Role of innovation ambidexterity in technology startup performance an empirical study

Full Terms & Conditions of access and use can be found atTechnology Analysis & Strategic ManagementISSN: Print Online Journal homepage: www.tandfonline.com/journals/ctas20Role of innovat

Full Terms & Conditions of access and use can be found at

Technology Analysis & Strategic Management

ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/ctas20

Role of innovation ambidexterity in technologystartup performance: an empirical study

Ambreen Khursheed & Faisal Mustafa

To cite this article: Ambreen Khursheed & Faisal Mustafa (2024) Role of innovation

ambidexterity in technology startup performance: an empirical study, Technology Analysis & Strategic Management, 36:1, 29-44, DOI: 10.1080/09537325.2021.2020235

To link to this article: https://doi.org/10.1080/09537325.2021.2020235

Published online: 24 Dec 2021.

Submit your article to this journal

Article views: 675

View related articles

View Crossmark data

Citing articles: 4 View citing articles

Role of innovation ambidexterity in technology startup performance: an empirical study

Ambreen Khursheed and Faisal Mustafa

Faculty of Management Studies, University of Central Punjab, Lahore, Pakistan

Innovation ambidexterity is equivocal and this concept holds a significant position in the literature of strategic entrepreneurship as it binds together the crucial dimensions of organisational learning theory Ambidexterity is recognised as a source for achieving higher levels offirm performance; however, scholars left a gap in examining how innovation ambidexterity is achieved, particularly so in startups Therefore, this study investigates how a balance between exploratory and exploitative innovations can lead towards better startup performance under the moderating effects of absorptive capacity and environmental dynamism This is a cross-sectional empirical research on young incubating technology startups and the data were analysed using partial least square structural equation modelling (PLS-SEM) Our results highlight that innovation ambidexterity plays a crucial role in achieving better startup performance This research contributes to the ambidexterity literature by providing quantitative evidence at a large scale, delineating the procedure through which entrepreneurs can achieve innovation ambidexterity The research enriches the organisational learning theory by providing a more inclusive understanding of innovation ambidexterity concept.

After developing several studies on organisational ambidexterity, recent research has shifted its attention to exploitative and explorative innovations (Wang and Zhang2020) As in the existing changing and uncertain business environment, entrepreneurs across the world are obliged to con-tinuously update, reinvent and improve their startups for their survival in the markets (Chung, Yang, and Huang2015) Moreover, the fourth industrial revolution is going to begin so this will transform the way businesses innovate, learn technologies and exploit their resources The energy and pace of this revolution are turbulent and unmatched Therefore, practitioners and researchers believe that only technological advancement can assistfirms to compete in dynamic environmental conditions (Giones and Miralles2020; Khursheed et al.2020) Therefore, it is significant to consider the way a firm interacts with their suppliers, employees and customers Furthermore, in the existence of una-voidable effects of COVID-19, the innovation processes of businesses have become an essential requirement for survival (Mustafa et al.2021).

In the literature of organisational learning, exploratory and exploitative learning present two crucial and most influential categories of learning (Wang and Zhang2020) Exploratory learning pre-sentsfirm behaviours categorised by discovery, search and experimentation, whereas the exploita-tive learning presentsfirm behaviours comprising of refinement, efficiency and implementation.

CONTACTAmbreen Khursheedambreen.khursheed@ucp.edu.pk2024, VOL 36, NO 1, 29–44

https://doi.org/10.1080/09537325.2021.2020235

However, empirical studies have not clarified the significant elements or moderating mechanism in the strategic entrepreneurial factors-innovation ambidexterity relationship, which might provide ambiguous outcomes Wefill this gap by presenting a new multidimensional model of innovation ambidexterity, investigating the startup performance using the theoretical framework of organis-ational learning theory Therefore, this research bridges the gap by aligning essential and conflicting strategic entrepreneurial factors to balance explorative and exploitative innovations leading to enhance startup performance.

In this study, we investigated young technology startups as they are more vulnerable to exogen-ous shocks indicating their survival is highly dependent on the way they balance market exploitation and exploration– innovation ambidexterity (IA) – and their capability to absorb knowledge – absorp-tive capacity (AC) and the capability to manage environmental challenges– environmental dyna-mism (ED).

