METHODOLOGY A. Sample and Questionnaire DesignThe sample used for this research consists of various types offirms, i.e., project-based and more functionally organized firms,with integrated and separated innovation activities, in the construction, information technology, and engineering industries.From the Reach database, across these industries, 1200 firmswere selected, all firms with more than 50 employees. The innovation managers of these firms were invited by telephone toparticipate in this research.To make comparisons across industries and different types offirms possible, an innovation project was defined as a projectin which a new product or service was developed and commercialized for more than one customer. Project-based types oforganizations also develop innovations in the projects executedto customer order, for example, an engineering firm builds anew factory for a customer. These types of innovations wereexcluded from this research.B. Response Rate and Sample BiasTwo different internet-based questionnaires were used. Thefirst was sent to the innovation manager. The second questionnaire was sent to project leaders. Of the 1200 firms, 720 innovation managers agreed to participate by phone. Of these managers203 (28%) filled out the online questionnaire. These innovation managers named 257 innovation projects and provided thee-mail addresses of 213 project leaders. One hundred and fortyeight of these project leaders responded (69%). Some projectswere deleted because the name provided by the project leaderdid not match the name mentioned by the innovation manager.The used dataset contains information of 142 projects. Of theseprojects, 96 are pairs within the same firm. Using MANOVA,no overall significant difference was found between firms thatprovided one or two projects (Wilks lambda = 0.83, F 1.62p > 0.05).Response bias was investigated by comparing the total sampleof firms (203) that answered the first questionnaire and the subsample of firms (95) that answered both questionnaires (Wilksslambda = 0.94, F=1.44, n.s.). There was a significant difference (p < 0.05) for firm strategy only. The firms in our samplefollowed a more progressive strategy based on the criteria ofMiles and Snow [50].Due to list-wise deletion, the number of cases used in theanalyses varies between 113 and 142 at the project level andbetween 86 and 95 at the organizational level.C. VariablesOriginally, each variable was derived from the literature.However, during the pretests in the various industry settings,a substantial number of questions had to be adapted to be applicable, relevant and clear to the respondents in the differentindustries.Several tests were applied to verify the validity of the modifications. Cronbachs alpha was used to verify the reliability ofeach reflective construct. For the formatively indicated type oforganization construct, the procedure described by Mackenzieet al. [51] was used. The unidimensionality and discriminantvalidity of the constructs was verified using confirmatory factor analyses. Discriminant validity was tested using a methodthat is assigned to Anderson and Gerbing as well as Bagozzi andPhillips [52]. This method compares the difference in chi-squarebetween pairs of constructs for an unconstrained and constrainedmodel, using covariance based structural equation modeling. Tosupport discriminant validity at the 5% significance level, thedifference in chi-square value between the constrained and unconstrained model needs to be larger than 3.84 [52].If applicable, data from all 203 firms were used to verifythe validity of the variables. These results are shown in squarebrackets.Performance is the projects commercial outcome, asperceived by the innovation manager. Asking for objective performance measures negatively impacts the response rate in innovation surveys [13], while using perceptions thereof leads tosimilar results [13], [53].To avoid common method bias, the innovation managersscores of performance were used [54]. Innovation managerstypically have a broader perspective [55] and, therefore, providea more reliable assessment of external market-related aspects ofinnovation projects [15].Performance is measured using the items for financial andcustomer-based performance of Griffin and Page [56]: the perceived impact of the new product or service on competitiveadvantage, the match with client needs, adherence to profit targets, and adherence to revenue targets. Market share was omittedafter the pre-test, as it appeared difficult to answer for the moreproject-based type of firms [see also 57, p. 46]. Perceived gainin reputation in the area of the new product or new service wasadded. This is an important aspect of performance in projectbased [58] and in service firms [59]. The Cronbachs alpha ofthe performance construct is = 0.86 Cross-functionality of the innovation team is the perception ofthe project leader regarding the cross-functionality of the innovation team. This practical approach was chosen, because innovation teams typically have fluid boundaries [6], which makes itdifficult to create a composite