Antecedents of strategic ambidexterity in firms’ product innovation: External knowledge and internal information sharing

ABSTRACT The conflict between “faster, better and cheaper” and high environment uncertainty forces firms to choose between different product innovation strategies. This study shows that firms can develop strategic ambidexterity in product innovation (SAPI) through diverse product innovation strategies, by effectively exploiting internal and external resources as well as their combination. The work explores the relationship between the interplay of external knowledge sources and internal information sharing for SAPI in 600 small- and medium-sized Chinese manufacturing enterprises. The results show knowledge breadth and depth and information sharing are positively related to SAPI, which tends to increase when knowledge depth is associated with information sharing. The study contributes to new product development (NPD) literature by introducing the concept of strategic ambidexterity in product innovation, and exploring the roles of external knowledge sources and internal information sharing, and their complementarity, on the ability to develop firm-level SAPI. It further provides implications for open innovation studies.


Introduction
Product innovation plays a central role in a firms' competitiveness.However, the high failure rates for firms' internal new product development (NPD) (Kuester et al., 2012) and the high external uncertainty of the business environment (Bingham et al., 2014) pose challenges for firms in translating product innovation into performance.Appropriate strategies for product innovation thus become vital for firms pursuing competitive advantage (e.g., Cankurtaran et al., 2013;Shinkle & McCann, 2014), such as time-based strategies 1 (Cankurtaran et al., 2013;Stanko et al., 2012), market launch and related entry strategies (Kuester et al., 2012), and product quality strategies (Tellis & Johnson, 2007).
However, potential trade-offs may arise, particularly when a firm promotes these strategies simultaneously.Internally, one such trade-off ensues from the conflicts between different product innovation strategies (PIS).For example, a shorter NPD time does not always imply higher product quality or lower development costs (Tellis & Johnson, 2007).Therefore, managers need to decide among product information strategies by exploring an optimal balance of "faster, better and cheaper" (Stanko et al., 2012;Swink et al., 2006).Externally, another trade-off may arise due to the environment uncertainty surrounding the firm regarding opportunity capture, which calls for managers to make appropriate strategic choices in products to exploit opportunities in the environment (Bingham et al., 2014;Nadkarni & Narayanan, 2007).
Prior literature has adopted a paradoxical lens to address these trade-offs, originally tracing back to the metaphor story The Hedgehog and The Fox 2 (Berlin, 2013).For the internal trade-off, two alternative scope strategies have been highlighted for firms (Carroll, 1985;Carroll & Swaminathan, 2000): "generalist" firms can rely on a variety of new product segments by broadening resource bases, while "specialist" firms tend to capitalize on accumulated resources with a reduced number of NPDs (Barroso & Giarratana, 2013;Jourdan & Kivleniece, 2017).For the external trade-off, two types of strategic choices fit the fast-changing environment to capture the opportunity: "opportunist" firms tend to retain strategic variety for flexibility while "strategist" firms are more disciplined in following a strategic focus (Bingham et al., 2014;Nadkarni & Narayanan, 2007).
The paradoxical lens of firms addressing trade-offs has one basic implicit assumption regarding resource restriction, leading to a firm's inability to coordinate a wide range of different product information strategies in a tradeoff (Voss & Voss, 2013).This issue is particularly important in small and medium size enterprises (SMEs), as they face resource scarcity, unsystematic innovation activities, and inferior capabilities, which result in higher failure rates of innovations (e.g., Van de Vrande et al., 2009;Vanhaverbeke, 2017;Wynarczyk et al., 2013).However, the literature underemphasizes that a firm's resource restrictions could be alleviated, leading to the need to revisit the binary strategic choice of managers.Externally, many firms have sourced knowledge and resources by opening their traditional organizational 1 Cankurtaran et al. (2013) summarize similar concepts related to time-based strategies on new product development such as time-to-market, innovation speed, lead time, innovation time, product development time, project completion time, and total time.
boundaries (Chesbrough, 2003;Laursen & Salter, 2006), particularly when firms are embedded in an environmental resource-munificence condition such as Industry 4.0 (Amezcua et al., 2013).Internally, firms with strong organizational capabilities can structure, bundle and leverage resources effectively to achieve their strategic purpose of NPD (Sirmon et al., 2007(Sirmon et al., , 2011)).Further, the potential interactive and integrating effects between external knowledge and internal capabilities for product innovation have also been considered (e.g., West & Bogers, 2014).
We accordingly contribute to the paradoxical lens conversation by contending that an additional way to tackle the trade-offs is to adopt both strategic variety and intensity among different product information strategies at the same time.Firms that do so develop what we term "strategic ambidexterity" in product innovation.We further surmise that firms can develop strategic ambidexterity in product innovation (SAPI) as long as they are able to build on their internal and external resources.For SMEs, operating with multiple NPD strategies it is conducive to increasing their success rate.Correspondingly, external and internal resources used for one single purpose could be adapted for new strategic applications extending the potential array of outputs that attenuate resource constraints (Voss & Voss, 2013).Therefore, vigorous firms are not only likely to choose a generalized strategy to improve efficiency, but also emphasize the development of specialization.
Therefore, we generate our main research question: What are the antecedents of a firm developing SAPI and how do these antecedents matter to the firm's SAPI?We contribute to the NPD literature by developing SAPI and further exploring both internal and external antecedents of SAPI, particularly in the context of SMEs, which revisits the resource restriction assumption on firms dealing with strategic trade-offs (e.g., Stanko et al., 2012;Voss & Voss, 2013).Moreover, the construction and antecedents' exploration of SAPI brings new knowledge to the NPD literature by extending the focus from product innovation performance (e.g., Cankurtaran et al., 2013;Kuester et al., 2012) to SAPI through the pursuit of diverse product information strategies.The literature reports only limited evidence on guiding choices in firms' different product information strategies (Acur et al., 2010), and particularly regarding the underexplored interaction effects between internal and external antecedents on product innovation strategic choice (Vega-Jurado et al., 2008).To address this gap, we combine literature on NPD, open innovation and the resource orchestration view, and make contributions by exploring the roles of external knowledge sources, internal organizational capabilities on information sharing, and their complementarity, on the firm-level SAPI.Moreover, we contribute to open innovation literature by extending the openness measure (involving the breadth and depth from Laursen & Salter, 2006) to the NPD context of product information strategies, which represents a novel strand of the existing product innovation studies (e.g., Acur et al., 2010;Cankurtaran et al., 2013), and highlights the role of organizational capabilities (information sharing) beyond the R&D capabilities that were skewed for product innovation in prior studies (e.g., Vanhaverbeke et al., 2014).

