practices – whether they succeed or fail (Sirmon et al., 2007; Sirmon and Hitt, 2009). When a company’s practices produce positive results, the organisation usually interprets this as confirmation that it aligns with market expectations and environmental demands. Consequently, these routines are preserved or even strengthened, on the assumption that continuity will maintain success.

Conversely, when organisational practices fail to deliver the expected outcomes, a reflective and analytical process must be initiated. This enquiry seeks to identify the root causes of failure, evaluate the assumptions underlying the failed routines and generate actionable insights for future decision-making (Madsen and Desai, 2010; Dorfler and Baumann, 2014). In this sense, the experience of failure is not merely a setback, but a potential trigger for deep learning. The more frequently and systematically a company engages with its failures – rather than ignoring or concealing them – the more likely it is to develop adaptive capabilities that prevent similar outcomes in the future (Leoncini, 2016; Ferreira et al., 2020).

This learning dynamic becomes especially important when companies pursue innovation initiatives, which are inherently uncertain and carry a high risk of underperformance. When innovation projects fall short of expectations, organisations face the decision of whether to continue, suspend or abandon the initiative. In such instances, it is vital that any temporary pause in the project is not seen as a failure but rather as an opportunity for organisational learning and strategic reassessment (Ahn et al., 2018; Zouaghi et al., 2018; Ferreira et al., 2024). This reframing enables the company to examine what went wrong, identify overlooked contingencies and make necessary adjustments to improve the project’s course.

Through this reflective process, organisations may come to recognise that certain obstacles were not anticipated in the initial planning stages, leading them to abandon the project altogether if it is deemed no longer viable (Radas and Bozic, 2012). On the other hand, the same process might reveal that the innovation initiative holds more promise than initially perceived, provided that specific corrections are made to address the underlying issues (Dorfler and Baumann, 2014). Thus, failure in innovation does not necessarily signal the end of a project; rather, it may mark a critical juncture for transformation and growth.

Nevertheless, it is important to recognise that innovation failures do not always lead to future innovation. In some cases, repeated or traumatic failures can create psychological and organisational barriers that prevent further innovative efforts. This phenomenon, sometimes called “innovation trauma” or “innovation block,” can seriously weaken a firm’s willingness to take risks, lowering its overall innovation capacity (Zollo and Winter, 2002; Byrne and Shepherd, 2015). Therefore, while learning from failure is a powerful driver of long-term resilience and competitiveness, it requires an organisational culture that is open to experimentation, accepts the inevitability of setbacks and actively turns them into strategic advantages. Hence, our second research hypothesis emerges:

Companies do not exist alone; they coexist with all stakeholders in their environment. In this cohabitation, the importance of cooperation for knowledge transfer to promote innovative activities is increasingly recognised and verified (Acs et al, 2013; Schroll and Mild, 2012; Podmetina et al., 2016). The existing literature shows that the intensity of this cooperation and this relationship with KS is strictly linked with the development of dynamic capabilities to cope with turbulent environments (Cruz-González et al., 2015; Audretsch and Hinger, 2014; Zouaghi et al., 2018).

If OK allows companies to cross critical periods with a greater capacity for adaptation, guaranteeing their survival while managing to maintain their internal innovation capacities, even reducing the existence of innovation failures, the moderation of KS in this relationship (OK-innovation failures), allows the success of the same (Chesbrough and Garman, 2009; Garcia Martinez et al., 2019; Ferreira et al., 2020). KS can arise through cooperation in knowledge transfer with suppliers, customers, universities and even competitors, aiming to achieve success and reduce innovation failures (Belderbos et al., 2004; Un and Asakawa, 2015; Audretsch et al., 2025).