This research is motivated by following main issues such as there is no widely applicable theor-etical model of ambidexterity, which can be used by young incubating startups for achieving better performance Although, some of the past researchers presented a model of ambidexterity, their research lacks a complete operationalisation of the innovation ambidexterity concept (Soto-Acosta, Popa, and Martinez-Conesa2018; Wiratmadja, Profityo, and Rumanti 2020) To date, the understanding about how and when exploratory and exploitative learning can influence startup per-formance is not sufficient In the existing literature, there has been very limited investigation about the influence of the balancing strategy of innovations on the startup performance Therefore, the research motivation is to contribute to the existing literature by providing a comprehensive multi-dimensional ambidexterity model to startupfirms for integrating strategic entrepreneurial factors for overcoming the challenge of achieving ambidexterity.

The contributions of this research are threefold First, this research assists young entrepreneurs and incubation managers in understanding the significance of ambidexterity framework in achieving better startup performance Second, considering the existing theoretical claims about the role of organisational learning and its impact on young startup’s performance (Chung, Yang, and Huang

2015) Third, this research adds fresh insights into thefield of entrepreneurship about crucial mod-erating roles of absorptive capacity and environmental dynamism This study enriches the under-standing of professionals, academicians and bureaucrats to design strategic policies for startups leading to resolve the challenge of achieving ambidexterity The remainder of the research study is designed as follows First, the conceptual foundations are discussed and consequently, a theoreti-cal model is developed Then, the theoretitheoreti-cal model is empiritheoreti-cally tested with the help of partial least square structural equation modelling on the data obtained from 285 incubated startup representa-tives Finally, the results are explained along with the theoretical and practical implications of the study.

Theoretical framework

Various theories have been used in the past research studies in strategic entrepreneurship and ambi-dexterity research (Hughes et al.2021) In this study, we have used the theoretical foundation of ‘Organisational Learning Theory’ and it’s one of the most important streams ‘Organisational Ambi-dexterity’ The organisational learning perspective (OL) is rooted in the organisational behaviours and management history It is presented through organisational acts or routines and helps to improve the organisational performance and behaviours (Dixon, Meyer, and Day 2007) One of the most significant stream of organisational learning is the organisational ambidexterity (OA) and it refers to the capability of a firm to balance two competing activities (O’Reilly III and Tushman2013) In case of startups, ambidexterity represents the ability to simultaneously exploit available resources and explore new market opportunities The literature about ambidexterity within the context of young incubated technology startups is not properly developed Therefore, drawing on organisational learning theory with a particular focus on ambidexterity concept, this

research examines the relationship between strategic entrepreneurial factors with innovation ambi-dexterity and ultimately its effect on the startup performance.

Entrepreneurial orientation

Entrepreneurial orientation (EO) creates an action-focused prejudice that motivates the firm to pursue market-focused entrepreneurial options for innovation (Wang and Zhang2020) The litera-ture has highlighted that entrepreneurial orientation shares a close link with the organisational learning theory (Lumpkin and Dess 1996) By having a strong entrepreneurial orientation, a startup can develop a solid knowledge base, which will ultimately improve the startup’s capability to access, utilise and disseminate learned knowledge (Khan et al.2020) Therefore, using the theor-etical framework of OL theory, it is crucial to determine whether an entrepreneurially orientedfirm prefers innovation exploration or innovation exploitation or both Therefore, we propose the follow-ing hypotheses:

Hypothesis 1a: A positive relationship exists between entrepreneurial orientation and explorative innovation.Hypothesis 1b: A positive relationship exists between entrepreneurial orientation and exploitative innovation.

Technology orientation

Founders of technology startups primarily depend on the following two things: the ability of a business to respond towards evolving needs and the discovery of contemporary techniques to main-tain the benchmark of excellence Small ambidextrousfirms can design both incremental and radical innovation at the same time because of their learning capabilities (Tiberius, Schwarzer, and Roig-Dobón 2020) Therefore, it is crucial to determine the preference of a technologically oriented startup regarding explorative and exploitative innovations However, this area is still unexamined within the context of young tech-startupfirms of developing countries Thus, the following hypoth-eses are developed:

Hypothesis 2a: Technology orientation positively influences explorative innovation.Hypothesis 2b: Technology orientation positively influences exploitative innovation.