Strategic ambidexterity in product innovation
The term "ambidexterity" has been used to indicate a firm's ability to pursue two different initiatives simultaneously: exploitation and exploration in its original definition of manufacturing efficiency and flexibility, and induced and autonomous strategy processes (for a review, see, Gibson & Birkinshaw, 2004).We accordingly adopt a similar terminology -strategic ambidexterity in product innovation (SAPI).
Previous literature has highlighted how firms rely on different strategies for successful product innovations, involving time-based strategies (for example, Cankurtaran et al., 2013), intensive market launch and related penetration strategies (Kuester et al., 2012), superior product quality strategy (Tellis & Johnson, 2007), and customer orientation strategy (Lynn & Akgün, 1998).However, considering the potential incompatibility of the "faster, better and cheaper" features of NPD goals (Swink et al., 2006), uncertain and fastchanging environments (Bingham et al., 2014), and firms' scarce product innovation resources (Barroso & Giarratana, 2013), managers largely need to make a choice from the different product information strategies (Stanko et al., 2012).
The existing literature illustrates the above trade-offs in adopting different product information strategies.For example, some studies consider the tradeoff between variety and intensity of horizontal product strategies under resource scarcity (Barroso & Giarratana, 2013;Combs et al., 2011), while others highlight the trade-off between strategic flexibility and strategic focus due to market uncertainty (Bingham et al., 2014;Nadkarni & Narayanan, 2007).Overall, there seems to be an inherent difficulty in combining a high number of product information strategies, due to an apparent incompatibility between exploration and exploitation (for example, Gibson & Birkinshaw, 2004), and the strategic concerns between variety and intensity in product segments (Barroso & Giarratana, 2013;Jourdan & Kivleniece, 2017).As suggested by Combe et al. (2012), variety should be balanced with stability in strategic focus, and the friction between strategic variety and strategic intensity is a key strategic challenge given the resource scarcity (Brozovic, 2018;Combe et al., 2012).In contrast, recent contributions highlight the possibility for complementarity between different strategies, pointing to building ambidexterity for firms (for example, Z. Wei et al., 2014).Extending the ambidexterity view to the strategic concerns in NPD contexts, we define SAPI as the ability of firms to carry out strategic variety and intensity among a wide number of product innovation strategies at the same time.Notably, we expect firms to develop SAPI the more they expand efforts for external resources via open innovation (Chesbrough, 2003) or when firms build internal organizational capabilities to orchestrate resources properly (Sirmon et al., 2007(Sirmon et al., , 2011)).

Knowledge sources and strategic ambidexterity in product innovation: Open innovation
Open innovation is defined as "a paradigm that assumes that firms can and should use external ideas as well as internal ideas, and internal and external paths to market, as firms look to advance their technology" (Chesbrough, 2003, p. xxiv), and where knowledge flows purposively across firms' boundaries via outside-in, inside-out, or coupled modes (Enkel et al., 2009).Given resource scarcity, firms, particularly SMEs, must search for, identify and use diverse external knowledge to follow through strategy and innovation (Laursen & Salter, 2006).By accessing knowledge sources, firms expand their resource base and create an open resource repertoire to extend horizontal product strategies (Barroso & Giarratana, 2013), balance diverse product information strategies and alleviate uncertainty (Jourdan & Kivleniece, 2017).
Previous studies have identified heterogeneous external knowledge sources including suppliers (Li & Vanhaverbeke, 2009), competitors (Lim et al., 2010), customers, andusers (Von Hippel, 2001), government agencies (Hewitt-Dundas, 2006), universities (Rajalo & Vadi, 2017), and intermediaries (Y.Zhang & Li, 2010), among others.These studies used two concepts -knowledge breadth and knowledge depth -to represent how and the extent to which a firm purposively uses external knowledge in innovations (Laursen & Salter, 2006;Love et al., 2014).Knowledge breadth is the quantitative characterization of external knowledge sources, indicating the number of the focal firm's diverse partners, and knowledge depth is the qualitative characterization of a firm's external knowledge links with specific partners.
Recognizing the inherent complexity of NPD and environment uncertainty, we contend that external knowledge sources are needed for firms to promote SAPI.With the increase in number of diverse knowledge sources, a firm has more opportunities to access useful and valuable external knowledge on different product innovation strategic activities (Love et al., 2014).In addition, a firm with diverse external knowledge sources can promote resource deployment in multiple market settings without losing its internal basis (Combs et al., 2011), bring flexibility to the mix of product information strategies (Lant et al., 1992), and reduce the likelihood of routine dependence or organizational inertia (Nadkarni & Narayanan, 2007).When facing environmental change, a firm with broad knowledge can bundle diverse knowledge sources flexibly to improve adaptation and fit between the knowledge breadth and the combined product information strategies for competitive advantage (Bingham et al., 2014;Sirmon et al., 2007).Therefore, we propose: Hypothesis 1: The breadth of external knowledge sources used in firm innovation strategies is positively related to strategic ambidexterity in firm product innovation.
Knowledge depth is also important for a firm implementing SAPI when searching for external knowledge.First, knowledge depth builds the rigid resources for firms' specific product information strategies that are valuable in a narrow set of strategic markets (Combs et al., 2011).Unlike the diverse knowledge sources drawn from to fit the combined product information strategies, managers are inclined to transform unique knowledge sources into a specific product development process to increase the relevant success rate (Laursen & Salter, 2006).Second, a firm learns and absorbs knowledge from partners by promoting knowledge depth.Specific knowledge is situationdependent and related to specific partners, with which the firm establishes long-term interactions and a basis of shared learning and trust (Laursen & Salter, 2006).Within intensive interactions, a firm identifies valuable knowledge sources by long-term learning and screening, and acquires suitable knowledge as crucial assets for product innovation (Nadkarni & Narayanan, 2007).Third, prior studies have demonstrated the preference of innovative firms to focus on knowledge depth derived from a small number of sources (Laursen & Salter, 2006), such as steering valuable resources and concentrating external attention.This reduces the possibility of misunderstanding, erroneous interpretation and misapplication of external knowledge sources impacting product innovation (Luca & Atuahene-Gima, 2007), accelerates flow and transformation of external knowledge sources necessary for product information strategies and their portfolios, and maintains robustness and stability by strategic consistency (Combe et al., 2012).All these encourage firms to promote product information strategies via the "stick to their knitting" approach (Combs et al., 2011).Therefore, we propose: Hypothesis 2: The depth of external knowledge sources used in firm innovation strategies is positively related to strategic ambidexterity in firm product innovation.