Cooperation with suppliers and customers allows firms to learn different skills of updating, modifying and absorbing the information transmitted by the market, thus reducing the failure of their innovative activities, making them stronger than their competitors (Fritsch and Franke, 2004; Miotti and Sachwald, 2003; Walsh et al., 2016). Cooperation with universities provides firms access to state-of-the-art technologies and knowledge tailored to the firm’s specific circumstances (Tsai, 2009; Stefan and Bengtsson, 2017; Wang et al., 2025). The connection of firms with this knowledge spillover allows them to develop innovative projects with a greater likelihood of success while reducing their development time (Chiaroni et al., 2008; Garcia Martinez, 2019; Lee et al., 2021). The connection with competitors and access to their resources and capabilities provide companies with access to specific technologies that may be more difficult to develop (van Beers and Zand, 2014; Braunerhjelm and Svensson, 2024; Stojčić et al., 2024). Thus, we verify that the relationship between OK and innovation failures is moderated by KS. In this sense, we present our third research hypothesis:

Several studies show a direct relationship between the degree of project failure and the probability that a company will abandon or fail to integrate innovative projects in the future; this is called innovative trauma (Välikangas et al., 2009; Kamoto, 2017; Colombelli et al., 2024). This position is, however, contested; for example, D’Este et al. (2017) show that OK gained from past failures decreases the probability of failure in future innovative projects. This knowledge can be acquired through failures in innovative processes and KS (Leoncini, 2016; Rhaiem and Amara, 2021; Barboza, 2024). After failures, in the context of future innovation, the moderation of KS allows companies not only to acquire new knowledge but also to learn from other companies’ failures, thereby avoiding their own future failures (Ingram and Baum, 1997; Leoncini, 2016; Yang et al., 2024).

It is precisely when economic conditions are more uncertain to invest in new knowledge that companies most need to insert KS, to guarantee their survival in the long term (Di Minin et al., 2010; Velu, 2015; Cerrato et al., 2016; Audretsch et al., 2024) and also to boost their innovative projects, ensuring their success (Lichtenthaler, 2011; Garcia Martinez et al., 2019). Thus, KS not only reduce failures in innovative projects but, in case they do occur, through their moderating effect, they help companies in new innovation processes, making them continue their innovative activity (Ferreira et al., 2020). Thus, we formulated our fourth research hypothesis:

Figure 1 contains the conceptual model we propose to test in our investigation.

The data set used in this study comes from the Community Innovation Survey (CIS), which has been widely used in innovation research (Mohnen et al., 2008; Duarte et al., 2017; Ferreira et al., 2023). The CIS is a harmonised questionnaire based on the Oslo Manual, with methodological recommendations specified by Eurostat and includes questions on a range of topics related to innovation activities. Our study used the CIS2018, which covers 2016–2018 and includes 100,115 companies with more than 10 employees from 14 EU countries (Bulgaria, the Czech Republic, Germany, Estonia, Greece, Spain, Croatia, Hungary, Lithuania, Latvia, Malta, Portugal, Romania and Slovakia).

We analysed data from 37,110 companies that reported product (goods or services) or process innovation or innovation expenditures. The sample includes companies from all sectors of activity, with 50.5% in manufacturing, 25.8% in services and 11.1% in trade. Table 1 presents the distribution of companies by country across CIS2018 and our study.

We proxy Future Innovation Capability by the percentage change in innovation expenditure from 2018 to 2019. This forward-looking, investment-based indicator reflects the firm’s revealed commitment to re-engage in innovation after setbacks. By design, it ranges from −100% (full disinvestment) upwards, with no fixed upper limit; negative values indicate contraction, zero signifies stability and positive values suggest increased commitment. We adopt this proxy because it is objective, comparable across firms and industries and aligns with knowledge-based and real-options views in which resource commitment signals prospective innovative effort. We recognise that spending is not equivalent to innovation capability or output; alternative outcome-focused measures (e.g. sales from new products, patenting) can be useful but often face delays, limited coverage or sectoral bias.

To measure OK, we use three items related to importance (0 – not important, 1 – slightly important, 2 – moderately important and 3 – highly important) assigned to staff work rotation planning in different functional areas; regular brainstorming sessions for staff to consider improvements within the company; and cross-functional workgroups or teams (combined into different work areas or functions) (Leoncini, 2016; Ferreira et al., 2020). The value of OK is obtained by averaging the three items (ranging from 0 to 3).

