Review status
This summary is based on the full article text.
The DOI and source URL have been verified through the official INFORMS article page.
The article is an empirical study published in Organization Science. It tests the ambidexterity hypothesis using survey data from manufacturing firms in Singapore and Penang, Malaysia. The study uses factor analysis, hierarchical regression, path analysis, ANOVA, and sensitivity analysis to examine whether explorative and exploitative innovation strategies jointly influence sales growth.
Research question
How do explorative and exploitative innovation strategies jointly influence firm performance?
More specifically, the article asks whether firms perform better when they pursue both explorative and exploitative technological innovation strategies, and whether imbalance between the two strategies is harmful for sales growth.
Hypotheses
The article tests four hypotheses.
Hypothesis 1a proposes that there is a positive interaction effect between explorative and exploitative innovation strategies on firm performance.
Hypothesis 1b proposes that relative imbalance between explorative and exploitative innovation strategies is negatively related to firm performance.
Hypothesis 2 proposes that firms specializing in explorative innovation strategy exhibit larger performance variation, relative to their mean performance, than firms specializing in exploitative innovation strategy.
Hypothesis 3 proposes that ambidextrous firms, defined as firms scoring high on both explorative and exploitative innovation strategies, exhibit smaller performance variation, relative to their mean performance, than firms specializing in explorative innovation strategy.
Method
The study uses survey data from manufacturing firms in Singapore and Penang, Malaysia, collected during 1999-2000.
The sampling frames came from the Economic Development Board of Singapore and the Penang Development Corporation. Questionnaires were sent to CEOs of 1,872 manufacturing firms in Singapore and 950 manufacturing firms in Penang. After removing incomplete, doubtful, or contradictory responses, the authors obtained 371 valid responses from Singapore and 192 valid responses from Penang. This produced response rates of 19.8% and 20.2%.
The analysis focuses only on innovating firms. A firm was classified as innovating if it had introduced a new or substantially improved product or a new or substantially improved production process in the previous three years. Using this threshold, the final sample consisted of 206 innovating manufacturing firms: 137 from Singapore and 69 from Penang, Malaysia.
The dependent variable is sales growth rate, measured as self-reported compounded average sales growth over the previous three years, from 1996 to 1999. The authors justify this three-year window because most firms in the sample had relatively short innovation cycles. Only 1.5% of firms reported average innovation project duration of more than three years, while 68.9% reported average project duration of less than one year. Similarly, 15.0% reported average payback periods of more than three years, while 66.0% reported payback periods of less than two years.
The authors validated the survey-based sales growth measure against archival data for a subsample of 90 firms. Survey-based sales growth correlated strongly with archival-based sales growth, r = 0.821, p = 0.000. It also correlated with return on sales, r = 0.351, p = 0.001; return on assets, r = 0.276, p = 0.018; ROS growth, r = 0.398, p = 0.000; and ROA growth, r = 0.418, p = 0.000.
Explorative and exploitative innovation strategies were measured with eight Likert-scale items asking how important different innovation objectives had been during the previous three years. Explorative innovation captured objectives such as introducing new generations of products, extending product range, opening new markets, and entering new technology fields. Exploitative innovation captured objectives such as improving existing product quality, improving production flexibility, reducing production cost, and improving yield or reducing material consumption.
Factor analysis reduced the eight items into two dimensions: explorative innovation strategy and exploitative innovation strategy. The two-factor solution explained 65% of the variance. Reliability was acceptable for both constructs: Cronbach’s alpha = 0.752 for explorative innovation strategy and 0.807 for exploitative innovation strategy. Confirmatory factor analysis supported discriminant validity between the two dimensions.
The study also includes two intermediary innovation performance variables. Product innovation intensity is measured as the percentage of total annual sales consisting of new or improved products introduced over the previous three years. Process innovation intensity is measured as the percentage of annual production volume using new or improved processes introduced over the previous three years. Natural logs of both measures were used to reduce skewness.
