Financial projections in innovation selection: The role of scenario presentation, expertise, and risk

Innovation project selection is a decision of major relevance to firms. Errors in this decision may have serious consequences for firms, especially as many firms struggle with optimizing innovation project selection decisions. In their pitches to innovation decision-makers, project teams invariably present financial projections on their innovation projects, which often include bestand worst-case scenario presentation. Despite the potential influence the presentation of such financial projections has on firms’ innovation project selection decisions, this topic has not received sufficient attention in the literature. This study examines the role of scenario presentation on financial projections in innovation project selection by conducting two conjoint experiments among 2,425 managers and 11 follow-up interviews with senior executives. First, the findings of this study suggest that firms should help project teams present smallrather than large-range scenarios. This is important for at least the 57% of firms surveyed in this study where project teams are reported to present ‘too wide’ and ‘too extreme’ scenarios. Second, firms seeking to promote transformational innovation in their innovation pipeline should make the presentation of small-range scenarios required for an innovation proposal to be presented to a project selection committee. This is relevant for 79% of surveyed firms that would like to select more transformational than core innovation projects and especially for the half of which that currently do not require scenario presentation. Third, project teams with less expertise should develop scenarios analytically rather than intuitively and convey the project’s strategic merit to decision-makers to help increase innovation project selection likelihood.


Introduction
Innovation project selection is a decision of major relevance to marketers (Cooper, 1979). Marketers are typically members of committees of innovation decision-makers who make project selection decisions and, thus, play a pivotal role in innovation project selection decisions (Ernst, Hoyer, and Rubsaamen, 2010). While prior literature has abundantly covered innovation generation decisions, innovation selection as a decision is understudied. Many firms struggle with innovation selection (Birkinshaw, Bouquet, and Barsoux, 2011). Innovation selection decisions nevertheless are consequential. For instance, firms may err on the prudent side and select too many core innovation projects that target existing markets with products that are similar to their current offerings and too few transformational projects that target new markets with new products. Or, firms may take too much risk and neglect core innovation, giving up their leadership position in the market. Most firms, however, are heavily oriented towards core innovations and would like to select more transformational innovations in their innovation pipeline (Day, 2007).
In their pitches to innovation selection committees, innovation project teams invariably include financial projections, such as expected cumulative cash flows. Practitioners suggest that the presentation of such financial projections and metrics is likely to influence firms' innovation selection decisions (Christensen, Kaufman, and Shih, 2008;Nagji and Tuff, 2012). However, this influence, or the channels through which it operates, has not yet been demonstrated empirically.
Interestingly, such financial projections may include best-and worst-case scenarios (Schoemaker, 1993(Schoemaker, , 1995. In a survey of 1,268 executives, Bain & Company (2017) reports that the prevalence of such scenarios across various decision areas is about 19%. Focusing on innovation project selection decisions specifically, we find, in the present study, that project teams present scenarios in about half of our surveyed firms. 1 The business press debates the usefulness of scenario presentation for project selection. On the one hand, decision-makers may feel that presentation of scenarios improves their ability to understand and manage project risk (Koller, Lovallo, and Williams, 2012). On the other hand, decision-makers may perceive scenario presentation as an ''abdication of leadership" that casts doubt on the project team's capability to successfully execute the project (Roxburgh, 2009). These conflicting views suggest that the effects of scenario presentation on project selection decisions are complex and require more scholarly research.
Prior literature examines how project selection is influenced by, for example, information in the innovation proposal, characteristics of decision-makers and project teams, and the design of selection processes (see Table 1). We extend this literature in two key respects. First, we contribute to prior research that examines how the presentation of information in innovation proposals influences project selection. Prior work shows that factors such as information length and tone, presentation of demos and prototypes, influence project selection (Kirsch, Goldfarb, and Gera, 2009;Lu et al., 2019;Reitzig and Sorenson, 2013). No study, however, examines specifically the effects of scenario presentation on project selection decisions. Our paper addresses this gap. Nor do prior studies in this area distinguish information presentation effects for different types of innovation (one exception to some extent is Lu et al. (2019) who categorize innovation based on its novelty from the customers' perspective -see Table 1). We distinguish types of innovation from the firm's perspective as suggested by the literature for studying innovation development because it is the firm that must allocate resources to develop selected projects (Danneels and Kleinschmidt, 2001;Sethi, Iqbal, and Sethi, 2012). In so doing, we show that the effect of scenario presentation on project selection decisions is different for core vs. transformational innovation projects.
Second, the literature on the innovation fuzzy front-end focuses predominantly on innovation idea generation and less on innovation project selection. The scant research on project selection (e.g., Criscuolo et al., 2017;Girotra, Terwiesch, and Ulrich, 2010;Keum and See, 2017) applies various theoretical lenses to explain project selection decisions but does not examine the decision-making process empirically. Moreover, the literature tends to disregard the importance of the decision-making context. Our paper addresses these gaps. Specifically, to describe the decision-making process, we adopt attribution theory and posit that scenario presentation influences the decision-makers' perceptions about project team expertise and project risk which, in turn, influence decision-makers' project selection decisions. We then examine how these influences are contingent upon two relevant decision-making contexts in which the project selection decision is taken: (1) the method of scenario development (intuitive vs. analytical) and (2) the strategic merit of the innovation (i.e., its benefits beyond the project's financial payoff).
We conduct two choice-based conjoint experiments among 745 and 1,680 managers (Study 1 and 2, respectively) and 11 complementary interviews with senior executives. We contribute to managerial practice as to when to make the presentation of scenarios required; whether to help project teams narrow down the range between scenarios; when project teams should develop scenarios intuitively or analytically; and when conveying the project's strategic merit is key.

Background
The literature distinguishes several stages of the innovation development process (e.g., Cooper, 2006;Kornish and Ulrich, 2014;Sethi and Iqbal, 2008). The process starts with concept development, a stage where project teams generate ideas about solutions to customer needs or opportunities to use new technologies (Ernst, Hoyer, and Rubsaamen, 2010;see Fig. 1). Promising ideas are selected to receive funding and proceed to the innovation development stage, where ideas turn into tangible projects (Olson, Walker, and Ruekert, 1995). The most promising projects are selected to move into the implementation stage, during which the innovation is implemented.
To obtain funding for their project, project teams pitch the results of their progress through these stages to innovation decision-makers who make ''go/no-go" decisions at each stage (Cooper, 2006;Sethi and Iqbal, 2008;see Fig. 1). Our focus is on the project selection decision after the concept development stage, which is a decision taken under high uncertainty. 1 We extracted this finding from the survey in Study 2, which we report in Section 5 below. V. Avagyan, N. Camacho, W.A. Van der Stede et al. International Journal of Research in Marketing xxx (xxxx) xxx Table 1 Key empirical studies on innovation project selection.