Financial resources

The literature has proved thatfinancial resources are drivers of innovative products/services but limited studies have investigated the role offinancial access in explorative and exploitative inno-vations (Zhu et al.,2020) Moreover, it is not investigated by prior research studies whether entrepre-neurs prefer to invest in exploratory innovation or exploitative innovation Therefore, it is very important to determine how access tofinancial resources facilitates a firm in explorative and exploi-tative innovations Therefore, the following hypotheses are proposed:

Hypothesis 3a: Accessingfinancial resources shares a positive relationship with explorative innovation.Hypothesis 3b: Accessingfinancial resources shares a positive relationship with exploitative innovation.

Relational resources

Relational resources mainly comprise networks between thefirm and other organisations such as vendors, rivals, government institutions and clients It has been witnessed that the reconfiguration of the resources accessible to new technology startups from its internal network ties allows thefirm

to make the latest explorative endeavours for its present and future customers, but averts resource utilisation in the existing exploitative practices of the organisation (Voss, Sirdeshmukh, and Voss

2008) Additionally, very limited studies have been done to relate two crucial concepts of relational resources and organisational learning theory Therefore, tofill this gap, this research proposes that a firm’s knowledge transfer capability is analysed through its externally acquired knowledge at the expense of the knowledge distributed internally under the theoretical framework of organisational learning Therefore, the following hypotheses are formulated:

Hypothesis 4a: Relational resources share a negative relationship with explorative innovation.Hypothesis 4b: Relational resources share a positive relationship with exploitative innovation.

Relational embeddedness

Relational embeddedness comprises of values including mutual trust and the relationship between individuals orfirms The literature revealed that as the relational embodiment grows over the period, the transference of knowledge decreases with time (Dezi et al 2019) Hence, considering the accepted significance of relational embeddedness in achieving better firm performance, it is crucial to empirically investigate this relationship for young incubating startups Thus, we have pro-posed the following hypotheses:

Hypothesis 5a: A negative relationship exists between relational embeddedness and explorative innovation.Hypothesis 5b: A positive relationship exists between relational embeddedness and exploitative innovation.

Innovation ambidexterity and startup performance

Existing entrepreneurship discourse builds a strong argument on the belief that firms pursuing exploratory and exploitative innovations are more credible to accomplish higher returns as com-pared to those who pay attention to just one factor while neglecting the other (Hughes et al.

2021) Furthermore, effective exploration in one technological discipline can support the exploita-tion activities of the corresponding discipline However, the attractiveness of young technology startupfirms to excel in accomplishing innovation ambidexterity is the opportunity to secure the advantages of the exploration while balancing exploitation at the same time (Cho, Bonn, and Han

2020) Prior studies have mainly investigated firms in technologically advanced countries, thus, leaving a gap for empirical studies on startups within the context of developing markets Therefore, this study aims tofill this gap by proposing the following hypothesis:

Hypothesis 6: A positive relationship exists between innovation ambidexterity and technology startupperformance.

It is important to carefully measure innovation ambidexterity Therefore, this research uses the following equation for measuring innovation ambidexterity developed by Hughes et al (2021):

Innovation Ambidexterity = S (Explore × Exploit) – 




















(Explore− Exploit)2

The moderating role of environmental dynamism

Environmental dynamism refers to the rate of variation and the unpredictability of environmental conditions (Yap et al.2020) Therefore, to understand the relationship between innovation ambidex-terity and startup performance it is crucial to understand the impact of external environmental con-ditions (Wiratmadja, Profityo, and Rumanti2020) In this research, we have selected environmental dynamism as a moderator because new ventures must assess the environment and update their

businesses with the latest capabilities to succeed in dynamic markets Therefore, it is crucial to under-stand how environmental dynamism will affect the relationship between innovation ambidexterity and technology startup performance Therefore, we have proposed the following hypothesis:

Hypothesis 7: Environmental dynamism moderates the relationship between innovation ambidexterity andtechnology startup performance.