Control variables include firm size, firm age, location, export share, R&D spending intensity, ownership nationality, and technology class. The technology classes distinguish high-, medium-, and low-technology firms.
The authors use hierarchical regression and path analysis to test the ambidexterity hypotheses. They also use ANOVA to compare performance variation across four groups: no-emphasis firms, exploitative firms, explorative firms, and ambidextrous firms. White-heteroskedasticity-robust estimates are reported because the variance tests indicate heteroskedasticity. The authors also run Heckman’s two-stage procedure to check sample selection bias. The inverse Mills ratio is not significant, suggesting that sample selection bias is not a serious issue in the analysis.
Results / key findings
The study finds evidence consistent with the ambidexterity hypothesis.
First, explorative and exploitative innovation strategies affect innovation performance differently. Explorative innovation strategy is positively related to product innovation intensity. In the innovation performance regressions, the coefficient for explorative innovation strategy predicting product innovation intensity is significant across specifications, including 0.156, p = 0.002, and 0.124, p = 0.030 when R&D spending intensity is included. Explorative innovation strategy is not significantly related to process innovation intensity.
Exploitative innovation strategy is related to both product and process innovation, although the effect is stronger and more consistent for process innovation. For process innovation intensity, exploitative innovation strategy is positive and significant, with coefficients around 0.160, p = 0.016, and 0.191, p = 0.007. For product innovation intensity, the effect is weaker, with coefficients around 0.106, p = 0.055, and 0.097, p = 0.071 before R&D spending is added.
R&D spending intensity is positively associated with product innovation intensity, with coefficients around 0.032, p = 0.029, and 0.031, p = 0.038. It is not significantly related to process innovation intensity. This suggests that formal R&D spending in this sample is more closely tied to product innovation than to process innovation.
Second, the regression results support Hypothesis 1a. The interaction between explorative and exploitative innovation strategies is positively related to sales growth. In the main sales growth regression, the interaction coefficient is 4.539, p = 0.035. Adding the interaction term increases R² from 0.161 to 0.175. This supports the idea that exploration and exploitation can complement each other when pursued together.
Third, the regression results provide weaker but still supportive evidence for Hypothesis 1b. The absolute difference between explorative and exploitative innovation strategies is negatively related to sales growth. The coefficient is -3.026, p = 0.074. This suggests that imbalance between exploration and exploitation is associated with lower sales growth, although the evidence is weaker than for the interaction effect.
Fourth, path analysis supports the same broad conclusion. In the fit-as-moderating model, the interaction between explorative and exploitative innovation strategy has a positive path to sales growth, 0.146, p < 0.10. Explorative innovation strategy is positively related to product innovation intensity, 0.140, p < 0.05. Exploitative innovation strategy is positively related to product innovation intensity, 0.196, p < 0.05, and process innovation intensity, 0.179, p < 0.05. Product innovation intensity and process innovation intensity are both positively related to sales growth, with paths of 0.177 and 0.142 respectively.
The fit-as-moderating path model has acceptable fit: chi-square = 146.887, degrees of freedom = 107, p = 0.006, normed chi-square = 1.373, GFI = 0.933, CFI = 0.952, NFI = 0.861, and RMSEA = 0.046.
Fifth, the fit-as-matching path analysis also supports the imbalance argument. The absolute difference between explorative and exploitative innovation strategy is negatively related to sales growth, with a path coefficient of -0.167, p < 0.10. The model fit is also acceptable: chi-square = 141.599, degrees of freedom = 107, p = 0.014, normed chi-square = 1.323, GFI = 0.936, CFI = 0.961, NFI = 0.873, and RMSEA = 0.043.