Innovation Failure is considered a binary indicator of abandoned innovation efforts (no/yes), representing projects or activities that were started but stopped before completion or commercialisation. This operational definition is commonly used to examine failure in innovation processes and has been used in previous research (e.g. Leoncini, 2016; Ferreira et al., 2020; Stojcic, 2024).

We operationalise the KS factor as a channel-based proxy for exposure to external knowledge – such as external research and deveopment (R&D), purchased services or materials for innovation and capital goods or intellectual property rights – combined with absorptive inputs (in-house R&D and innovation personnel) that enable assimilation (Ferreira et al., 2023). This specification captures both access to external knowledge and the capacity to absorb it, which aligns with the KBV and the absorptive capacity literature. We favour expenditure channels because they are objective and comparable across firms and industries, whereas direct observation of spillovers (e.g. patent citations, collaboration ties) is often incomplete or not harmonised at the firm level (Yi et al., 2021; Fu et al., 2022). In our data, the six indicators load on a single latent dimension (KMO = 0.719; variance explained = 62.5%), supporting the interpretation of a unified spillover-exposure construct.

We used dummy variables related to small and business economics (SME) to control the influence of firm size (0 – No; 1 – Yes). The other control variable identified was the sector in which the organisation operates. To measure the effect of the sector, we used dummy variables referring to the manufacturing industry (0 – No; 1 – Yes), commerce (0 – No; 1 – Yes) and services (0 – No; 1 – Yes) sectors (Leoncini, 2016; Ferreira et al., 2020; Ferreira et al., 2023).

All constructs are operationalised with CIS2018 items aligned with the Oslo Manual. OK combines three organisational routines – job rotation across functions, regular brainstorming and cross-functional teams – which capture learning/absorptive routines commonly used in CIS-based research. Innovation Failure follows the standard CIS dummy “Abandoned innovation activities”. Future Innovation Capability is the percentage change in innovation expenditure from 2018 to 2019, viewed here as a forward-looking, investment-based proxy of a firm’s ability to re-engage with innovation after setbacks. KS is summarised through a principal component derived from six CIS innovation-expenditure shares (including in-house R&D, external R&D, other innovation outlays, own personnel involved in innovation, purchased services or materials for innovation and capital goods for innovation), aligning with prior CIS studies of external knowledge access. We rely on the CIS instrument’s content validity and the well-established use of these items in the literature; we also retain the previously reported PCA diagnostics for KS (e.g. single-factor solution; KMO and explained variance) already included in this section.

Four regression analyses are combined to assess the hypothesis (Baron and Kenny, 1986). In Step 1, to test the negative impact of OK on innovation failures (H1), we used a logistic regression model (Model 1). Based on the questionnaire, we used abandoned innovation activities as the dependent variable and OK and control variables as independent variables. We included the interaction terms and used Model 3, a logistic regression model, to test H3. This hypothesis suggests that KS play a moderating role in the relationship between OK and innovation failures.

The function used in logistic regression to estimate the probability that a certain realisation j (j = 1, …,n) of the dependent variable occurs, that is, that companies abandon innovation activities $(P[Yj=1]=π^j$ can be expressed by $π^=eXβ1+eXβ$ where $π^$ is the vector of estimated probabilities, X is the matrix of independent variables and β is the vector of logistic regression coefficients (Cameron and Trivedi, 2005). As the logistic model was used, the Odds Ratio (OR) was estimated. The OR is a measure that quantifies the relationship between the presence of an independent variable and the probability of the event of interest. When the OR is less than 1, it indicates that the independent variable is associated with a decrease in the probability of the event occurring. Linearising this function with the logit transformation of the dependent variable, the econometric logistic regression model under analysis is obtained, estimating the following model:

In Step 2 of our analysis, we used a linear regression model (Model 2) to test our second hypothesis, which aims to investigate the impact of innovation capacity (H2). We used the variation in innovation expenditure between 2018 and 2019 (%) as the dependent variable, along with the innovation failure variable and control variables based on the questionnaire. To examine the moderating effect of KS on the relationship between innovation failures and future innovation capability (H4), we incorporated the interaction factor between innovation failures and KS in our analysis (Model 4). Therefore, the estimated econometric model was as follows:

To ensure that our statistical model is accurate and reliable, we conducted both residual and influence analyses (Cameron and Trivedi, 2005; Hair et al., 2018). Specifically, we examined the residuals to check for homoscedasticity, independence and normality. These factors are especially important when working with linear regression. We used residual-versus-fitted-value plots to identify patterns that may indicate heteroscedasticity or dependence in the data (Greene, 2018; Hair et al., 2018). In addition, we applied the Durbin–Watson test to detect autocorrelation in the residuals and used histograms and QQ-plots to confirm that the residuals were normally distributed (Cameron and Trivedi, 2005; Greene, 2018).

In the influence analysis, we focused on identifying any data points that may have significantly impacted our model. We did this by calculating leverage measures to assess how distinct the observations were from the overall data set and using Cook’s distance to identify any observation that may have had a potentially significant influence on the model’s coefficients (Cameron and Trivedi, 2005; Hair et al., 2018). Because data were collected across different countries, violating the assumption of independence of observations within each country, we used clustered robust standard errors with country as the clustering variable (Greene, 2018; Hair et al., 2018). Finally, we checked for possible multicollinearity between the variables used in the analysis. This involved examining the correlations between the variables used in the regression analysis and calculating the variance inflation factor (VIF) to assess the level of collinearity (Greene, 2018). These steps are fundamental to ensuring that our model is reliable and provides accurate estimates.

Table 2 contains the descriptive statistics of the study variables (except for sector and year dummy variables), correlation coefficients and VIF. Most companies in our sample (51.7%) were SMEs, 14.5% abandoned innovation activities and an average increase of 10% in innovation expenditure was expected between 2018.

We test H1–H4 with Models 1–4, respectively (Model 1 → H1; Model 2 → H2; Model 3 → H3; Model 4 → H4), with estimates reported in Tables 3–4. Table 3 contains the results of the logistic regression model (Model 1), the impact of OK on innovation failures (H1) and of a linear regression model (Model 2), which assesses the effect of innovation failures on future innovative capacity (H2).

With regard to H1 “The greater the OK, the smaller the innovation failures”, it appears that there is a statistically significant impact of the level of OK on the abandonment of the innovative activity (Model 1: OR = 0.67; p < 0.01), in which the higher the level of OK, the lower the probability of abandoning the innovative activity. These results confirm H1. Thus, our H1 is statistically supported, corroborating the results of other studies (Cohen and Levinthal, 1990; Zahra and George, 2002; Garcia Martinez et al., 2017).

As for H2, “Innovation failures have a negative impact on future innovative capacity”, there is a statistically significant effect of the abandonment of innovative activity on the variation of future investment in innovation (Model 2: B = −0.09; p < 0.05), in which companies that abandon innovative activity have significantly lower levels of future investments in innovation. These results support H2 and other investigations (Zollo and Winter, 2002; Byrne and Shepherd, 2015).

Table 4 includes the results of the logistic regression model (Model 3) that analyses the moderating effect of KS on the relationship between OK in innovation failures (H3) and the linear regression model (Model 4) referring to the moderating effect of KS in the relationship between innovation failures and future innovation capacity (H4).

Regarding the control variables, it is only found that manufacturing has a significantly higher propensity to abandon innovative activity (Model 1: OR = 1.42; p < 0.05; Model 3: OR = 1.46; p < 0.05), as well as firms in the services sector (Model 1: OR = 1.48; p < 0.05).

In terms of H3, “The relationship between OK and innovation failure is moderated by knowledge spillovers”, based on Model 3, it is found that the interaction term OK and KS have no statistically significant negative impact on innovation failure (Model 3: OR = 1.07; p = 0.475). As other authors argue, these results do not support Hypothesis 3 (Chiaroni et al., 2008; Garcia Martinez, 2019).