Paper Empirical basis
Variables of focus Decision-making process (P) / context (C) Theoretical lens Key findings Girotra, Terwiesch, and Ulrich (2010) & Lab experiment (N = 44 students) & Group structure (team vs. hybrid) Decision-makers' problem solving and information processing Individuals working in a hybrid structure (first individually and then as a team) are better able to discern high quality ideas than individuals working in teams. Reitzig and Sorenson (2013) & Secondary data ( In their pitches to decision-makers, project teams often include financial projections, such as expected revenues, expected Cumulative Cash Flows (CCFs), or other relevant information for a normal-case scenario of the project. Project teams may also include financial projections of best-and worst-case scenarios in addition to the normal case, which we refer to as scenario presentation for short (Schoemaker, 1993).
To substantiate how such scenario presentation influences decision-makers' project selection decisions, we draw on attribution theory (Kelley and Michela, 1980) and interviews with senior executives that supported our theory development (see Section 7). At the most intuitive level, attribution theory explains the process someone (in our case, the innovation decisionmaker) goes through when seeking to discern the behaviors of others (the project team)-i.e., when seeking to attribute their behaviors to causes for them. For instance, decision-makers may make inferences about the project team's expertise or the project's risk based on whether the project team presents scenarios.
Attribution theory is especially relevant for decisions under high uncertainty where it is challenging for decision-makers to be ''in control" and discerning the causes of others' behaviors is especially valuable (Berscheid et al., 1976). Thus, attribution theory provides a suitable lens for our study because project selection happens under high uncertainty where innovation decision-makers, who control innovation development gates, must make inferences about project team behaviors based on limited information in the pitch (Sethi and Iqbal, 2008).
More specifically, to capture the relevant aspects of the attribution process, our model incorporates (1) dimensions of attribution, (2) decision-makers' expectations, and (3) boundary conditions. We focus on two dimensions of attribution: the communicator (i.e., the project team) and the stimulus (i.e., the innovation project) (Laczniak, DeCarlo, and Ramaswami, 2001). As key characteristics of the project team and the innovation project, that are pertinently linked with the controllability premise of attribution theory (Huang and Pierce, 2015), we focus on project team expertise and project risk. Perceived project team expertise refers to the knowledge the project team has related to the proposed innovation as perceived by the decision-maker (Chandy et al., 2006). Perceived project risk refers to the variance in possible outcomes of the innovation project as perceived by the decision-maker (Weber and Hsee, 1998). Innovation project selection requires that decision-makers ''go beyond the numbers" to assess the project team's ''general understanding of the project" (Jørgensen and Messner, 2010); that is, an assessment of the project team's expertise is pivotal. Furthermore, because project risk is a key concern of innovation decision-makers (March and Shapira, 1987), presenting alternative outcomes of the proposed innovation as scenarios is likely to trigger decision-makers' attributions about the project's risk.
In sum, we posit that scenario presentation influences the innovation decision-makers' perceptions of the project team's expertise and the project's risk which, in turn, influence decision-makers' project selection decisions (see our conceptual framework in Fig. 2).
Attribution theory holds that decision-makers' expectations influence how they generate attributions (Kelley and Michela, 1980). That is, innovation decision-makers generate explanations for project team's behaviors, such as whether they present scenarios, adjusted for whether they expect the project team to present scenarios in a given decision-making context. One aspect of a decision-making context is the type of innovation being pitched, which we focus on in Study 1. Specifically, the augmentation principle implies that decision-makers generate more attributions about project team expertise and project risk when scenarios are presented but were unexpected; that is, when the decision-maker deems scenario presentation as unsuitable or ''contraindicated" for the type of innovation being pitched. On the other hand, when scenario presentation is expected, the discounting principle implies that decision-makers lessen or discount their attributions about project team expertise or project risk because, in this case, there is another readily plausible cause for triggering scenario presentation (i.e., the type of innovation being pitched).
Thus, we suggest that the effects of scenario presentation on the perceived project team expertise and perceived project risk are contingent on the type of innovation. A commonly used framework in the innovation literature categorizes innovations by the newness of the intended market and the newness of the proposed product to the firm (Danneels and Kleinschmidt, 2001;Olson, Walker, and Ruekert, 1995;Sethi, Iqbal, and Sethi, 2012). The newness of the intended market is the extent to which the proposed product targets new markets and requires new sales and customer support system compared with the firm's existing products. The newness of the product is the extent to which the proposed product requires new technologies, engineering and design skills, and production processes compared with the firm's existing products. This framework is also frequently applied in practice (Day, 2007;Nagji and Tuff, 2012). We adopt this framework to distinguish between core and transformational innovation projects. Core innovations are incremental changes to a firm's current products to serve its existing customers. Transformational innovations are new offerings targeting new customers with new products.
Finally, attribution theory posits that boundary conditions, such as the decision-making context, makes decision-makers' attributions more or less influential in guiding their decisions (Green and Mitchell, 1979). That is, the extent to which attributions about project team expertise and project risk influence project selection decisions depends on the decision-making context. In Study 2, we incorporate two relevant aspects of the decision-making context: (i) the method project teams use to develop scenarios and (ii) the innovation project's strategic merit, both of which are grounded in prior work on scenarios (Schoemaker, 1992(Schoemaker, , 1993.

Scenario presentation and perceived team expertise
We consider how scenario presentation may affect perceived team expertise by way of positive or negative team expertise attribution, respectively.
Presentation of scenarios may enhance perceived team expertise. We refer to this as positive team expertise attribution. Decision-makers attribute the voluntary presentation of a worst-case outcome, together with the best case, to the expertise of the message source (Crowley and Hoyer, 1994). Presentation of scenarios indicates that the project team willingly accepts the possibility of setbacks and, thus, has adequate knowledge of customer needs and the product under various circumstances. When only the normal case is presented, decision-makers are still aware that a worst-case scenario may occur, and thus its omission from the pitch is likely to reflect negatively on the perceived expertise of the project team.
However, scenario presentation may also decrease the perceived team expertise, which we refer to as negative team expertise attribution. Decision-makers may attribute presentation of scenarios that depict alternative plausible futures to a lack of confidence and agreement among project team members about the project's strategies (Du et al., 2011), causing decisionmakers to perceive the project team as being less expert (Slotegraaf and Atuahene-Gima, 2011). In other words, unlike a single scenario (i.e., the normal case), alternative scenarios may be attributed to a lack of expertise by the project team to converge on a precise outcome of its project.
We first theorize that, ceteris paribus, positive team expertise attribution is stronger for transformational than for core innovations. Project teams encounter greater uncertainty and take longer when developing projects that are new to the firm (Olson, Walker, and Ruekert, 1995;Sethi and Iqbal, 2008). Therefore, setbacks are more likely and more consequential for transformational than for core innovations (Day, 2007). That said, presenting the possibility of setbacks to decisionmakers naturally runs counter to project team's interests. Consequently, presenting scenarios that do reveal a project's setbacks to decision-makers is more ''contraindicated"-and, according to the augmentation principle (Kelley and Michela, 1980), more likely to trigger positive team expertise attributions-in the case of transformational than in the case of core innovations. Thus, when project teams present scenarios, decision-makers are more likely to augment positive team expertise attributions for transformational than for core innovation projects.
Second, we also theorize that negative team expertise attribution is weaker for transformational than for core innovations. There is ''less relevant experience to draw on" when developing innovation projects that are new to the firm (Olson, Walker, and Ruekert, 1995). Thus, innovation decision-makers know that while project teams proposing core innovations have access to sufficient information to develop their project, project teams proposing transformational

Project Selection
Project teams pitch their project to innovation decision-makers who select which projects to move into the development stage. They may or may not present scenarios in their pitches.

Concept Development
Idea generation Analysis of trends, market changes, and market potential Assessment of needed funds, time, and risk related to the project Determination of desired innovation features Scenario analysis

Development of prototypes Preparation of commercialization concept Execution of prototype tests and validation
Evaluation of market acceptance

Go-to-Market Decision
Project teams present the results of completed activities and their go-tomarket proposal to innovation decision-makers who decide which projects to move into the implementation stage. 2 While H1-H3 represent expectations we had before any data gathering took place, H4-H5 were added based on reviewer feedback after Study 1. innovations may need to rely on scarce information and may have more grounds for disagreements among project team members about the best strategies to develop the project. Consequently, presenting scenarios that depict alternative plausible futures is more congruent with the expectations of decision-makers-and, according to the discounting principle (Kelley and Michela, 1980), less likely to trigger negative team expertise attributions-for transformational than for core innovation projects. Hence, taken together, scenario presentation is likely to increase the perceived team expertise more, or decrease the perceived team expertise less, for transformational than for core innovations. ''Increasing the perceived team expertise more" or ''decreasing the perceived team expertise less" are akin to having a more positive predicted relationship between scenario presentation and perceived team expertise for transformational than for core innovations. Stated as an empirically testable hypothesis, this gives:

Go/ No-Go Decisions
H 1 There exists a more positive relationship between scenario presentation and perceived project team expertise for transformational than for core innovations.