Information sharing and strategic ambidexterity in product innovation: Resource orchestration
Internal firms' organizational capabilities matter for SAPI, particularly when firms do well in resource orchestration.The resource orchestration view states that possessing valuable, rare, inimitable and nonsubstitutable resources (Barney, 1991) cannot guarantee competitive advantage, and firms further need to orchestrate these resources effectively (Sirmon et al., 2011).This involves structuring the portfolio of resources, bundling resources to build capabilities, and leveraging and enriching resources to develop firms' managerial capabilities to facilitate innovations (Sirmon et al., 2007).
Prior literature has highlighted the importance of the synergy of different product information strategies (Sirmon et al., 2011), involving speed, quality, andcost (Jiménez-Jiménez &Sanz-Valle, 2011).To effectively manage diverse product information strategies, appropriate organizational capabilities are necessary (Alegre & Chiva, 2013), referring to a firm's internal organizational management functions like "administrative styles, the formalization of internal communication systems, and the interdependence of work teams" (Vega-Jurado et al., 2008, p. 617).Further, the development of organizational capabilities for resource orchestration become even more critical in the presence of environment uncertainty and resource limitations (Sirmon et al., 2007).
Basically, a firm can be depicted as an information-process network combining different practices for information interactions (Luca & Atuahene-Gima, 2007).Information sharing, referring to the open culture for a firm facilitating cross-functional interactions (Song & Montoya-Weiss, 2001), is an important organizational capability for a firm's SAPI.First, information sharing creates an intraorganizational common understanding conducive to product innovations (Wu, 2008).Firms with better internal information sharing have transfunctional strategic beliefs, behaviors, and operational processes, helping their employees to better perceive and understand the innovative experiments on product innovation collectively (Y.Wei et al., 2013).This in turn enhances the joint problem-solving activities, encourages the development of common commitment to execute diverse product information strategies in one firm interactively (Acur et al., 2010;McEvily & Marcus, 2005), and reduces task conflicts among diverse groups in NPD contexts (Moye & Langfred, 2004).Second, information sharing facilitates the spread of intraorganizational knowledge for product innovation (Lin et al., 2013).Prior literature has demonstrated the positive effects of knowledge dissemination among a firm's employees and business units on both new product introductions (Darroch & McNaughton, 2002) and new idea generation for product development (Sáenz et al., 2012).By orchestrating vertical and horizontal information flows, a firm creates an unimpeded network for effective knowledge synthesis for product innovation among departments and between hierarchical managerial levels (Sirmon et al., 2011;Wooldridge et al., 2008).Finally, information sharing shapes an innovative climate for firm members' collective risk-taking activities (Acur et al., 2010).Highly innovative product strategies are accompanied by high technological and market uncertainty, potentially challenging the existing experience of firm members (Lau et al., 2010).Enhancing innovative climate and risk-taking commitment encourages the high involvement of a firm's product innovation process (Acur et al., 2010;Y. Wei et al., 2013).Therefore, we propose: Hypothesis 3: The extent of information sharing within a firm is positively related to strategic ambidexterity in firm product innovation.