measur based on the backgroundof individual team members.Cross-functionality of the innovation team takes into accountthe participation of different disciplines, functional departmentsand the involvement of marketing ( = 0.74). Marketing is singled out as a discipline. Teams consisting of various technicaldisciplines only could already be considered cross-functionalin more project-based types of organizations. In the contextof innovation projects, cross-functionality typically implies notonly technical disciplines, but also the involvement of marketing [60]. ,with significant factor loadings for all items.Type of organization captures whether an organization is morefunctional or more project-based. It assesses the organizationalstructure of the firm, its outputs, and skills [21], [25]. A formatively indicated construct was used to model these variousaspects in one construct [61], [62]. Innovation leaders answeredwhich organizational form described their organization best;an operational processes resembling a project-based productionprocess (reverse coded), having a mass production process, delivering customized goods and/or services (reverse coded), ordelivering standardized goods. Project-management capabilitiesand the organizational structure of the organization were usedas global indicators to identify the model [62], [63]. Leaving out items because of insignificant loadings is not considered good practice for formative constructs, asit alters the meaning of the construct [61], [63]. For that reason,all items were included in the construct.Degree of separation of the innovation activities is measuredat the organizational level and assessed by the innovation manager. The operationalization of this concept is in line with thenotion of integration-separation by Lawrence and Lorsch [64].It covers the extent to which the innovation unit has autonomyover its own budget, to which extent innovation activities areformalized, and to which extent the innovation activities areprimarily the responsibility of this innovation unit. Cronbachsalpha for this construct is [0.75] 0.72. The confirmatory factor analysis shows significant factor loadings for all items andexcellent fit (2 = 3.36, df = 1, GFI = 0.99, SRMR = 0.04,CFI 0.97) [2 = 7.26, df = 1, GFI = 0.98, SRMR = 0.04,CFI = 0.95].Organizational connectedness is measured at the organizational level and assessed by the innovation manager. The operationalization of this construct is taken from Pinto and Pinto [65].It is operationalized as the degree to which functional departments and disciplines acknowledged each others expertise andhow common this type of communication is within the organization. Chronbachs alpha is = [0.90] 0.91, and the confirmatory D. Discriminant Validity of Independent VariablesThe chi-square differences between type of organization, organizational connectedness, degree of separation of the innovation activities, and cross-functionality of the innovation teamare all larger than 3.84. The lowest value is found for type oforganization and organizational connectedness ( 2 = 22.9),supporting statistically significant discriminant validity betweenthe independent variables.E. Control VariablesThere are four control variables at the firm level, all assessedby the innovation manager. First, firm size is used as a controlvariable, since cross-functional innovation projects in large corporations may face other challenges than in small firms [3]. Toaccount for the nonnormality of the size measure, a logarithmictransformation was used [66]. Second, industry effects are takeninto account. From the Reach database, the first two numbersof the firms first listed industry code were used: IT (72), Engineering (74), and Construction (45). Three dummy variablesare included to account for these and a fourth group of firmsfrom adjacent industries. Third, strategy ( = 0.70, [0.62]) isincluded as a control variable, to account for differences between prospector or more defensive innovation strategies [50].Fourth, product versus service firms is used as a control variable,since many project-based firms are service firms and becauseservice firms tend to integrate their innovation activities [18],[67].At the project level three control variables are included. First,newness of the product or service ( = 0.80) is included, sincethe management of radically new innovation projects may differ from that of incremental innovations [68]. Second, the involvement of senior management ( = 0.68, based on [82])is controlled for. Their involvement has a positive influenceon performance [4]; moreover, senior management could evaluate the projects in which they are involved more favorably.Third, the experience of the project leader is included as contro lV. DISCUSSIONAlthough no one doubts the value and need of crossfunctional information, cross-functional teams turn out to bea good mechanism to insert such knowledge into an innovationteam only in specific organizational contexts. Lack of collaboration within teams has typically been used to explain the mixedempirical findings regarding the