Scenario presentation and perceived risk
We again consider two sides that may underlie the effect of scenario presentation on innovation decision-makers' risk perceptions. On the one hand, scenario presentation may reduce perceived project risk, which we refer to as the project risk-reducing attribution. Scenario presentation places bounds on the project's expected outcomes (Goodwin and Wright, 2001). An innovation proposal without scenarios may indicate to decision-makers that the project has high ambiguity and complexity that prevent the project team from conjecturing and presenting reliable scenarios (Ahmad, Mallick, and Schroeder, 2013;Hogarth and Kunreuther, 1995). Thus, when the project team does present scenarios, decision-makers are likely to infer that the project is less ambiguous and complex, thereby reducing its perceived risk (Cabantous et al., 2011).
On the other hand, scenario presentation may increase perceived project risk, which we refer to as the project riskincreasing attribution. Just like negativity bias suggests that decision-makers may consider negative information as more consequential than positive information (Baumeister et al., 2001), attribution theory suggests that worst-case outcomes are more attributed to the project and more influential to the decision-maker than best-case outcomes of similar magnitude (Chen and Lurie, 2013). Hence, scenario presentation may lead innovation decision-makers to put more weight on the worst-than on the best-case scenario, thereby increasing their perception of project risk.
We propose that, ceteris paribus, the project risk-reducing attribution is stronger for transformational than for core innovations. Due to the higher uncertainty involved with transformational projects, innovation decision-makers know that project teams are less able to conjecture the possible outcomes for transformational innovations (Olson, Walker, and Ruekert, 1995). Therefore, when the project team does present scenarios, it is likely to indicate to decision-makers that the project has less ambiguity and complexity which enabled the project team to conjecture scenarios. Thus, presenting scenarios has the potential to disambiguate transformational innovation projects and thereby reduce perceived project risk. In contrast, for core innovations, project teams should be more easily able to conjecture the possible outcomes of the project (Olson, Walker, and Ruekert, 1995). Hence, presenting such outcomes as scenarios is more expected, and-according to the discounting principle (Kelley and Michela, 1980)-less influential to reduce decision-makers' perceptions of project risk. We also theorize that project risk-increasing attribution is weaker for transformational than for core innovations. Transformational innovations are more likely to have negative outcomes than core innovations (Day, 2007;Olson, Walker, and Ruekert, 1995). Consequently, presenting scenarios that show a negative outcome of the project is more expected for innovation decision-makers-and, according to the discounting principle (Kelley and Michela, 1980), less likely to draw their attention to such an outcome-in the case of transformational than in the case of core innovations. Therefore, scenario presentation is less likely to trigger a project risk-increasing attribution for transformational than for core innovation projects.
Taken together, scenario presentation is likely to decrease the perceived risk more, or increase the perceived risk less, for transformational than for core innovations. Stated formally: H 2 There exists a more negative relationship between scenario presentation and perceived project risk for transformational than for core innovations.

Perceived team expertise, perceived risk and project selection
Decision-makers' attributions guide their behavior (Kelley and Michela, 1980). First, we hypothesize that perceived team expertise increases project selection likelihood. This is because decision-makers are likely to expect project teams with greater expertise to be better able to satisfy customer needs, address implementation problems, and thus, ensure the eventual success of their innovation (Chandy et al., 2006;Slotegraaf and Atuahene-Gima, 2011). Second, we hypothesize that perceived project risk decreases project selection likelihood. Indeed, there is broad support for the negative relationship between a project's risk and decision-makers' propensity to fund it based on the general premise of risk aversion (e.g., Cooper, 2008;March and Shapira, 1987). Specifically, decision-makers are more likely to refrain from funding a riskier project because possible large downside losses may affect the resources under their control as well as their power (Sethi, Iqbal, and Sethi, 2012). Riskier projects are also less likely to be delivered, thus adversely affecting decision-makers' reputation (Criscuolo et al., 2017). Thus: H 3a The greater the perceived team expertise, the higher the likelihood that a decision-maker funds the proposed innovation project. H 3b The greater the perceived project risk, the lower the likelihood that a decision-maker funds the proposed innovation project.

Method of scenario development and project selection
We argue that the way in which project teams develop scenarios-i.e., analytically vs. intuitively-affects the extent to which decision-makers' attributions of project team expertise and project risk influence project selection likelihood. Decision-makers learn about the method of scenario development through their evaluations of the innovation proposals (Jørgensen and Messner, 2010). Project teams that develop scenarios analytically rely on procedures, using objective data and systematic analyses, including benchmarks (Schoemaker, 1995;Sloman, 1996). These analyses typically use historical data on firm's prior launches of similar products and past applications of similar technologies, which project teams refer to in their proposals. Such information prompts decision-makers to use their existing knowledge of the reference products and technologies (Bolton, 2003), which causes them to less likely rely solely, and thus rely less, on attributions about project team expertise and project risk to guide project selection decisions.
In contrast, project teams that develop scenarios intuitively use affectively-charged judgments and ''gut instincts" (Dane and Pratt, 2007) instead of referring to firm's prior launches of similar products and past applications of similar technologies in their proposals. As such, decision-makers have fewer, if any, explicit reference points for their project selection decisions, causing them to rely more on their attributions about project team expertise and project risk. Hence, we expect stronger effects of perceived team expertise and perceived project risk on project selection likelihood when project teams develop scenarios intuitively than when they develop scenarios analytically.
H 4a The effect of perceived team expertise on the likelihood that a decision-maker funds the proposed innovation project is more positive when project teams develop scenarios intuitively than when they develop scenarios analytically. H 4b The effect of perceived project risk on the likelihood that a decision-maker funds the proposed innovation project is more negative when project teams develop scenarios intuitively than when they develop scenarios analytically.

Strategic merit of the project and project selection
We propose that the extent to which decision-makers' attributions about project team expertise and project risk influence project selection likelihood depends on the project's strategic merit. An innovation project has high strategic merit if V. Avagyan, N. Camacho, W.A. Van der Stede et al.
International Journal of Research in Marketing xxx (xxxx) xxx it offers benefits beyond its direct financial merit, helping the firm compete more effectively (Ernst, Hoyer, and Rubsaamen, 2010;Noordhoff et al., 2011). Such benefits include increasing the sales of other products in the firm's assortment, improving customer relationships, and developing employee talent. First, we argue that the effect of perceived team expertise on project selection likelihood is weaker for projects with high rather than low strategic merit. Firms concentrate their innovation development efforts and allocate more budget to innovation projects of high strategic merit (Cooper, 2013). This enables firms to support project teams to gather more information about the market and the competitors that may enhance project team's knowledge in the proposed innovation after such project is selected (Calantone, Schmidt, and Song, 1996). Firms may also involve external specialists with various functional skills to support the project team to develop such project that gets selected (Olson, Walker, and Ruekert, 1995). Hence, decision-makers are less reliant on attributions about project team expertise when the project has high strategic merit.
Second, we argue that the effect of perceived risk on project selection likelihood is also weaker for projects with high rather than low strategic merit. When evaluating a project in terms of its risk, decision-makers take into account the project's financial and strategic impact on the firm (Kaplan and Strömberg, 2004). Decision-makers are more likely to undertake a high-risk decision if such decision promises to deliver a high financial and strategic return (March and Shapira, 1987). Hence, all else equal, decision-makers are less reliant on their attributions about project risk and may accept a higher risk when a project has high strategic merit.
Taken together, we expect weaker effects of perceived team expertise and perceived project risk on project selection likelihood when decision-makers evaluate projects with high rather than low strategic merit.
H 5a The effect of perceived team expertise on the likelihood that a decision-maker funds the proposed innovation project is less positive for projects with high strategic merit than for projects with low strategic merit. H 5b The effect of perceived project risk on the likelihood that a decision-maker funds the proposed innovation project is less negative for projects with high strategic merit than for projects with low strategic merit.