External knowledge sources and internal information sharing
We expect the combination of external knowledge sources and internal information sharing to encourage SAPI, and find support for the complementary effect.According to the open innovation literature, sourcing both diverse and specific knowledge assets is central to focus the firm's overall strategic purpose and improve the strategic portfolio in product innovation (Laursen & Salter, 2006).However, this is not enough to guarantee the value capture from the outside-in processes, as complementary internal factors are needed to effectively facilitate the use of knowledge sources (West & Bogers, 2014).The resource orchestration view supports that firms willing to develop SAPI should be able to effectively orchestrate resources (Sirmon et al., 2011), and the practice of information sharing accordingly is one of the most important complementary mechanisms facilitating not only the direct knowledge assimilation and transformation in outside-in processes, but also the integration and reconfiguration of external and internal knowledge resources (Lin et al., 2013).The open culture that information sharing contributes to a firm helps overcome the "not invented here" syndrome by adopting external knowledge appropriately for product information strategies (Chesbrough, 2003), thus reinforcing the positive association between knowledge breadth and depth and SAPI.
Regarding the external knowledge breadth, it first creates an abundant knowledge asset base for firms' internal resource orchestration for SAPI through the pursuit of diverse product information strategies.By leveraging the intraorganizational common understanding (Wu, 2008), facilitating knowledge dissemination (Lin et al., 2013) and constructing an innovative climate (Acur et al., 2010) for product innovation, information sharing encourages the collective participation and common involvement in these inbound innovation-resourcing processes (Y.Wei et al., 2013;Wooldridge et al., 2008).Accordingly, it leverages the potential alignment of external diverse knowledge resources to the different product information strategies (Sirmon et al., 2011), which enables strategic flexibility and consistency of product innovation simultaneously (Brozovic, 2018), as resources emphasizing change may influence strategic variety whereas resources valuing discipline may affect strategic intensity (Combs et al., 2011).Moreover, the internal information sharing further enables firms to combine both diverse external knowledge and internal resources (e.g., ideas, processes, and knowledge) in novel configurations in different NPD contexts (Markovic & Bagherzadeh, 2018).Firms that foster intraorganizational information sharing ensure that the retrieved and absorbed diverse external knowledge sources can be assigned, structured and bundled to different business units (Foss et al., 2011), which improves the orchestration of external resources and internal capabilities for appropriate product innovation strategy portfolios (Zobel, 2017).Therefore, we propose: Hypothesis 4: The interaction between the extent of information sharing within a firm and the breadth of external knowledge sources used in firm innovation strategies is positively related to strategic ambidexterity in firm product innovation.
The combination of external knowledge depth and internal information sharing also improves SAPI.First, the embeddedness of interfirm interactions with a small number of knowledge sources can be seen as a firm's strategic resources (Crook et al., 2008), and information sharing enables the characteristics of these identified resources to be more understandable and assimilable by the firm's employees (Y.Wei et al., 2013).A firm, therefore, can deliberately reduce the intraorganizational knowledge asymmetry toward the specific external resources and encourage interdepartmental collaborations to help avoid the confusion of product flows (Lau et al., 2010) in order to improve the synergy among different product information strategies (Sirmon et al., 2011).In addition, the information sharing facilitates the effective bundling of internal resources with the identified external knowledge.As information sharing allows the absorbed identified external knowledge to spread across the firm (Foss et al., 2011), the firm could allocate the appropriate internal innovation activities to fit the specific external knowledge resources (Zobel, 2017) to take advantage of bundling and leveraging of combined external and internal assets for product innovations (Lau et al., 2010).This kind of combined benefit can be further accumulated regarding the deepening interactions with specific external knowledge sources (Nadkarni & Narayanan, 2007) and the thickening of the firm's orchestrating capabilities (Sirmon et al., 2011).Therefore, we propose: Hypothesis 5: The interaction between the extent of information sharing within a firm and the depth of external knowledge sources used in firm innovation strategies is positively related to strategic ambidexterity in firm product innovation.
Figure 1 shows the conceptual framework.

Data
We tested our hypotheses in the empirical context of SMEs in China's Zhejiang Province.China provides an appropriate setting due to its complex economies in dynamic transition (Zhou & Wu, 2010), with SMEs accounting for more than 99% of all firms contributing to China's economic growth and employment (Zeng et al., 2010).Zhejiang Province is one of the fastest-growing regions in China, experiencing rapid growth and ranked first in income per capita in China for 21 consecutive years. 3Zhejiang Province's industrial structure is a combination of traditional manufacturing industries (such as textiles and machinery) and hightech industries (such as internet, and bio instruments and equipment).
The sample was initially drawn from a survey administered to the Directory of Zhejiang Manufacturing Enterprises via post by the Zhejiang Economic and Information Commission between September and December 2013.One thousand manufacturing enterprises with total annual revenue less than 400 million RMB 4 were randomly selected from the directory of SMEs compiled by the provincial government.We mainly used perceptual measurements in the survey because firms are usually reluctant to share objective data due to confidentiality issues (Ward et al., 1996), and subjective survey data was used to compare firms from different industries (Ledwith, 2000).
The questionnaire was originally developed in English and then translated into Chinese.To avoid cognitive misunderstanding for conceptual equivalence (Hoskisson et al., 2000), we asked two researchers to carry out the translations independently and then made comparisons until consensus was reached.For content and face validity, we conducted face-to-face interviews with five senior managers for their corrections on the questionnaire, enhancing the relevance and completeness of the items (Zhou & Wu, 2010).The questionnaire had two main parts.The first part was on the firm's innovation activities (for example, different innovation strategies) and general demographic information (such as size and foundation year), and the second part had questions on organizational capabilities and external knowledge sources.To minimize problems arising from common method bias (Podsakoff et al., 2003), we assigned each part to two different respondents (one senior manager and one operational manager) in each firm, increasing response accuracy due to a shorter questionnaire (Y.Zhang & Li, 2010).The two respondents from each firm were selected based on criteria such as long tenure in the firm, and engaging in or being responsible for core businesses involving the firm's product innovation, organizational management, and external collaboration.
We obtained 711 preliminary responses from the sample of 1,000 manufacturing firms.After removing 111 questionnaires containing several missing data items, our final sample had full information for 600 firms.This size greatly exceeds the common recommendation of a large sample for advanced statistical analyses (MacCallum et al., 1999).Of the firms, 34.7% were in the light industry, 26.7% in the machinery industry, 13.7% in the textile industry, 13.3% in the electronics equipment industry, 6.7% in the bio instruments and equipment industry, and 5.0% in the hardware and software equipment industry.
A high response rate (in our case 60%) is commonly acknowledged as a good sign against nonresponse bias (Armstrong & Overton, 1977, p. 396).We further investigated possible nonresponse bias by comparing the size and age of participating (600) and nonparticipating firms.A one-way ANOVA analysis of variance showed no significant difference for firm size (F-statistic = 0.944; Sig.= 0.332) and age (F-statistic = 2.011; Sig.= 0.156).We also compared the same characteristics for the early 10% (60 firms) and the late 10% of respondents (Armstrong & Overton, 1977) and found no significant difference in firm age (F-statistic = 1.081;Sig.= 0.301) or size (F-statistic = 0.025; Sig.= 0.875).
Since the survey was completed by two respondents per firm, potential common method bias (CMB) problems were largely controlled for (Podsakoff & Organ, 1986).We applied Harman's single-factor test as a further check of CMB.Results show the first and most important factor explains 39% of the total variance, lower than the 50% threshold (Podsakoff et al., 2003), indicating the absence of CMB.In addition, as one of our interaction hypotheses was supported, we further argue the absence of CMB, as it is improbable for respondents to hold an "interaction-based theory" mindset that systematically biases their responses to produce these interaction effects (Aiken et al., 1991).
To check respondents' self-assessment of knowledgeability (Kumar et al., 1993), we adopted a seven-point Likert-type scale, anchored at "1 -a little involved" to "7 -heavily involved."The mean responses were 5.88 for senior managers and 5.84 for operational managers (with no significant difference: t = 1.901, p < .05).Both respondents for each firm were asked one common question on whether the firm exported products from China.The correlation test between their answers was checked for consistency (Nunnally & Bernstein, 1994;Y. Zhang & Li, 2010).The resulting correlation value of 0.913 (significant at p < .01)indicated good consistency.