Sixth, the ANOVA results support Hypotheses 2 and 3. Firms specializing in explorative innovation have the highest performance variation relative to mean performance. The explorative group has mean sales growth of 8.59, standard deviation of 21.15, and a standard deviation-to-mean ratio of 2.46. The exploitative group has mean sales growth of 9.96, standard deviation of 14.75, and a ratio of 1.48. The ambidextrous group has mean sales growth of 17.43, standard deviation of 27.64, and a ratio of 1.59. The no-emphasis group has mean sales growth of 10.13, standard deviation of 15.81, and a ratio of 1.56.
This pattern supports the article’s claim that exploration is riskier and produces more variable outcomes. Ambidextrous firms achieve the highest mean sales growth while avoiding the extreme variance-to-mean ratio seen among exploration-focused firms.
Seventh, the regression-based standard deviation analysis reinforces this interpretation. The actual standard deviation for the explorative group is above the upper bound of the 95% confidence interval, indicating unusually high variation relative to its mean. The exploitative and no-emphasis groups fall below the lower bound, indicating lower variation. The ambidextrous group’s actual standard deviation is within the 95% confidence interval and very close to its predicted standard deviation, indicating normal variation.
Eighth, the sensitivity analysis shows that ambidexterity has limits. When ambidexterity is defined using the median cut-off, the ambidextrous dummy is positively related to sales growth, 0.269, p = 0.008. When the cut-off becomes stricter, requiring firms to score in the upper 3/8 for both strategies, the effect remains positive but weaker, 0.193, p = 0.049. When the cut-off becomes very strict, requiring firms to score in the upper quarter for both strategies, the relationship becomes nonsignificant, 0.077, p = 0.363. The authors interpret this as possible evidence that pursuing both exploration and exploitation extremely aggressively can create organizational difficulties.
Ninth, the study finds that very low levels of both exploration and exploitation should not be treated as true ambidexterity. When the authors remove firms scoring in the lowest 15% on both strategies, the imbalance result becomes stronger. In the regression, the imbalance coefficient improves to -3.841, p = 0.064, and in the path analysis the effect improves to -0.214, p = 0.044.
Overall, the evidence supports the idea that exploration and exploitation can jointly improve performance when firms manage them as complementary innovation strategies. However, the article also warns that ambidexterity is not simply “high everything.” Extremely aggressive pursuit of both strategies may create coordination and organizational tensions that weaken the performance benefit.
Practical implications
For managers, the article gives a clear resource allocation lesson: exploration and exploitation should be managed together, not treated as mutually exclusive choices.
Exploration matters because firms need to enter new product-market domains, open new markets, and develop new technology fields. Exploitation matters because firms also need to improve quality, production flexibility, cost efficiency, and material yield. The study suggests that firms grow more strongly when they combine these logics rather than overcommitting to only one.
Managers should track innovation portfolios along exploration-exploitation dimensions. Traditional metrics such as product versus process innovation, basic versus applied research, or R&D spending intensity are useful but incomplete. He and Wong’s eight-item innovation strategy measure provides a useful starting point for asking whether innovation investments are aimed at new domains or improving existing positions.
The results also warn against overemphasizing exploration. Exploration-focused firms had the highest performance variation relative to mean performance. This does not mean exploration is bad. It means exploration carries more risk and should be combined with exploitation so that firms can benefit from new opportunities while maintaining operational and market discipline.
The ambidextrous firms in the sample had the highest average sales growth, 17.43, compared with 10.13 for no-emphasis firms, 9.96 for exploitative firms, and 8.59 for explorative firms. For managers, this suggests that the payoff comes not from avoiding exploration or exploitation, but from combining them effectively.
The sensitivity analysis adds a practical warning. When the threshold for ambidexterity became very strict, the performance relationship disappeared. This suggests that managers should not assume that maximum exploration plus maximum exploitation is automatically best. The harder challenge is finding a workable balance that the organization can actually absorb.
For practitioners, useful diagnostic questions include:
- Are innovation projects mainly aimed at new product-market domains or improving existing product-market positions?
- Does the innovation portfolio contain both explorative and exploitative initiatives?
- Is the firm overexposed to risky exploration without enough exploitation discipline?