Other variables
We control for several additional variables. First, we control for the age, gender and education of the innovation decisionmaker (Chen, Yao, and Kotha, 2009). Second, in line with research in marketing and finance, we control for decision-makers' (1) familiarity with financial concepts, (2) knowledge in investing, (3) quantitative background, and (4) training in metrics (Hoffmann and Broekhuizen, 2010;Mintz and Currim, 2013). Third, following prior research, we control for the decisionmaker's risk attitude (Weber, Blais, and Betz, 2002) and depth of elaboration (Nenkov et al., 2009) in our analyses. Fourth, we control for the complexity of the project evaluation task (Lee and Shavitt, 2009). Finally, we control for the interaction between perceived project team expertise and perceived project risk in project selection models.

Field interviews with multinational companies and pretest
In this study, we test our hypotheses H 1 -H 3 using an online choice-based conjoint experiment with experienced managers as subjects. We went through several steps to ensure the practical relevance and realism of our experiment. Specifically, we conducted several interviews at four multinational companies to validate the importance of scenario presentation on innovation project selection decisions (Web Appendix A). We pretested our measures and experimental stimuli on a sample of 251 undergraduate students attending degree programs at one of our universities to verify the clarity of our experimental task. To further ensure the realism of the operationalization of our factors, we calibrated our stimuli based on 31 actual innovation proposals and on the advice from two senior innovation executives from Alcatel-Lucent Bell Labs. These innovation proposals cover a wide array of applications, ranging from new hardware components, new software and mobile/IOT applications, to entirely new business models.

Subjects
We contracted instantly to recruit and incentivize participants to our conjoint experiment from its online executive panel in the US. Subjects were considered eligible for our study if they (1) were sufficiently knowledgeable about innovation within their firm (at least 6 on a 10-point scale), (2) worked for a firm with 500 employees or more, and (3) did not work in financial institutions, insurance or consulting. Out of 1,215 potential respondents who clicked on the link of our study, 373 (30.7%) were ineligible and 69 (5.7%) exited the study early, resulting in 773 participants. We then conducted additional checks and removed 28 multivariate outliers from our data 3 (see Web Appendix B).
Our final sample consists of 745 managers with an average age of 39.7 years, 40.5% female, 91.5% holding a bachelor's degree or higher, and an average tenure at their firm of 11.5 years. The sample participants were highly knowledgeable about innovation within their firm (an average of 8.7 on a 10-point scale) and reported high knowledge and experience of investing in general (5.3 and 5.2, respectively, on a 7-point scale).

Experimental procedures
At the start of our experiment, subjects read a vignette depicting a hypothetical situation in which they worked for a leading global telecommunications equipment supplier Telforce and had to select which innovation projects to fund (see Web Appendix B). To ensure the truthfulness of participants' project selection decisions, we informed participants that the best performing participant would earn a reward of $250 (see Web Appendix B for our reward mechanism).
Next, we administered our choice-based conjoint task. We created six unique innovation projects (or profiles, in conjoint terminology) based on our experimental manipulation of two factors: type of innovation (core vs. transformational) and scenario presentation (no scenarios, small-range, and large-range scenarios). We distinguished small-range and large-range scenario presentation because prior studies suggest that decision-makers' attributions may vary depending on the range between scenarios (Du et al., 2011;Schoemaker, 1993). Our experimental design thus enables us to test empirically whether our theoretical arguments regarding the moderating effect of type of innovation in H 1 and H 2 hold for different ranges between scenarios.
We constructed all possible pairs of these six projects (Louviere and Woodworth, 1983), resulting in 15 choice tasks. In each choice task, the participant had to evaluate two innovation projects described by their financial projections and by a description of the type of project. Next, she had to select which of the two projects she would fund. Our pretests revealed that 15 choice tasks did not constitute an inordinate cognitive burden on participants. We randomized the order of pairs and the order of profiles within each pair to rule out order effects (Criscuolo et al., 2021).
We manipulated type of innovation by varying the extent to which the proposed innovation targets new markets with a new product. We ensured that transformational innovations, as compared to core innovations, had (1) a higher CCF in the normal-case scenario and (2) a lower probability of occurrence of the normal-case scenario. 4 Specifically, following Day (2007), we designed the CCF in the normal-case scenario for transformational innovations to be $30 M after five years with a low probability of occurrence (20%), whereas for core innovations, we designed the CCF in the normal-case scenario to be $7.5 M after five years with a high probability of occurrence (80%). We based this value on the CCFs of the transformational projects among the Alcatel-Lucent Bell Labs' projects we examined. Finally, we also based the required investment and target return of each project on the projects from Alcatel-Lucent Bell Labs (see all conjoint profiles in Web Appendix B).
To manipulate scenario presentation, we again used Alcatel-Lucent Bell Labs' innovation projects. About 40% of the projects we examined contained scenarios. In the ''no scenarios" level, we presented CCFs only in the normal-case scenario without any mention of best-or worst-case scenarios. To determine the ''small-range" and ''large-range" scenario levels, we proceeded as follows. We consider the size of the scenario range as the difference between the CCFs in the best-and worst-case scenarios relative to the CCFs in the normal-case scenario. We used the Alcatel-Lucent Bell Labs' projects with below-median scenario range to construct the small-range scenario level, which is equal to 4/3 of the CCFs in the normal-case scenario. For the large-range scenario level, a similar procedure resulted in a range that is four times larger than the range in the small-range scenario level (i.e., 16/3 of the CCFs in the normal-scenario case).

Measures
After the 15 pairwise project selection decisions, we presented again the six unique projects, one at a time and in a random order, and measured the perceived team expertise and the perceived risk of each project (see Web Appendix B). We used three items (i.e., expert, experienced, and knowledgeable) to measure the perceived team expertise on seven-point scales (adapted from Ohanian, 1990), and a numerical rating scale ranging from 0 (''not at all risky") to 100 (''extremely risky") to measure perceived project risk (Weber and Hsee, 1998). After conducting our experimental task and measuring expertise and risk perceptions, we administered the rest of our measures (we provide all measures, their respective sources, and their reliabilities in Web Appendix E). We adopted scales widely used in prior research for all our constructs. To ensure truthfulness and reduce inattentive or careless responses, we used different scale formats and guaranteed respondents full anonymity (Podsakoff et al., 2003).