Dependent variables and estimation method
We examined the relationship between leverage of external knowledge sources, internal organizational capability, and their interplay on the ability of the firm to combine different product information strategies simultaneously, namely SAPI.The respondents were asked to rate the importance of the different product information strategies they adopted during the period 2010 to 2013 on a seven-point Likert-type scale anchored at "1 -Not at all important" to "7 -Extremely important."Items relating to a broad number of different product information strategies were constructed (Acur et al., 2010;Carson et al., 2012): (1) frequent introduction of new products (Intensive launching PIS); (2) being first to introduce new products (First-mover PIS); (3) timely introduction of new products into the market (Time-to-market PIS); (4) develop and introduce products of the highest quality (Differentiation PIS); (5) using new products to penetrate markets (Market penetration PIS), and ( 6) introduce new products that meet customers' needs well (Target PIS).The main dependent variable was then created following Bozeman & Gaughan (2007) and Tartari et al. (2014), by computing an index combining both variety and intensity of product innovation strategies (PIS index) as follows: where i is the firm, s is the type of product innovation strategy (with s taking an integer value between 0 and 6), and r s is the rating received by the type of product innovation strategy s in the Likert-type scale; p s is the frequency of firms' use of each type of product innovation strategy in the sample and is defined as: where f s is a dummy variable taking the value of 1 if the firm considers the product innovation strategy s as relevant (by assigning it a value of 5 or higher) and 0 otherwise; n is the firm and N is the total number of firms in our sample (N =600).As highlighted in Tartari et al. (2014), this measure has the attractive feature of considering the "difficulty" of certain activities by incorporating the frequency of product innovation strategy nonoccurrence (1-p s ).
A high PIS index reflects reliance on a wide range of strategies, thus proxying well for our theoretical construct of SAPI.The scores for this measure range from 2 to 10.5, following a close to normal distribution.The type of information provided by this measure is illustrated in the following two contrasting cases from our sample: a firm with a score of 2.03; and a firm with a score of 10.48.The low-score firm just relies on two product information strategies as it deems the others not relevant.Moreover, the two adopted strategies are seen as low importance to the firm as it rated both of them 2 out of 7. The second example refers to a firm with a portfolio of product information strategies distributed across all six possible PIS types.In this case the adopted strategies were all rated as extremely important with a score of 7 out of 7.
Our dependent variable is restricted both by lower (zero is the minimum number of product information strategies available) and upper bounds (six is the maximum number of strategies available).Following Papke and Wooldridge (1996) and Laursen and Salter (2014), we normalize the PIS index to vary between 0 and 1 and apply a fractional logit regression estimation as follows: where the PIS index is the variable constructed following the procedure above, 5 Θ(.) is the logistic function, x is the vector of key explanatory variables for firm i (see the following section), and z is the vector of controls for firm i (see section on control variables below).

5
We also adopted a more standard PIS measure as the sum (average) of scores for each type of strategy and obtained results with no qualitative difference.

Explanatory variables
For our key explanatory variables, we capture the breadth and depth of external knowledge sources in the firm's innovation strategy based on prior literature (Laursen & Salter, 2006).We use the following 14 external knowledge sources: suppliers of equipment, materials, or components; clients or customers; competitors; consultants; universities or other higher education institutes; private research institutes; professional conferences, meetings; trade associations; technical and trade press; fairs and exhibitions; joint ventures; government agencies; intellectual property organizations; and lead users.
Knowledge breadth takes a value from 0 to 14 based on the number of knowledge sources for firm innovations.Knowledge depth ranges from 0 to 14 based on the number of knowledge sources the firm rates as highly important.Both measures show a good degree of internal consistency with Cronbach's alpha of 0.87 for Knowledge breadth and 0.91 for Knowledge depth.The extent of correlation between the two measures is in line with prior studies such as Laursen and Salter (2006), supporting the expectation that they measure two different empirical constructs.
Our second set of explanatory variables refers to the importance of a firm's information sharing, which is regarded as a typical organizational capability.Previous studies do not provide a unified scale to define organizational capability involving information sharing, but rather use different constructs to emphasize specific parts of the concept (Alegre & Chiva, 2013;Vega-Jurado et al., 2008;Xu et al., 2012;Yam et al., 2011).For this reason, instead of relying on an existing set of measures, we adopt Churchill's (1979) framework and develop our measures following the procedure below.First, we decided on a pool of items relating to organizational practices that subtend organizational capability based on prior literature.Next, we asked industry experts to review and modify the selected items.The revised set of items was included in the questionnaire to rank their importance level.
We ran factor analysis on the 18 different items to synthesize the common factor information underlying organizational capability, and obtained three predicted factors: Information sharing, Lean management, and Training and Learning. 6,7Previous literature facilitates the interpretation of these three constructs (Souitaris, 2002;Templeton et al., 2002;Xu et al., 2012).The first factor is labeled Information sharing and proxies for the ability of the firm to share knowledge within the organization (Y.Wei et al., 2013).The second group, Lean management, contains a range of items relating to attempts to streamline the organization's processes to reduce or eliminate non-value-