effectiveness of such teams,and moderators that impact collaboration between team members of various backgrounds have been extensively researched[27], [30], [31], [75]. However, lack of collaboration cannot explain why for instance cross-functionality contributes negativelyto performance in the services industry [9]. This finding can onlybe explained by organizational context factors that moderate therelationship between the cross-functionality of a team and innovation performance, regardless of the level of collaboration atthe team level. Investigating the latter requires multilevel studies that include organizational level moderators and team levelvariables. In spite of the call for such studies [6], [10], [11],this paper is one of the first multilevel studies that investigatesteams in their organizational context, to study the effectivenessof cross-functional innovation teams.Our results show that organizational context moderatesthe performance of cross-functional innovation teams. Crossfunctionality of the innovation team contributes to theperformance of innovation projects in organizations with a morefunctional organization, or upon higher levels of separation ofthe innovation activities. Such a context is typical for manufacturing firms, which have dominated the innovation managementliterature in the past [12]. Cross-functionality of the innovationteam does not contribute to performance when there is a lackof connectedness at the organizational level. The organizationalcontext in which innovation teams operate thus matters, anda cross-functional team may not always be the most suitablemechanism to bring cross-functional information to a team. Forexample for project-based firms, or firms who have integratedtheir innovation activities in their day-to-day operations, puttingexperts with similar backgrounds on an innovation team maywork better.Few have challenged the usefulness of cross-functionalinnovation teams, but there are high costs associated withsuch collaboration [76] and high value outcomes arerare [77].Most of the ambiguous findings regarding the effectiveness ofcross-functional innovation teams have been found in the serviceliterature [9]. Many service firms have their innovation activities integrated within their operational activities [67], which isalso the case in our sample (see Table I). Integration of innovation activities could thus explain why Henard and Szymanski[9] found no contribution of cross-functional innovation teamsto innovation for service firms in their meta-analysis. Anconaand Caldwell [19] found a negative effect in high-tech firms.Many high-tech firms make use of a project-based organizational structure [21], [25].Studies that found a positive effect between the crossfunctional teams and innovation performance have typically been based on manufacturing firms, for example, themeta-analysis of Brown and Eisenhardt [4] had a bias towardsmanufacturing firms [12]. The meta-analyses of Evanschintzkyet al. [7] only includes new products studies, this exclusionof new services studies indicates a bias toward manufacturing firms. Froehle et al. [78] found a positive effect for crossfunctional teams in service firms, yet their sample had a biastoward service firms that use a more formal innovation process. Their findings could therefore be the result of the use offormal, and thereby often separated innovation structures [79].Taking organizational context into accounts thus helps explainambiguous findings of the past.A. Limitations Although the R, R2 , and R2 adjusted values in this study aretypical for team studies, or studies using independent evaluatorsto assess the dependent variable, these values indicate that alot of variance is left unexplained. Commercial performance isa distant outcome measure, which may explain why the R2 israther low. Less distant, proxy outcomes such as project performance could reduce the level of unexplained variance, however,project performance is not a good indicator for commercial success of an innovation project [15]. To reduce the unexplainedvariance, we recommend that future studies use a sample that isless diverse, for instance by focusing on various types of firmswithin a single industry such as IT.B. Practical ImplicationsCross-functional knowledge is important for all innovations.However, the effectiveness of cross functional innovation teams,as a mechanism to provide access to such knowledge, turns out tobe context specific. The bias in the literature for manufacturingfirms [12]which typically have a functional organization anda separate innovation unitmay have led to the impression ofcross-functional innovation teams as best practice [80].Project-based types of organizations that benefit less fromcross-functional innovation teams may need to integrate theirinnovation efforts to make the expert innovation teams effective. Summerize the article, are thoughtful, and analyze the content or question asked, make connections to other content and real-life situations ,extend discussions already taking place, or pose new possibilities or opinions not previously voiced