Model specification and estimation
Let us denote each project by j = 1,. . .,6 and each choice set by k = 1,. . .,15. We denote project j appearing in choice set k by j(k). We estimate two hierarchical linear models to test hypotheses H 1 and H 2 , which pertain to the effects of scenario 4 Note that our operationalization of the type of innovation factor encompasses (i) the extent to which the proposed innovation targets new markets with a new product, (ii) the CCF in the normal-case scenario, and (iii) the probability of the normal-case scenario. This operationalization follows prior literature (e.g., Day 2007;Rosenkopf and McGrath 2011). Thus, manipulating type of innovation orthogonally to the probability of normal-case scenario is infeasible. As such, our intent is not to test separate effects of each of these components of type of innovation. (1)) and on perceived risk (Eq. (2)). We control for several subject-specific control variables (Controls i ). We include subject-and projectspecific error terms in the model (e ij,E and e ij,R , for the expertise and risk equations, respectively), which we assume to be independently normally distributed with mean zero. To capture unobserved heterogeneity among subjects, we model the parameters capturing the effects of our conjoint factors as random coefficients with their own error terms (t i,E for expertise and t i,R for risk). We assume that these error terms are independently and normally distributed with mean zero and are independent of the residuals (e ij,E and e ij,R ). We estimate these models using restricted maximum likelihood.
To model managers' project selection decisions, we use a mixed logit specification with repeated choices (Revelt and Train, 1998). The mixed logit model circumvents several limitations of the standard logit model by allowing for random taste variation, unrestricted substitution patterns, and correlation in unobserved factors over choice sets and does not impose the independence from irrelevant alternatives assumption (Train, 2009).
Let us denote the utility that manager i obtains from selecting project j appearing in choice set k by U ij(k) . To test H 3a and H 3b , we express U ij(k) as a function of manager i's perceptions regarding project j's team expertise (EXPERTISE ij(k) ) and risk (RISK ij(k) ), the interaction between expertise and risk and an unobserved individual, project and choice set specific error term (e ij(k) ) that captures unexplained variance in managers' preferences for different innovation projects (Eq. (3)). To allow for heterogeneity in managers' preferences for observable project characteristics, we model the parameters (b) in our model as random coefficients with their own error terms (t i ). We allow for different variances for these error terms and assume that they are independently and normally distributed. We assume that the error terms (e ij(k) ) are iid extreme value, independent of observed variables and coefficients.
Eq. (4), in turn, tests the same relationship while including scenario presentation, type of innovation and their interaction in the model, allowing us to account for the influence of scenario presentation on innovation project selection decisions beyond its influence through project team expertise and perceived risk of the project: To estimate the model parameters in Eqs. (3) and (4), we assume that decision-makers make utility-maximizing project selection decisions. We use the utility functions specified in Eqs. (3) and (4) to derive, using the conditional logit formula, the probability that manager i chooses to fund project j in choice set k. We estimate our models using simulated maximum likelihood. The resulting log likelihood does not have a closed form solution. Hence, we implement the procedure proposed by Revelt and Train (1998) to simulate the likelihood function value. We use 50 Halton draws in our simulations. Table 2 presents the estimation results for Study 1. Unless otherwise mentioned, we use effects coding for the type of innovation factor (where core is coded as À1 and transformational as 1) and dummy coding for the scenario presentation factor (where small-range scenario is coded as 1 if small-range scenarios are presented, 0 otherwise; and large-range scenario is coded 1 if large-range scenarios are presented, 0 otherwise). For this and subsequent analyses, we compute variance inflation factors (VIFs) for all independent variables. The maximum VIF is 3.2, indicating that multicollinearity is not a threat.

Results and discussion
Scenario Presentation and Perceived Expertise. Columns 2 and 3 of Table 2 present the results of our perceived expertise model (Eq. (1)). We find that the model has a better fit (see Table 2; AIC = 12,795; BIC = 12,833) compared to a model without random coefficients. In line with H 1 , we find that the effect of scenario presentation on perceived team expertise is more positive for transformational than for core innovation projects both for small-range scenario presentation (b = 0.07, p < .05) and large-range scenario presentation (b = 0.06, p < .05).
Additional findings deserve mention. First, although we made no formal prediction, we examine the effects of scenario presentation on the perceived expertise for small and large ranges. We find that small-range scenarios increase the perceived expertise (b = 0.16, p < .01) but large-range scenarios decrease the perceived expertise (b = À0.11, p < .01).
Second, to identify the marginal effects of small-and large-range scenarios on perceived expertise for core and transformational innovations, we dummy code the type of innovation (i.e., the focal level of type of innovation is coded as 0 and the other level as 1) and then re-estimate Eq. (1). We find that, for core innovations, small-range scenarios increase (b = 0.09, p < .05), whereas large-range scenarios decrease, perceived team expertise (b = À0.17, p < .01). For transformational innovations, while small-range scenarios increase the perceived expertise (b = 0.23, p < .01), the effect of large-range scenarios is not significant (b = À0.05, p > .10).
Third, when no scenarios are presented, decision-makers perceive the project teams of transformational innovations as less expert than those of core innovations (b = À0.22, p < .01).
Fourth, for the control variables, we find that decision-makers' depth of elaboration on potential outcomes (b = 0.14, p < .01) and risk-seeking attitude (b = 0.38, p < .01) have a positive influence, whereas decision-makers' age has a negative influence (b = À0.01, p < .01) on the perceived team expertise. Finally, decision-makers' general knowledge in investing (b = 0.10, p < .05), quantitative background (b = 0.05, p < .05) and level of training in metrics (b = 0.11, p < .01) all have a positive influence on perceptions of expertise.
Scenario Presentation and Perceived Risk. Columns 5 and 6 of Table 2 present the results of our perceived risk model (Eq. (2)). We find that the model has a better fit (see Table 2; AIC = 40,378; BIC = 40,416) compared to a model without random coefficients. Supporting H 2 , we find that the effect of scenario presentation on perceived project risk is more negative (for small-range scenarios; b = À2.53, p < .01) and less positive (for large-range scenarios; b = À1.27, p < .05) for transformational than for core innovations.
Project Selection Decisions. Columns 2 and 3 of Table 3 present the parameter estimates of the model described in Eq. (3). Supporting H 3a and H 3b , we find that perceived expertise increases (b = 0.18, p < .01), while perceived risk decreases (b = À0. 02, p < .01), the likelihood of innovation project selection. The coefficient for the interaction between the perceived expertise Notes: We use effects coding for the type of innovation factor (where core = À1 and transformational = 1). We use dummy coding for the scenario presentation factors (i.e., small-range scenario = 1 if small-range scenarios are presented, 0 otherwise; large-range scenario = 1 if large-range scenarios are presented, 0 otherwise). We use dummy coding for gender (gender = 1 for males, 0 for females). a. We computed the significance levels for the main parameters (i.e., the population means) from t-tests. b. We computed the significance levels for the variances of the random coefficients from z-tests. * p < .10, ** p < .05, *** p < .01. and the perceived risk is positive and significant (b = 0.002, p < .05). However, due to the nonlinearity in logit specifications, the sign and significance of the interaction coefficient may not indicate the true direction and true statistical significance of the interaction effect (Hoetker, 2007). Hence, we examine the interaction effects in Table 3 using the simulation-based approach suggested by Zelner (2009). 5 We confirm that the perceived expertise has a positive and significant moderating effect on the effect of perceived risk on project selection likelihood. Columns 5 and 6 of Table 3 present the results of the model described in Eq. (4). We find that the full model, which includes the effects of scenario presentation, type of innovation and an interaction term between these variables, has the best fit (see Table 3; AIC FULL = 12,558 vs. AIC BASELINE = 14,299; BIC FULL = 12,663 vs. BIC BASELINE = 14,339). The full model also has the best predictive validity using a hold-out sample (hit rate of 67.2% for the full model and 61.2% for the baseline model, both of which compare favorably to chance, which is 50%). 6 In line with H 3a and H 3b , the effect of perceived expertise on project selection likelihood remains positive and significant (b = 0.10, p < .01) and the effect of perceived risk on project selection likelihood remains negative and significant (b = À0.01, p < .01). The moderating effect of perceived expertise on the effect of perceived risk on project selection likelihood remains positive and significant (b = 0.001, p < .10).
Discussion. In summary, in Study 1, we find that the effect of scenario presentation on perceived team expertise is more positive for transformational than for core innovations (H 1 ). The effect of scenario presentation on perceived project risk is more negative for transformational than for core innovations (H 2 ). Moreover, we find that perceived team expertise positively affects, and perceived project risk negatively affects, the likelihood of project selection (H 3 ). In addition, smallrange scenarios increase perceived expertise and reduce perceived risk, whereas large-range scenarios decrease perceived expertise and increase perceived risk.
Study 1, however, accounts for the type of innovation as one aspect of the context in which project selection decisions are taken. This raises the question whether our findings from Study 1 generalize to different decision-making contexts. In addition, in Study 1, we make some assumptions in our experimental procedures to enhance its internal validity. These assumptions, however, might somewhat restrict our findings' weight. Thus, we design a second study to address these issues and further strengthen both the robustness and relevance of our findings. Notes: The dependent variable is a binary choice variable indicating whether or not a participant decided to fund an innovation project. Given the nonlinearity in logit specifications, we center Expertise and Risk at the mean and use effects coding for the type of innovation (where core = -1 and transformational = 1) and the scenario presentation factors (i.e., small-range scenario = 1 if small-range scenarios are presented, 0 if large-range scenarios are presented, and À1 otherwise; large-range scenario = 1 if large-range scenarios are presented, 0 small-range scenarios are presented, and À1 otherwise). a. We computed the significance levels from z-tests. * p < .10, ** p < .05, *** p < .01. 5 We explain in detail this approach in Web Appendix D. We check the interpretation of interaction effects using this approach for all logit specifications reported in this paper. All focal results pertaining to the interaction effects hold. 6 To compute these hit rates, we re-estimate our baseline and full models on a calibration sample of respondents' first thirteen choice decisions and predict their remaining two choice decisions.