6
The factors were obtained on the basis of three diverse factor extraction methods including the Iterated Principal Factors Method, the Maximum Likelihood Method, and the Principal Components Method.Details and results of factor extraction are available in the online supplemental material.7 For a robustness check, we also used the average value of the items entering each factor instead of the predicted factor and obtained robust results.
added activities via an assessment of processes and projects within the organization (Narasimhan et al., 2006).The third group is named Training and learning, such as upskilling through specialized training or practices to favor learning among peers (Song et al., 2003).We used the factor Information sharing as the main regressor and retained the other two factors Lean management and Training and learning as controls, not only regarding our theoretical consideration, but also because Information sharing is the first and most important explanatory factor. 8These three measures show good degrees of internal consistency with Cronbach's alpha greater than the threshold value of 0.7 (Kline, 2015): 0.92 for Information sharing, 0.92 for Lean management, and 0.89 for Training and learning.We implemented a set of analyses to assess the construct validity of organizational capability.First, we ran a confirmatory factor analysis on items, showing that all factor loadings are statistically significant (p < .001)with values for the standardized loadings greater than 0.5 indicating the unidimensionality of our measures (Anderson & Gerbing, 1988).Next, we checked convergent validity by computing the composite reliabilities (CR) and average variance extracted (AVE) for the three constructs of organizational capability (Fornell & Larcker, 1981): for information sharing CR=0.87 and AVE=0.43; for Lean management CR=0.88 and AVE=0.58; and for Training and learning CR=0.85 and AVE=0.59.In all of the cases the CR was greater than 0.8 and the AVE was close to the threshold value of 0.5, showing adequate reliability and convergent validity.Finally, to evaluate the constructs' discriminant validity, we compared the unrestricted model to the restricted model 9 by fixing a correlation value of one between each pair of constructs.Our constructs showed good discriminant validity as the unrestricted model had better fit and the chi-squared test was not rejected at standard confidence levels.As the absolute values related to the three constructs of organizational capability were all smaller than the square roots of their AVE, we further demonstrated good discriminant validity (Fornell & Larcker, 1981).

Control variables
We minimized omitted variable bias by including a large set of controls.We first controlled for the above predicted two factors of organizational capability, Lean management and Training and learning.Lean management was regarded as an organizational resources bundle for product value-chain improvement that contributes to diverse product information strategies (Narasimhan et al., 2006), such as time-to-market acceleration, manufacturing cycle time 8 Information sharing explained 56.16% in principal component factor analysis, 54.03% in maximum-likelihood factor analysis, and 68.5% in iterated principal factor analysis.9 The model is composed of the three constructs of organizational capability and the one construct of technological capability measured by a Likert-type scale.
reduction, productivity leveraging, quality security, and manufacturing costs control (White et al., 1999).Training and learning was seen not only as a valuable resource for motivated employees to implement a firm's strategies (Chadwick et al., 2015), but also contributed to teamwork cooperation that positively influences product innovation (Collins & Smith, 2006).We controlled for the firm's technological capabilities to further capture the complex nature of the firm's product innovation profile as in prior literature (Vega-Jurado et al., 2008;Zhou & Wu, 2010).Following Zhou and Wu (2010), we evaluated the ability of the firm to rely on various technologies via factor analysis on five different items of technological capabilities, involving (1) Acquiring relevant technological information; (2) Identifying new technological opportunities; (3) Responding to technological change; (4) Mastering state-of-art technologies; and (5) Developing a series of innovations constantly.The resulting predicted factor was then included as a control variable. 10,11 The extent of Research and Development (R&D) investment has been shown to be relevant to product innovation not only as a provider of knowledge inputs for a firm's exploration activities (Love et al., 2014), but also as a key proxy for its absorptive capacity central to external knowledge adoption (Cohen & Levinthal, 1990).We thus included R&D intensity as a control variable to avoid potential confounding effects.As it is difficult to obtain exact R&D investment data from China's manufacturing firms (Von Zedtwitz, 2005), we used the proportion of employees employed in R&D activities within the firm.
We considered a set of characteristics that might influence innovation resources, incentives and engagement in different product information strategies: export orientation (Export oriented), measured as a dummy variable equaling 1 if the firm exported products abroad and 0 otherwise (Y.Zhang & Li, 2010), and firm size (Log Size) measured as the natural logarithm of the number of employees in the reference period (+1).Human capital also plays an important role in product innovation (for example, Kato et al., 2015), and we included a dummy, Human capital, taking the value of 1 when the CEO respondent of the firm has a bachelor's degree or above and 0 otherwise (Kato et al., 2015).We also included among the controls (natural log transformed and +1) firm's age (Log Age), relating to experience in developing product innovations (for example, Schneider & Veugelers, 2010).
We also included five dummies for the effect of industry characteristics on the abovementioned six industries.Finally, we considered confounding effects stemming from the geographical specificities of the area where the firm operates.We controlled for this by including a set of 10 geographical dummies at the city level of aggregation.
Table 1 reports descriptive statistics and a correlation matrix for variables.In general, correlation among the independent variables was low, and the variance inflation factor ranged between 1.04 and 2.23 (well below the threshold value of 5), suggesting the absence of multicollinearity problems.