Regarding H4, “The relationship between innovation failures and future innovative capacity is moderated by knowledge spillovers”, Model 4 reveals that the interaction variable between innovation abandonment and KS has a statistically significant positive effect on future investment in innovation (Model 4: B = −0.04; p < 0.01), i.e. the higher the level of KS the lower the negative impact of abandonment of innovative activity on future investment in innovation. These results support H4, as other research argues (Lichtenthaler, 2011; Garcia Martinez, 2020).

Finally, we present our validated model in Figure 2.

Taken together, the estimates indicate that ex ante prevention via OK (H1) leads to a negative association between failure and investment in short-term innovation (H2). There was no moderation of KS on the OK → failure relationship (H3); however, a mitigating effect of KS was observed after failure (H4).

Innovation failures have long been examined by scholars (Shepherd et al., 2011; Eggers, 2012; Shepherd et al., 2013; Khanna et al., 2016), yet their actual organisational consequences – both positive and negative – remain insufficiently clarified. While some studies highlight the potential for learning and renewed innovation effort (Leoncini, 2016; Ferreira et al., 2020), others emphasise that failure may generate long-term disengagement, risk aversion and organisational paralysis (Shepherd, 2003; Cannon and Edmondson, 2001). This conceptual ambiguity underscores the need for more systematic, empirically grounded investigations (Gavetti and Levinthal, 2000; Zollo and Winter, 2002; Byrne and Shepherd, 2015).

Innovation failures are inherent to organisational life, particularly when understood as unforeseen contingencies that disrupt strategic intentions. In the context of increasingly volatile and complex environments, understanding how firms process and respond to failure has become essential. Knowledge – both internal and external – plays a pivotal role in shaping these responses, determining whether firms recover, adapt and continue innovating or instead stagnate.

Using CIS data and both linear and logistic regression models, we examine (i) how OK influences the likelihood of innovation failure, (ii) how innovation failure affects future innovation capacity and (iii) whether knowledge spillovers (KS) moderate these relationships.

Our results show that OK significantly reduces the likelihood of innovation failure (Model 1: OR = 0.67; p < 0.01), reinforcing the idea that internal knowledge capabilities act as a preventive mechanism. However, KS do not significantly moderate this relationship (Cefis and Marsili, 2012; Howell, 2015), suggesting that when firms already possess strong internal knowledge systems, external knowledge provides limited additional protection.

Conversely, innovation failure negatively affects future innovation activity (Model 2: B = −0.09; p < 0.05), providing evidence that firms may not automatically learn from failure and can become hesitant to re-engage in innovative initiatives (Välikangas et al., 2009; Kamoto, 2017). Yet, when KS are introduced into the model (Model 3: OR = 1.07; p = 0.475; Model 4: B = −0.04; p < 0.01), their buffering effect becomes clear: KS increase the probability that firms re-enter the innovation process after having experienced failure. Spillovers therefore help firms reinterpret failure constructively, reducing psychological, strategic and informational barriers to re-engagement and supporting recovery (Leoncini, 2016; Rhaiem and Amara, 2021).

Overall, our findings argue for a dual role of knowledge: OK is central in preventing innovation failures, while KS are essential in overcoming their consequences.

Our findings emphasise the need to examine how firms engage in knowledge transfer processes to reduce the risk of innovation failures and, when such failures occur, how knowledge enables them to recover and maintain their future innovation capacity. This relationship has not yet been empirically studied in the existing literature – indeed, Rhaiem and Amara (2021) explicitly called for future research to investigate this very issue.

Our research shows that firms need to actively learn from past innovation failures to improve their future innovation efforts. When failure is not followed by structured knowledge transfer processes, organisations often struggle to extract valuable lessons, which impairs their ability to innovate confidently. This insight addresses a significant gap in previous studies, which have largely ignored the dual role of knowledge in both preventing innovation failures and reducing their long-term negative impacts. Our research helps clarify whether firms can independently turn failure into learning and resilience, or if they rely on external knowledge support to do so.