Study 2
Study 2 aims to replicate the findings of Study 1 with different experimental procedures. Study 2 also extends Study 1 by accounting for two additional aspects of the project selection decision-making context: (i) the method that project teams use to develop scenarios and (ii) the strategic merit of innovation projects as hypothesized in H 4 and H 5 .

Subjects and experimental procedures
Similar to Study 1, we used a large-scale online conjoint experiment with experienced innovation decision-makers to test our hypotheses. We pretested our measures and experimental stimuli on a sample of 192 managers. We contracted ResearchNow to run the fieldwork of the pretest and the full experiment among innovation managers from its online executive panel in the US. We used the same sampling, screening and data cleaning criteria as in Study 1. This resulted in a final sample of 1,680 managers who were 42.1 years of age on average, 31% female, 92.7% holding a bachelor's degree or higher, and an average tenure at their firm of 12.9 years. Participants were highly knowledgeable about innovation in their firm (8.4 on a 10-point scale). They also reported high knowledge and experience of investing in general, 5.3 and 5.2, respectively, and high knowledge of investing in innovation projects, 5.0, on a seven-point scale.
In Study 2, we employed the same experimental procedures as in Study 1 with the following exceptions. First, in Study 1, we deliberately manipulated the type of innovation in a way that core and transformational innovations had the same expected return after five years in the normal-case scenario. In Study 2, we relaxed this assumption as follows. The normal-case scenario of a transformational innovation project had a 25% (rather than 20%) probability of occurring. Consequently, in this study, the expected return of transformational innovations (25% Â $30 M) was higher than that of core innovations (80% Â $7.5 M). Our scenario presentation factor followed the same manipulation as in Study 1 with the exception that we explicitly stated that the worst-and best-case scenarios were equally likely to occur.
Second, we modified our instruction vignette to inform subjects that Telforce would accept that its entire innovation budget might be lost because of investments in projects that were not successful (Web Appendix C). To ensure truthful responses, we informed subjects that the experimental projects were disguised real projects and that we would compare their project selection decisions with the ranking of projects by Telforce's CTO and a senior innovation manager. The best performing subject would be the one whose project selection decisions were closest to this ranking, and we would reward her or him with a $250 Amazon gift card.
Third, besides manipulating scenario presentation (no scenarios, small-range, and large-range scenarios) and type of innovation (core vs. transformational) to replicate our tests for hypotheses H 1 to H 3 , we manipulated two additional factors to test hypotheses H 4 and H 5 : the method of scenario development and strategic merit. We manipulated scenario presentation and type of innovation within subjects, as in Study 1. We manipulated the method of scenario development and the strategic merit between subjects. In other words, each participant was randomly assigned to one of the four conditions based on our 2 Â 2 manipulation of the scenario development method (intuitive vs. analytical) and strategic merit (high vs. low), and had to make project selection decisions among combinations of the six innovation project profiles.
To manipulate the method of scenario development, we provided information whether project teams used their intuition (''intuitive") or conducted a thorough analysis of the different scenarios they presented (''analytical" - Dane and Pratt, 2007;Schoemaker, 1995). To manipulate strategic merit, we varied the extent to which the proposed innovations generated benefits for the firm beyond their financial merit (Ernst, Hoyer, and Rubsaamen, 2010;Noordhoff et al., 2011). Specifically, in a ''high" (''low") strategic merit level, all projects generated a high (low) increase in the sales of other products in the firm's assortment, improvement of current customer relationships, and development of employee talent (see Web Appendix C).
Fourth, in the choice tasks of Study 1, we did not allow a no-choice option. One may argue, however, that in practice decision-makers may sometimes not be willing to fund any project during project selection decisions. To account for this, we allowed a no-choice option in our second study. To ensure design efficiency, we generated Bayesian D-optimal choice design (Sandor and Wedel, 2001). This resulted in 11 choice tasks consisting of three possible choice options in each set: two innovation projects and a no-choice option (see Web Appendix C). We randomized the order of choice sets for each participant.

Measures
We used the same measurement approach as in Study 1 with two exceptions. First, we measured perceived team expertise and project risk before project selection decisions. For each of the 11 choice tasks, we measured participants' perceptions regarding the team expertise and the risk of the two innovation projects in a choice task. Hence, for each participant, we obtained 22 evaluations of perceived expertise and risk. Second, we measured these two variables using a 10-point scale with the same scale type and response format to facilitate participants' task (Web Appendix C). After measuring perceived expertise and risk and obtaining participants' project selection decisions, we inventoried the same control variables as those obtained in Study 1.

Model specification and estimation
We follow the same general model specification and estimation procedures as in Study 1 with the exceptions that we account (i) for the method of scenario development and strategic merit manipulations in all our models and (ii) for the presence of a no-choice alternative in our project selection models (see Web Appendix C for details).