Results
In Table 2, we examine the relevance of the ability to combine different product information strategies (namely SAPI) across sectors.We also examine the level of knowledge breadth and depth, internal organizational capabilities, and technological capability in each industry.The average value of our PIS index is 0.699 (with a maximum possible value of 1), mainly driven by an above average value of 0.738 for the machinery industry, and values below the average for the light, textile, electronics, bio equipment and hardware and software industries.For external knowledge sources, although firms in our sample tend to use more than nine external knowledge sources for innovation (the average value of Knowledge breadth is 9.428), they draw from only about two sources in an intensive way (the average value of Knowledge depth is 2.263).For industry differences, the machinery industry by far has the highest level of external search breadth and depth as well of the product information strategies index, while the hardware and software industry shows a good level of adopting knowledge breadth (the value is above the average).Moreover, different industries rely on different combinations of organizational capabilities.For example, the electronics and machinery industries have higher levels of in-house information sharing, and the electronics, bio equipment, hardware and software, and machinery industries ascribe more importance to lean management.Conversely, the bio equipment, machinery and electronics industries rely more on training and learning activities.Finally, the bio equipment, machinery and light industries exhibit higher levels of technological capability.
Table 3 reports the main results, presenting the fractional logit regression estimate for our proxy of SAPI.Model 1 is the baseline model including the set of all control variables only.Model 2 further adds our main explanatory variables.For the controls, Model 1 shows that the two types of organizational capabilities -Lean management and Training and learning -are positively associated with firms' ability to combine different product information strategies, in line with prior literature regarding them as valuable organizational resources for product innovation (Chadwick et al., 2015;Narasimhan et al., 2006).Model 1 further indicates that technological capability is significantly positively correlated with the ability to combine different product information strategies.This resonates well with the strategic   Fractional logit regressions with robust standard errors in parentheses.The dependent variable in all columns is a measure combining variety and intensity in product innovation strategies and developed following the methodology of Bozeman & Gaughan (2007) and Tartari et al. (2014).Significance levels: *p < .05,**p < .01,***p < .001.
management literature showing a complementarity between strategic flexibility and technological capability (Zhou & Wu, 2010).Similarly, an exportoriented firm is more likely to rely extensively on a higher number of product information strategies.Interestingly, firm size, human capital and R&D intensity are not relevant.Model 2 in Table 3 shows the main focused results: the effect of external search breadth and depth and organizational capabilities on the product information strategies index.Knowledge breadth and Knowledge depth both exhibit significantly positive relationships with the PIS index (β = 0.015, p < .01 and β = 0.021, p < .01,respectively) with no significant difference between the two effects (χ 2 [1] = 0.17, p > .1),supporting Hypotheses 1 and 2. The information sharing presents a significantly positive correlation with the PIS index (β = 0.127, p < .001)supporting Hypothesis 3.
Hypotheses 4 and 5 refer to the interactions between external knowledge breadth and internal information sharing, and external knowledge depth and internal information sharing.Model 3 and Model 4 in Table 3 report our second specification, which adds a full set of two-way interactions between Knowledge breadth and depth and Information sharing of organizational capability to the former model accordingly.A positive sign would mean that the two constructs reciprocally reinforce; a negative sign would point to a substitution effect.We did not find any significant effect pointing to the interplay between Knowledge breadth and Information sharing in Model 3, which did not support Hypothesis 4. In contrast, Model 4 has a significantly positive coefficient for Knowledge depth X Information sharing, pointing to a reciprocally reinforcing effect (β = 0.025, p < .001) that supports Hypothesis 5.
We provide a graphical representation of the result and plot the average marginal effects of knowledge depth for various information sharing values with 95% confidence intervals in Figure 2, providing additional evidence on the complementarity effect of Hypothesis 5. When all the other variables are at their mean values, the average marginal effects of knowledge depth are negative (but not statistically significant) at values of information sharing below the mean; however, they turn positive and are increasing for higher values of information sharing.The confidence intervals show that the average marginal effects of knowledge depth are significant only for values of information sharing above the mean.