Our findings indicate that while OK plays a critical role in reducing the likelihood of innovation failures, overcoming their consequences and regaining innovative momentum requires the complementary presence of knowledge spillovers (KS). Firms embedded in knowledge-rich environments are better positioned to recover from setbacks and resume innovation because such contexts foster constructive reflection and learning. By cultivating an open culture that encourages analysis and discussion of failure, organisations can identify new pathways and reignite innovation efforts. Conversely, the absence of a supportive knowledge ecosystem significantly constrains learning from failure, reinforcing that the detrimental impact of failure on future innovation capacity is particularly pronounced in firms lacking access to external knowledge sources.

Rhaiem and Amara (2021) emphasised the need for research that empirically tests how the KBV of the firm informs learning from innovation failure, particularly regarding knowledge development and storage. Our study directly responds to this call by proposing a model that identifies specific stages of the innovation failure process where knowledge has the most significant influence. In doing so, we contribute to the emerging literature on the role of knowledge in failure recovery, highlighting that, as argued by Ferreira et al. (2020) and Testa et al. (2024), firms that do not effectively capture and use knowledge from past failures are unlikely to sustain innovation over time.

An important implication of our findings is that, during the early stages of innovation, where firms already possess substantial OK, the role of KS in reducing innovation failures is relatively limited. While OK is indispensable for preemptively mitigating certain types of failure, it cannot eliminate them on its own. In contrast, in the post-failure phase, KS become critical for rebuilding innovation capacity and restoring momentum. This reinforces the need for firms to deliberately establish mechanisms that enable learning from failures – particularly within environments that promote psychological safety, encourage candid feedback and facilitate open discussion of sensitive issues. Such learning processes are most effective when supported by constructive conflict, in which diverse, even opposing viewpoints are welcomed and debated productively.

Our research also makes a substantive contribution to the literature on organisational learning. Firstly, we evidenced that firms rarely learn from failure in isolation. Without access to KS, companies often fail to re-range in innovation following a setback. Conversely, when supported by external knowledge flows, firms recover more effectively, providing theoretical evidence that experience alone is insufficient for learning. As highlighting Chen et al. (2025) and Stojcic (2024), codified knowledge plays a stronger influence on organisational learning than raw experience, which can remain inert without deliberate reflection. Extracting valuable lessons from past failures requires time, effort and managerial commitment.

Furthermore, although OK contributes meaningfully to reducing innovation failures, our findings reveal that many firms lack a robust culture of idea-sharing. This is reflected in the limited moderating effect of KS on the relationship between OK and innovation failure. In contrast, our data show that, after a failure, the presence of KS – particularly in an open-dialogue climate – enables organisations to re-enter an innovative trajectory.

A further theoretical implication of our study is the identification of a gap in the literature: the lack of explicit theoretical models addressing innovation failure (Rhaiem and Amara, 2021). Our work addresses this gap by applying the KBV framework to explain when and how knowledge transfer becomes most critical after innovation failure. In doing so, we offer a clearer understanding of the timing and mechanisms through which knowledge fosters resilience and sustains innovation.

Finally, our research contributes to the growing literature on organisational adaptation and learning during strategic reorientation. Specifically, we emphasise the importance of idea exchange and constructive conflict in shaping strategic decisions following failed innovation attempts. The presence of KS facilitates productive internal dialogue, enabling employees to express differing views, challenge assumptions and collaboratively develop new pathways forward. As Salvato and Vassolo (2018) argue, the ability to respect diverse perspectives and co-create adaptive solutions is a critical factor in organisational renewal.

Our research on how firms can effectively learn from innovation failures offers practical insights that are highly beneficial to organisations, particularly because, as previously highlighted, there remains a noticeable gap in evidence-based, actionable guidance for managers and practitioners (Zollo and Winter, 2002; Byrne and Shepherd, 2015). Although the academic literature increasingly acknowledges the importance of organisational learning from failure, it is also evident – both from our findings and those of prior studies – that many companies struggle to undertake this learning process independently (Amankwah-Amoah and Wang, 2019; Costa et al., 2023). In practice, the ability to derive meaningful lessons from innovation failures is often hindered by internal barriers, such as fear of blame, a lack of structured reflection and the absence of supportive knowledge-sharing mechanisms.