Results and discussion
Replication Tests of H 1 and H 2 . Columns 2 and 3 and columns 5 and 6 of Table 4 present the results of our perceived expertise and risk models. We again find that our perceived expertise and risk models fit the data well (see Table 4; AIC E = 142,040 and BIC E = 142,091; AIC R = 152,060 and BIC R = 152,111) compared to models without random coefficients. The results in Table 4 replicate Study 1 and are consistent with our hypotheses H 1 and H 2 both for small-and large-range scenario presentation. One difference, as compared with Study 1: the effect of large-range scenarios on perceived team expertise is positive (b = 0.65, p < .01).
Replication Tests of H 3 . Columns 2 and 3 and columns 5 and 6 of Table 5 present the parameter estimates of the baseline and full project selection models. We again find that the full model has the best fit (see Table 5; AIC FULL = 29,318 vs. AIC BASELINE = 29,769; BIC FULL = 29,443 vs. BIC BASELINE = 29,814) and the best predictive validity, again using the last two choices as a holdout sample (hit rate of 68.4% for the full and 67.1% for the baseline models, both of which compare favorably to chance, which is 33.33%).
The results in Table 5 replicate H 3a and H 3b . One difference, as compared to Study 1, is that the coefficient for the interaction between perceived team expertise and perceived project risk is positive but not significant in both the baseline and full project selection models.
Method of Scenario Development and Project Selection. In support of H 4a , we find that the effect of perceived expertise on project selection likelihood is more positive when project teams develop scenarios intuitively than when they develop sce- Notes: We use effects coding for the type of innovation factor (where core = À1 and transformational = 1). We use dummy coding for the scenario presentation factor (i.e., small-range scenario = 1 if small-range scenarios are presented, 0 otherwise; large-range scenario = 1 if large-range scenarios are presented, 0 otherwise), method of scenario development (i.e., method of scenario development = 1 if scenarios are developed intuitively and 0 if scenarios are developed analytically), and strategic merit (i.e., strategic merit = 1 for the high level, 0 for the low level). We use dummy coding for gender (gender = 1 for males, 0 for females). a. We computed the significance levels for the main parameters (i.e., the population means) from t-tests. b. We computed the significance levels for the variances of the random coefficients from z-tests. * p < .10, ** p < .05, *** p < .01.
narios analytically (b = 0.04, p < .05; see columns 5 and 6 in Table 5). We do not find a significant moderating effect of the method of scenario development on the effect of perceived project risk on project selection likelihood (b = -0.02, p > .10). Hence, we do not find support for H 4b . Strategic Merit and Project Selection. Supporting H 5a , we find that the effect of perceived expertise on project selection likelihood is less positive for projects that have a high rather than low strategic merit (b = À0.06, p < .01). However, strategic merit does not significantly moderate the effect of perceived risk on project selection likelihood (b = À0.02, p > .10). Hence, we do not find support for H 5b .
Discussion. In summary, Study 2 replicates our tests for hypotheses H 1 to H 3 with different experimental procedures and accounting for two additional project selection decision-making contexts. Study 2 also enriches our insights by showing that the effect of perceived expertise on project selection likelihood is more positive when project teams develop scenarios intuitively rather than analytically (H 4a ), and when projects have low rather than high strategic merit (H 5a ).

Robustness checks and further analyses
One may conceive that the perceived expertise of a project team influences decision-makers' project risk perceptions. To account for this possibility, we re-estimate our perceived risk models controlling for the perceived team expertise (Web Appendix F; AIC = 40,314 and BIC = 40,353 in Study 1; AIC = 152,037 and BIC = 152,089 in Study 2). Our focal results remain robust with one borderline exception. In Study 1, the effect of small-range scenarios on perceived risk (which was significant at the 5% level) becomes insignificant (p = .11).
Similarly, in Study 2, we conduct further analyses to study whether (1) the method of scenario development and (2) the strategic merit of the innovation project moderate the effects of scenario presentation on perceived expertise and perceived risk. We do not find these factors to have an influence on the effects of scenario presentation on perceived expertise and risk (AIC E = 142,050 and BIC E = 142,101; AIC R = 152,073 and BIC R = 152,124).
Finally, we compare the proposed models with different model specifications. First, we run our project selection models without controlling for the interaction term between perceived expertise and risk in project selection models (AIC FULL = 12,560 and BIC FULL = 12,656 in Study 1; AIC FULL = 29,318 and BIC FULL = 29,434 in Study 2). Second, we reestimate our perceived expertise and risk models including an additional variable that indicates the respondents' progress Notes: We center Expertise and Risk at the mean. In line with Haaijer, Kamakura, and Wedel (2001), we account for the presence of no-choice alternative as follows: (i) we use effects coding for the type of innovation (where core = À1 and transformational = 1) and the scenario presentation factors (i.e., small-range scenario = 1 if small-range scenarios are presented, 0 if large-range scenarios are presented, and À1 otherwise; large-range scenario = 1 if large-range scenarios are presented, 0 if small-range scenarios are presented, and À1 otherwise); (ii) we code all attribute levels of no-choice alternative as 0; (iii) we add a no-choice constant that = 1 if an alternative is a no-choice option, and 0 otherwise. We use a random coefficient for the no-choice constant to account for the correlation among utilities for different alternatives that each decision-maker evaluates (Train 2009). We use dummy coding for the method of scenario development (i.e., method of scenario development = 1 if scenarios are developed intuitively and 0 if scenarios are developed analytically) and strategic merit (i.e., strategic merit = 1 for the high level, 0 for the low level). a. We computed the significance levels from z-tests. * p < .10, ** p < .05, *** p < .01.

Follow-up interviews: complementary evidence
To provide direct process-related evidence for our theoretical arguments, we conducted follow-up interviews with senior executives at 11 multinational companies from several industries, all of whom had extensive experience in project selection decisions (Web Appendix A). First, our interviewees suggested that scenario presentation influences project selection through processes that are consistent with, respectively, the positive and negative team expertise attributions and the project risk-reducing and risk-increasing attributions: ''It [scenario presentation] shows, it is not a proof but it shows, that at least the team has done quite a bit of work to get to know what the market is, how the product would behave in the market, and so on. . ." (A global head of innovation at a multinational energy company) [Positive team expertise attribution] ''Sometimes, scenarios might mean that the team is not realistic. . . . It might lead to a perception of the team not having made selections whether they want to start with the project." (A global talent acquisition manager and a former head of the innovation incubator office of a global tire manufacturer) [Negative team expertise attribution] ''The advantage of scenarios is that it . . . makes the project less uncertain. . . When I think of scenarios in a practical sense, it allows you [the decision-maker] to think how to overcome them [the problems]." (An innovation manager at a European food and beverages producer) [Project risk-reducing attribution] ''If you put in front of people a positive and negative case, they will only see the negative. They will not see the positive." (A senior research scientist in an R&D department of an American multinational healthcare company) [Project riskincreasing attribution] Second, consistent with our theoretical accounts, many interviewees believed that scenario presentation is more beneficial for transformational than for core projects. As an interviewee indicated, presenting scenarios in transformational projects is more challenging and, thus, project teams get more credit when they do so: ''If it is an incremental innovation, the level of uncertainty is lower, and the [project] teams are obviously more capable of doing scenarios easily. If you are working in areas of high uncertainty, so more transformational innovations, what we see is that the teams even do not know where to start from, and scenarios require a high level of abstraction." (An innovation manager at a European food and beverages producer) Third, many of our interviewees found project team's expertise to be more important when the project team develops scenarios intuitively than when they develop scenarios analytically: ''If you move to the analytical way, you kind of neutralize the [role of] expertise of the team because you say I am using analytics and more efficient tools." (A global head of innovation at a multinational energy company) Fourth, virtually all interviewees suggested that project team expertise is less influential for projects with high strategic merit. Consistent with our theoretical argument, an innovation manager at a European food and beverages producer explained that the firm buys ''expensive historical data" and supports project teams to survey potential customers for projects with high strategic merit, thus decreasing the importance of project team expertise to decision-makers. The following quote illustrates this insight further: ''For projects in our . . . [strategic areas], I worry less about the expertise of the people of the project because I know I would be more on top of things afterwards." (A chief technology officer at a European online retailer) In sum, our interviews not only buttress our theoretical arguments underpinning our hypotheses, they also enrich our insights. In addition, in our interviews, we extract pertinent managerial implications, which we discuss next.