The moderating effect of absorptive capacity

Absorptive capacity is defined as ‘a set of organizational routines and processes through which businesses systematically acquire, transform and implement knowledge’ (Zahra and George

2002) Past studies have used the lens of the absorptive capacity for explaining innovation with the help of different theoretical models; however, considering the theoretical underpinning of the OL theory, a research gap exists for more empirical studies analysing the moderating effect of absorptive capacity on the relationship between innovation ambidexterity and startup performance Therefore, this researchfills this gap by proposing that the impact of innovation ambidexterity on a startup performance will be stronger in the presence of higher levels of absorptive capacity Thus, the following hypothesis is proposed:

Hypothesis 8: Absorptive capacity moderates the relationship between innovation ambidexterity and technol-ogy startup performance.

Research methodology

This is a cross-sectional empirical research conducted to examine and test hypotheses deduced from the existing entrepreneurship literature We have used the quantitative survey method The popu-lation comprised of incubating young technology startups from Pakistan We selected incubation centres geographically due to the following two reasons:first, cities present incubation clustering and second, our target sample is comprised of incubated technology startups This restricted us to only two main cities of Punjab where recently a large number of technology startups are launched We have used a non-probability sampling technique, as this technique is more practical and conducive for researchers conducting surveys In our study, we collected data from the startup representatives, therefore, purposive sampling method is the most appropriate method We have calculated sample size using Bentler and Chou (1987) and Hair et al (2006) method As per this method, for calculating an appropriate sample, a minimum offive and a maximum of 15 subjects per item are required to be selected The questionnaire used in this research comprised of 54 indicators so the total sample size is (54 × 6) 324.

Data collection method

For the collection of data, an online and self-administered survey from startup founders, cofounders and vice presidents was conducted The questionnaire was pretested from entrepreneurship pro-fessors for detecting any inadequate wording and this helped in the improvement of the instrument We also conducted a pilot test with 45 startup founders incubated in a private incubation centre of Lahore, Pakistan to inspect the statistical aspects of the research’s measures None of those 45 ques-tionnaires were incorporated in thefinal analysis of the study We have selected the items of the questionnaire from the past-validated constructs with slight changes made to the language of some items for clarity.

This research uses a seven-point Likert scale to assess the participants’ agreement or disagree-ment with the statedisagree-ment In total, 324 questionnaires were distributed, of which 297 questionnaires were received, of which 12 were incomplete Therefore, we got 285 questionnaires usable for the analysis of data, presenting a response rate of 88% The respondents were ensured about the

privacy of their data by informing them that the survey will not collect any information leading to their identification and the results will only be used for research purposes The respondents were required to access the questionnaire online, so their survey completion was considered as an evi-dence of their consent to participate (Table 1).

Statistical procedure

This study has tested the structural and measurement model relationships with the help of partial least square structural equation modelling (PLS-SEM) (Lin et al.2020) We have selected the PLS-SEM as it is the most suitable approach when the focus is on the best prediction of relationships among the variables, which is the objective of our study.

Common method variance

To address, common method variance (CMV) concerns, we have designed our questionnaire com-prising of three separate sections, section (a) comprised of questions regarding the demographic characteristics The section (b) comprised of questions regarding the business and section (c) com-prised of questions about entrepreneurial characteristics For limiting CMV, we arranged the ques-tionnaire items randomly, mixed sequence of questions, and used neutral wording for helping respondents in understanding the questionnaire.

We have also applied Harman’s one-factor test Our findings confirmed that common method variance has not contaminated the data Kock (2015) reported that variance inflation factors (VIF)

Table 1.Demographic data of the startupfirms.

are produced for all latent constructs and it is recommended to use a threshold of 10 for identifying the common method variance After examining the outer variance inflation factors, all constructs presented a VIF <10 This ensured that the common method variance (CMV) has not affected our results Thus, we used the research model for the data analysis For testing the reliability and validity, we conducted confirmatory factor analysis (CFA) The results of the measurement model are pre-sented inTable 2 The goodness offit index demonstrated acceptable levels as shown inTable 3 Our results indicated the best fit of our model The study’s confirmatory factor analysis and partial least square structural equation modelling are performed on the SmartPLS 3.0 software We have used the SmartPLS software because it is one of the most efficient softwares for testing complex models having reflective and formative models.