Our findings reveal that companies are more likely to resume innovative activities following a failed project only when KS are present. These KS act as critical enablers of recovery by introducing external perspectives, resources and shared experiences that help companies make sense of what went wrong and how to improve. In contrast, when organisations choose to internalise or conceal their failures – typically due to fear of reputational damage or leadership stigma – they lose valuable opportunities to reflect, adapt and evolve (Leoncini, 2016). Such stigma often results in what we term “innovation trauma,” a condition in which companies become risk-averse and avoid re-engaging in innovative ventures altogether.

Thus, the central practical implication of our research is that failure should neither be ignored nor punished. Instead, managers should view failure as an integral component of the innovation process – a source of feedback rather than defeat. They must foster a culture that encourages open discussion and transparency around setbacks, not suppresses them. In doing so, firms can create an environment where constructive conflict – open, honest dialogue involving differing viewpoints – is welcomed and channelled towards organisational growth and continuous improvement.

In this context, tolerance for failure becomes not a sign of weakness, but a strategic asset. Managers are encouraged to treat failures as diagnostic tools: opportunities to reassess assumptions, revise strategies and realign goals. Our data suggest that in the absence of knowledge spillovers, companies are less likely to return to innovation after failure, underscoring the essential role of external and cross-boundary learning mechanisms in enabling recovery and future innovative capacity.

To move beyond reactive responses and towards proactive learning, we recommend that firms adopt a “strategic closure manual” for failed innovation projects (Corbett et al., 2007). This manual serves as a structured framework for intentionally terminating unsuccessful initiatives, ensuring that all relevant data – performance indicators, unmet targets, missteps and contributing factors – are systematically captured. The goal is not only to assign responsibility where necessary but also to codify lessons learned to inform future innovation processes.

Overall, our research emphasises that learning from failure must be intentional and supported – not left to chance. Organisations that institutionalise reflection, promote transparency and facilitate knowledge exchange – both internally and externally – are better equipped to overcome setbacks, prevent innovation paralysis and develop long-term resilience and adaptability in an increasingly dynamic business environment.

As with any model or methodological approach, certain limitations and assumptions warrant further discussion. To address these, we propose some avenues for future research that could yield insights beyond the scope of our study. Specifically, future work may explore the content and mechanisms of learning and knowledge transfer in greater depth. While our analysis focused on the moderating role of KS before and after innovation failures, subsequent studies examine failures across different organisational functions as some may be inherently more complex or ambiguous than others. Although we adopt standard CIS-based operationalisations for comparability and parsimony, alternative proxies exist in this literature. For OK, studies often use latent indices of organisational routines (e.g. factor scores) or the CIS “organisational innovation” indicator; for KS, measures such as cooperation breadth (number of external partners) or intensity indicators like external R&D are common. Innovation failure could also be broadened to include postponed or downsized activities, while future capability might be captured through log-change in innovation spending, binary increases or revenue shares from new products. We retained our current measures to preserve cross-study comparability and explicitly note these choices as scope conditions for future research.

Further investigation into technological versus non-technological failures is warranted, as the latter tends to be more ambiguous and influenced by external factors such as consumer behaviour, whereas technological failures often stem from internal faults. Conversely, a technological failure occurs due to a fault within the company. Future research could also examine how lessons learned are codified and disseminated across organisational levels, as well as the potential relational costs of failure analysis. Understanding when and how such analyses negatively impact firms would provide valuable managerial guidance.

Methodologically, future research could benefit from incorporating qualitative approaches – such as interviews or case studies – to capture the nuanced social dynamics and communication practices that shape learning after failure. Comparative studies across industries or firm sizes may also reveal distinctive patterns in knowledge sharing and adaptation.

Ultimately, our research sought to critically examine the interplay between knowledge, innovation failure and future innovation. In high-complexity, high-uncertainty contexts, the path to successful innovation often involves inevitable setbacks. While firms may tolerate failure, they risk repeated missteps in areas that receive insufficient attention. Only organisations that deliberately extract lessons from failure and institutionalise learning processes are positioned to transform setbacks into a source of competitive advantage.