Theoretical contributions and implications
Our research contributes to the literature on the innovation fuzzy front-end. In this space, we first contribute to the nascent but still scant literature on innovation project selection (e.g., Criscuolo et al., 2017;Girotra, Terwiesch, and Ulrich, 2010). Most studies that focus on the innovation front-end investigate factors that facilitate idea generation, not project selection. We show that presenting financial projections using scenarios influences project selection decisions. In doing so, we highlight the importance of scenario presentation as an important antecedent for project selection and cast new light on growing evidence that characteristics of an innovation proposal (e.g., length and tone, inclusion of demos and prototypes) influence project selection (Lu et al., 2019;Reitzig and Sorenson, 2013). Future research could extend our study by examining how other relevant information presentation choices influence project selection decisions (e.g., disaggregation of revenues and costs, presentation of customer survey results, use of new technologies such as Augmented Reality for pitching, etc.).
Second, by drawing on attribution theory we add a pertinent lens to the literature to examine project selection decisions. We demonstrate that attribution theory can be consistently applied to predict how the presentation of financial projections triggers decision-makers' attributions of project team expertise and project risk which, in turn, influence their project selection decisions. We therefore believe that future research could fruitfully extend attribution theory into other aspects of the ''black box" behind the project selection decision process.
Third, we examine how our theorized relationships between scenario presentation and project selection decisions vary depending on the decision-making context, such as the type of innovation being pitched (Rosenkopf and McGrath, 2011). Prior literature examines the effects of innovation proposal characteristics on project selection but tends to disregard how such effects vary for different types of innovation (e.g., DeRosia and Elder, 2019;Kirsch, Goldfarb, and Gera, 2009). Our results suggest that scenario presentation is more advantageous for transformational than for core innovation projects. This adds to the literature by showing that innovation proposal characteristics do not influence the selection of all projects in the same way. This is important for scholars who might examine how other aspects of innovation proposals also differentially affect (core vs. transformational) project selection.
Fourth, our results suggest that the effects of decision-makers' perceptions on project selection are also contingent on the decision-making context. We found support for the effects of two additional contextual factors: the method of scenario development and the project's strategic merit. Most studies on project selection do not explicitly incorporate the influence of decision-making context. Accounting for the decision-making context thus extends prior work and enriches our understanding of project selection decisions. Specifically, our results suggest that the same project may have systematically different chances of being selected depending on how the context interplays with decision-makers' perceptions. This is pertinent to scholars who might examine other contextual factors that play a role in project selection decisions to enhance the relevance of our scholarly research with respect to inevitably complex settings.

Managerial implications
Our findings offer four substantive implications to managers. First, our findings suggest that firms should help project teams present small rather than large-range scenarios. Small-range scenario presentation increases perceived team expertise and reduces perceived project risk, whereas large-range scenario presentation increases perceived project risk (both studies), though its effect on perceived team expertise is inconclusive. Large-range scenarios may thus lead decision-makers to avoid selecting an innovation project and, instead, default to a ''do nothing" position, which is hardly desirable (Christensen, Kaufman, and Shih, 2008).
Although the implication to present small-range scenarios may seem apparent, it is not evident in a sizable number of firms where we observe that project teams present 'too wide' and 'too extreme' scenarios (57%; Study 2). Our interviewees highlighted several reasons why project teams present large-range scenarios. For instance, some project teams may not be ''focused enough" or may not do ''strong pre-work to validate the assumptions of their project". Project teams may also fail to collect sufficient information on potential customers, possible competitive actions, and major costs of maturing their project which are especially uncertain for transformational innovations, and thus, pose quite a challenge. What then can firms do to make their likely practice come closer to the best practice?
Practically, firms can take specific actions to encourage small-range scenario presentation, often by way of setting expectations and/or providing resources and training. For instance, regarding expectations, we learned from the follow-up interviews that firms can inculcate in their project teams a mindset and/or issue guidance that it pays to ''do their homework" and narrow down the range of expected outcomes. Moreover, firms can offer resources to help project teams collect and analyze the competitive and market data they need to develop small-range scenarios. Firms may also coach project teams on various techniques, such as performing a project pre-mortem, to validate the assumptions driving scenarios and, thus, narrow down the range.
Second, our study suggests that firms seeking to promote transformational innovation should make the presentation of small-range scenarios required for an innovation proposal to be presented to a project selection committee. Indeed, based on our attribution theory predictions, we find that the effect of scenario presentation on perceived team expertise is more positive, and its effect on perceived project risk more negative, for transformational than for core innovations (both studies). Thus, presentation of (particularly small-range) scenarios increases the probability that decision-makers select transformational projects both when decision-makers face a forced choice among projects (Study 1) and when they do not (Study 2). Thus, our attribution-based model informs managers that project teams' behaviors differentially influence decision-makers' attributions and, in turn, their project selection decisions for different types of innovations, which may have systematic consequences on the composition of a firm's innovation pipeline.
Our recommendation is relevant for a sizable number of firms. In practice, decision-makers cannot fund all projects they evaluate due to budget constraints and often need to choose between alternative projects. Moreover, most of our respondents (79%; Study 2) indicate that senior management at their firms would like to select more transformational than core projects to strengthen their innovation pipeline. Yet, at present, only about half of the project teams at their firms always or almost always present scenarios.
Third, we find that projects with an expert team have a higher likelihood to be selected than projects with a less expert team (both studies). One of our interviewees corroborated this assertion. He mentioned that a project team with high expertise ''knows what it is doing" as ''they have expertise and experience under their belt" which leads decision-makers to ''trust their projects." As project teams have less expertise, however, we find that they should especially use analytical scenario development to help increase project selection likelihood (Study 2). Based on our attribution theory account, this happens because the context of analytical scenario development makes decision-makers' attributions of project team's lower expertise less influential in guiding project selection decisions. This is pertinent because novice project teams ''tend to fall in love with their own ideas" and thus experience high levels of positive affect that reduces their ability to engage in analytical thinking (Baron and Ensley, 2006). Thus, project teams with less expertise are especially prone to developing scenarios intuitively. Yet, our results indicate that this is bound to damage the likelihood of their project selection.
Fourth, lack of expertise may prevent project teams from discovering and articulating their project's strategic merit. However, as project teams have less expertise, we find that it is especially important to convey the project's strategic merit to decision-makers in the innovation proposal to help increase project selection likelihood (Study 2).
The latter two points are consequential for both project teams and firms because teams with the least propensity to develop scenarios analytically and ascertain a project's strategic merit depend on these two features for their projects to be selected. For project teams this implies that they should be aware of their ''blind spots" and to try and remedy them. For firms this implies that they should ''level the playing field" for their project teams, as one of our interviewees put it.

Limitations and further research
Our study has limitations, some of which offer pertinent opportunities for future research. First, in our experimental manipulations, we assumed that the CCFs in the normal-case scenario are at the midpoint of the CCFs in the best-and worst-case scenarios. Even though this is a standard practice in decision-making studies (e.g., Ho, Keller, and Keltyka, 2002), it would be worthwhile to focus on the role of skewed scenarios in innovation project selection decisions.
Second, one may ask whether some project teams may interpret the guidance of presenting small-vs. large-range scenarios as an excuse to withhold information about their project's true uncertainty. However, managers we interviewed suggested that such practice would be frowned upon and detrimental for project team reputation, making the practice unsustainable over time. In addition, it is not the purpose of our research to delve into the moral or ethical implications of such practices; we merely discuss the implications of presenting small-or large-range scenarios. Further research and scholarly discourse, however, could fruitfully address this question.
Finally, although we deemed an experiment to be the best possible method given the objectives of our study, it would be interesting to test if the effects of scenario presentation on project selection decisions hold in a retrospective study. However, it is difficult to conceive of a method to make this feasible at sufficient levels of reliability and validity.
Overall, we offer this study as a new look into the rather understudied innovation project selection process. By studying the effect of scenario presentation on innovation project selection decisions, our work offers new insights, though many interesting and important research questions remain. Marketing scholars would be remiss to neglect the role of different types of information that project teams present at different stages of innovation development process.