We have used the PLS algorithm under basic settings including weighting scheme (path) with 300 iterations with a stop condition of 10 to 7 (1.0E-07) The significance of p-value, t value and path analysis along with accelerated confidence bootstrap intervals and 95% bias correlated was achieved through selecting bootstrapping method with a subsample of 5,000 and 5% one-tailed significance.

Measurement model analysis

This study has tested the measurement model using the standard process recommended by Hair et al (2017) including convergent validity, internal consistency reliability, discriminant validity and indicator loadings.

Indicator loadings

The measurement of indicator loadings explained the limit to which an indicator is described through its latent constructs The required assessment of factor loadings must be understood in

Table 2.Measurement model results.

ConstructFactor loadingst-valueAVECronbach’s alpha Composite reliability (CR) Rho (ρA)Entrepreneurial orientation0.841–0.8992.140.7520.8120.8310.75

Access tofinancial resources0.813–0.8952.550.7160.8470.8580.71Access to relational resources0.817–0.8992.480.7650.8170.8770.73

Table 3.Fit statistics of the measurement model and confirmatory analysis.

Note: CMV, common method variance; CFA, confirmatory factor analysis.

principle following loadings in the factor analysis Therefore, factor loadings having significant values of >0.70 are required to describe at least 50% of an indicator variance or significant figures >0.50 in exploratory studies (Hair et al.2017).Table 2shows values of all loaded indicators of latent constructs were higher than the acceptable threshold (0.50) We also confirmed the significance of indicator loadings through the bootstrapping method, which resulted in a t-statistic value >1.96 (two-tailed test; p < 05) This indicated a higher level of indicator reliability.

Internal consistency reliability

It was evaluated through Dijkstra-Henseler’s rho (ρA), composite reliability and Cronbach’s alpha value The results revealed that the value of our latent constructs in all selected three reliability tests is more than the required threshold of 0.70 Thus, the results fulfilled the requirements.

Convergent validity

It was assessed through average variance extracted (AVE) and its score is more than the threshold value of 0.50 (Fornell and Larcker 1981) The results showed all latent constructs had values of AVE >0.50 as shown inTable 2.

Discriminant validity

For testing discriminant validity, it is recommended to check whether the values of the AVE square root are greater than the correlations between each latent variable and other latent variables (Fornell and Larcker1981) Confirming this recommendation, our results revealed the square root of each AVE (presented on the diagonal) was more than the related correlations present among constructs This confirms discriminant validity is present for latent constructs examined in this study Further-more, by assessing the cross-loadings, we also accessed how much the assessments of various latent constructs deviated in the measurement model Finally, using the heterotrait-monotrait (HTMT) approach, we determined discriminant validity as the HTMT index having a score <0.85 or 0.90 is considered acceptable (Henseler, Ringle, and Sarstedt2015) The HTMT analysis inTable 4

shows that all latent constructs completely fulfilled the required criteria as all values were <0.85 (see scores above the diagonal) Thus, the result confirms each latent construct examined a unique subject.

Structural model

We examined the structural model by evaluating the standard criteria comprising of path coe ffi-cients, predictive relevance Q2, coefficient of determination (R2 value), overall goodness-of-fit and collinearity (Schlägel and Sarstedt2016).

Collinearity assessment

Before measuring path analysis, it is crucial to check the issue of full collinearity among the latent variables Therefore, in the structural model, an assessment of full collinearity of all predictors was performed through variance inflation factors (VIFs) for detecting multicollinearity problems The results revealed that all values of VIF remained lesser than the required threshold ranging from 5 to 10 Hence, no issue of multicollinearity was found and this proved that collinearity has not affected the results of the structural model.

Note: SD, Standard deviation; VIF, Variance inflation factor; The items below the diagonal present correlation values between constructs; Items above the diagonal at HTMT values; Diagonal itemspresent square roots values of average variance explained (AVE) *p < 0.05 **p < 0.01.