- Is the firm overcommitted to exploitation and underinvesting in future opportunities?
- Are product innovation and process innovation both being measured?
- Does the organization have structures, routines, and leadership capacity to manage the tension between exploration and exploitation?
- Are managers trying to push both strategies so aggressively that coordination problems may outweigh the benefits?
Theoretical implications
The article contributes to research on organizational ambidexterity by providing direct empirical evidence that exploration and exploitation can be complementary in technological innovation strategy.
It extends March’s exploration-exploitation distinction into a measurable innovation strategy framework. Rather than treating exploration and exploitation only as abstract learning processes, the article operationalizes them as strategic priorities in technological innovation.
The article also contributes to dynamic capabilities thinking. The findings support the idea that firms need to combine different strategic logics to adapt and perform. Exploration supports movement into new domains, while exploitation strengthens current positions and improves efficiency.
The study also links ambidexterity to absorptive capacity. Exploration can expand the knowledge base, while exploitation can help firms use, refine, and commercialize what they know. The positive interaction between the two supports the idea that exploration and exploitation can form a mutually reinforcing learning cycle.
The article also contributes to innovation management by showing that exploration and exploitation influence product and process innovation differently. Explorative innovation strategy primarily affects product innovation, while exploitative innovation strategy affects both product and process innovation.
Finally, the study adds nuance to ambidexterity theory by showing both benefits and limits. Balance appears beneficial, but extremely high emphasis on both exploration and exploitation may create organizational difficulty. This supports a more careful view of ambidexterity as a delicate managerial capability rather than a simple prescription to do more of everything.
Limitations
The study uses survey data, so key measures depend on CEO self-reports and respondent recall.
The study is cross-sectional and cannot fully establish causality. It shows that ambidextrous innovation strategies are associated with sales growth, but it cannot prove that ambidexterity directly caused higher growth.
The study focuses on manufacturing firms in Singapore and Penang, Malaysia. The findings may not generalize fully to service industries, large global firms, digital platforms, public organizations, or firms in very different institutional settings.
The dependent variable is three-year sales growth. This time horizon fits the sample because most firms reported short innovation project durations and payback periods, but it may be too short for industries where exploration takes five to ten years to affect performance.
The operationalization of exploration and exploitation uses eight Likert-scale items. These items capture important innovation objectives, but they may not cover the full conceptual breadth of exploration and exploitation.
The study controls for broad technology classes and industry effects, but it does not include fine-grained measures of market dynamism, technological turbulence, competitive intensity, or organizational design.
The authors could not fully control for prior-period sales growth for the full sample. They checked archival data for 44 firms and found no significant correlation between prior and current sales growth, but endogeneity cannot be fully ruled out.
The study does not directly examine which organizational structures, leadership practices, or resource allocation routines allow firms to manage ambidexterity successfully.
Future research
Future research could test the ambidexterity hypothesis in other industries, countries, and institutional contexts.
Longitudinal studies could examine whether the balance between exploration and exploitation predicts firm performance over longer periods, especially in technology-intensive sectors where returns may take more than ten years to appear.
Future research could develop more detailed measures of explorative and exploitative innovation strategy, including digital innovation, platform development, sustainability transformation, artificial intelligence, and business model innovation.
Researchers could examine how the optimal balance between exploration and exploitation changes under different levels of market dynamism, technological turbulence, competitive pressure, and environmental uncertainty.
Another useful research direction would be to study organizational designs that make ambidexterity manageable, such as structural ambidexterity, contextual ambidexterity, semi-structured processes, innovation portfolios, and senior team integration.
Future research could also examine why extremely high levels of both exploration and exploitation may weaken the performance benefit. This could involve studying coordination costs, managerial overload, resource conflicts, cultural tensions, or strategic inconsistency.
Finally, future studies could connect exploration-exploitation balance to broader outcomes beyond sales growth, such as profitability, survival, innovation quality, patent impact, new product success, resilience, and long-term competitive advantage.