Conclusions and discussion
Our study examined the effect of firms' external knowledge breadth and depth, and internal information sharing, on the ability of firms to promote SAPI.In addition, we tested how the interactions of external knowledge breadth and depth and internal information sharing affect SAPI.The results show that firms' external knowledge breadth and depth, and internal information sharing, have positive effects on the ability to combine different product information strategies.Moreover, firms' internal information sharing and external knowledge depth, rather than knowledge breadth, reciprocally reinforce each other to positively correlate with SAPI.We make the following contributions.
Theoretically, we first contribute to the NPD literature on antecedents on the ability to combine different product information strategies.We add to the extant literature by conceptualizing strategic ambidexterity in product innovation.Regarding NPD, prior literature mainly holds the assumption of firms' resource scarcity, particularly for SMEs, and accordingly highlights the existence of trade-offs in product innovation decisions (for example, Barroso & Giarratana, 2013).The trade-offs involve potential internal conflicts of diverse product information strategies to achieve "faster, better and cheaper" (Stanko et al., 2012;Swink et al., 2006), and difficulty in making appropriate product innovation strategies fit the external uncertainty of firms' business environment (Bingham et al., 2014).Firms thus tend to choose between product diversification based on a broad resource base and a strategic focus with narrow resource accumulation (Barroso & Giarratana, 2013;Bingham et al., 2014;Jourdan & Kivleniece, 2017).Our work revisits this paradoxical lens addressing the above trade-offs and argues that firms can develop SAPI, combining both variety and intensity of different product information strategies, as long as they are capable of orchestrating external and internal resources (Sirmon et al., 2007), which points to a third viable option to product innovation.We take a step further and investigate the external and internal antecedents of SAPI, which have attracted little attention in prior NPD literature (Acur et al., 2010;Vega-Jurado et al., 2008), and add depth to the NPD literature by exploring the relationship between external knowledge sources and internal capabilities and SAPI.Notably, we show that the following factors have a positive effect on firm SAPI: the capacity of the firm to (i) expand knowledge sources beyond organizational boundaries (Alexy et al., 2018;Laursen & Salter, 2006); (ii) leverage information sharing for internal resource orchestration (Sirmon et al., 2007, 2011), and(iii) ensure complementarity between firms' external knowledge depth and internal information sharing.In addition, we extend the NPD research in the context of Chinese manufacturing SMEs, with firms facing limited natural resources and experiencing immense and rapid changes in their business environment.As prior studies mainly consider the role of antecedents on Chinese manufacturing firms' product innovation performance (for example, Jeong et al., 2006;J. Zhang & Wu, 2017), we advance attention on SMEs in particular by focusing on the antecedents which matter to firms' ability to deal with diverse product information strategies, and further explore the role of external knowledge sources, internal information sharing, and their complementarity on firms' SAPI in a combined framework.
Second, we add to the open innovation literature.Studies on how knowledge breadth and knowledge depth affect firms' performance are regarded as a key issue in the field (for example, Laursen & Salter, 2006).We extend this research stream to the NPD context by targeting the role of knowledge breadth and depth on firms' product innovation strategic choices.We further show the complementary role of organizational capabilities (internal information sharing) for external knowledge sourcing (external knowledge depth), tentatively supporting the frontier opportunities proposed by Vanhaverbeke et al. (2014), targeting firms' functions beyond R&D capabilities in open innovation (we highlight organizational capabilities).As is the case with administrative styles, internal communication systems, and teams' interdependence (Vega-Jurado et al., 2008), organizational capabilities can be treated as internal hidden resources relating to bundles of non-R&D activities in open innovation contexts (West & Bogers, 2014).In this sense, information sharing plays an important integrating mechanism -also complementing West & Bogers's (2014) innovation integration review -helping firms profit from external sources.We further differentiate this integrating mechanism, and argue that it matters only to firms focusing on small numbers of external knowledge sources but cannot affect firms engaging in knowledge breadth.We surmise that this may be because the specific and focused knowledge is more likely to be translated, assimilated and transformed to consensus by firms in the information-sharing process (Wu, 2008), while the information sharing ineffectively reinforces the variety of knowledge sources' benefits.Based on this combined consideration of external knowledge sources and internal organizational capability on information sharing, we not only directly respond to Alexy et al.'s (2018) claim that future open innovation research should emphasize how firms jointly deploy external resources and internal capabilities for value creation and capture, but also steer the open innovation literature by connecting it with existing management theories (Vanhaverbeke et al., 2014) We provide relevant practical implications for managers as our primary stakeholder (Fisher, 2020).First, managers should consider an ambidextrous vision in the strategies they adopt for product innovation, coping with the potential trade-offs across different product information strategies and the changing environmental uncertainty.We argue that in addition to relying on an "either/or" dichotomy logic by acting as generalist or specialist firms coordinating different product information strategies, or by behaving as opportunists or strategists to adapt to the fast-changing environment to capture opportunity, managers could deliberately and strategically cultivate their strategic vision according to an ambidextrous logic, playing as a hybrid actor carrying out strategic variety and intensity simultaneously through the pursuit of diverse product information strategies.To do so, managers -particularly the ones who are in charge of SMEs -need to be aware of the challenge of resource constraints to developing SAPI.Accordingly, they need to primarily overcome resource constraints by leveraging external resources and/or enhancing organizational capabilities like information sharing by effectively orchestrating internal resources.Further, the combination of specific internal and external factors seems to reinforce each other.The mutually reinforcing effect between knowledge depth and information sharing we observed supports focusing on a reduced number of external sources -while facilitating internal information sharing -to help further develop SAPI.
Our study has some limitations that suggest avenues for future research.First, the cross-sectional design of the study may have a reverse causality problem.However, we suggest this is not a significant concern for three reasons.First, the existing literature in general argues that the external knowledge and internal capabilities, particularly information sharing, matter to firms' product innovations (for example, Laursen & Salter, 2006;Lin et al., 2013;Vega-Jurado et al., 2008).Second, it is difficult to explain the reverse causality of the significant interaction effects of our study, which offer stronger support to our hypotheses (Simons & Peterson, 2000, p. 109).Third, the faceto-face interviews with five senior managers in our study indicate that firms have engaged in exploring knowledge sources and practicing information sharing independently and prior to developing product information strategies, as these activities are more convenient, playing a fundamental role for firms accumulating strategic and operational resources.Nonetheless, we need to construct a longitudinal design to assess the hypotheses and uncover relevant dynamic mechanisms in our empirical research framework.
The second limitation is that our work focuses on a single country, China, which weakens the external validity of our results.Future work should aim to extend our analysis to a number of countries to obtain greater generalizability.Our third limitation is only including six types of product innovation strategy in our empirical setting as more product innovation strategies might be available to firms.
values (absolute value) above 0.1 are significant at standard confidence levels (p < .5);Descriptive statistics for Information sharing, Lean management, Training and learning and Tech capability refer to the average value of the seven-point Likert-type scale.

Figure 2 .
Figure 2. Average marginal effects of knowledge depth with 95% confidence interval.
. Here we embrace open innovation with the resource orchestration view, and indicate that their combination reflects how firms benefit from orchestrating both external and internal organizational resources.Finally, we further extend the open innovation literature's geographical context.Using data from Chinese manufacturing SMEs, we extend the open innovation frontier proposed by Vanhaverbeke et al. (2014) toward non-Western and SME contexts.

Table 2 .
Product innovation strategies, external knowledge leveraging, organizational capabilities and technological capability by industry (n = 600).

Table 3 .
Determinants of the balanced index of product innovation strategy (n = 600).