A literature review and integrated framework for the determinants of crowdfunding success

Crowdfunding is an innovative strategy for financing a new business venture from the general public instead of seeking funds in traditional ways, such as issuing bonds or bank lending. This study aims to identify the determinants affecting the success of a crowdfunding campaign and how different measurements for crowdfunding success, different crowdfunding models, and the selection of subdivided determinants influence the determinants’ impacts on crowdfunding success. We set the disciplines in the search strategy to select studies related to crowdfunding success. Ultimately, 94 empirical papers are selected to reveal the different findings for the determinants of crowdfunding success; based on this information, we construct an integrated framework for future research. There has been much research on project- and creator-related factors; however, many of these factors have inconsistent relationships with crowdfunding success due to varying measurements of success. In particular, different measurements used within the same study for determinants or crowdfunding success may also produce inconsistent results. In addition, different crowdfunding models of a project have been found to induce additional findings. Our review of the determinants of crowdfunding success and the definitions of the determinants, as well as the proposed integrated framework, can help focus future work on relatively new or unique determinants rarely addressed in the existing literature. This work provides practical implications for both theory and practice, and directions for future research.

The first online crowdfunding platform, ArtistShare, was launched in 2001, and users began to create crowdfunding projects in 2003. Since then, according to Massolution (2015), almost 1250 crowdfunding platforms have been launched worldwide. Kickstarter and Indie-GoGo are two famous platforms launched in 2009 and 2008, respectively ). However, not all projects on such platforms can achieve their goals to be successfully funded, especially for Kickstarter. Kickstarter is an "All-or-Nothing" platform in which the entrepreneurs on the platform receive nothing unless the funding goal is achieved ; Da Cruz 2018). The platforms that employ the "Keep-it-All" model (e.g., IndieGoGo and GoFundMe) allow creators to obtain funds even if their projects fail to realize their initial goals ). The three main parties (creators, backers, and platforms) of crowdfunding projects all hope for crowdfunding success. Creators hope to achieve the funding goals to get the money to carry out their business (Frydrych et al. 2014). Backers hope to successfully support a crowdfunding project to obtain material or spiritual benefits ( Johnson et al. 2018). In addition, prior research has conducted a literature review on crowdfunding and its success, which mainly focuses on reviewing the categories of crowdfunding and its success, and explaining the determinants of crowdfunding success within several types (Moritz and Block 2016; Popescul et al. 2020; Yuan et al. 2016). The most closely related review work on crowdfunding success in our study is the research of Kaartemo (2017) and Shneor and Vik (2020). Specifically, Kaartemo (2017) identifies and reviews four main determinants of crowdfunding success: project-, creator-, backer-, and platform-related factors. It synthesizes and evaluates the findings in empirical research by providing examples of these determinants. The researcher explains the effect of each determinant on crowdfunding success by reviewing the research findings of each representative study. Shneor and Vik (2020) identify the general trends and research gaps concerning independent variables based on each primary crowdfunding model (i.e., reward-, equity-, loan-, and donationbased) separately. Then, they further build several integrated frameworks for the influential independent variables based on each main crowdfunding model. The independent variables have similar measurements and persistently significant effects in a direction.
Our work serves as both a complement and a contrast to prior research in several ways. First, we aggregate the definitions and measurements of the determinants, as well as the effect of each determinant on crowdfunding success (i.e., positive, negative, or nonsignificant effects) within four types of factors (i.e., project-, creator-, backer-, and platform-related factors). Second, considering that different measurements of crowdfunding success and different crowdfunding models may lead to additional findings for the effect of a determinant, we identify how the factors influence the performance of project fundraising by considering different measurements of success and different crowdfunding models. Third, we subdivide some determinants and investigate whether different definitions of a determinant could influence the findings. Finally, we synthesize the relevant findings into a general framework that can assess the determinants of crowdfunding success. The framework comprises the platform and crowdfunding models, the main measurements of crowdfunding success, the classification of the determinants, the research methods, and the gaps in need of future attention. More importantly, it shows how different determinants affect crowdfunding success and how the measurement of crowdfunding success determines the research method. This study presents a comprehensive understanding of achieving crowdfunding success, which fills several research gaps by analyzing mixed empirical findings on the determinants of crowdfunding success.
We first plan the review by setting the research goals, literature search strategy, and selection criteria of the studies according to Hossain et al. (2019) and Leidner (2018). Ultimately, 94 empirical studies are selected. We then conduct a review and report our findings by assessing the measurements of crowdfunding success, the choice of research methods, the platform involved in each study, the sample size of each study, the definition of each determinant, and the findings regarding the determinants' impacts on crowdfunding success. Finally, we propose an integrated framework to show the current research and gaps in future attention.
Our work yields the following key findings. First, there are eight main methods for measuring the success of crowdfunding projects. The four most widely used measurements include the dichotomous variable Funding Success and the continuous variable Funds Raised, Success Ratio, and Number of Backers. The other measurements include Time to Funding, Pledge/Backer Ratio, Decision to Invest, and Overfunding. A specific measurement determines the choice of research method in that logistic/logit regression and probit regression are widely used in the literature on Funding Success. In contrast, linear regression is used for continuous measurements such as Funds Raised and Success Ratio. Second, we find that most studies widely analyze project-and creator-related determinants, whereas platform-and backer-related factors rarely appear in most studies. They warrant more intensive study in future research. Third, in different studies or even within the same paper, various measurements of crowdfunding success, different crowdfunding models, and different subdivided definitions of a determinant can lead to additional findings for the determinants' impacts on crowdfunding success.
Following is an outline of the remainder of this paper. "Research methodology" section describes the research methodology, including the research goals, search strategy, and inclusion and exclusion selection criteria. "Overview of selected studies" section provides an overview of the selected studies, including literature identification, measurements of crowdfunding success gleaned from the available literature,

Research methodology
Given the popularity of crowdfunding platforms among small entrepreneurs, it is essential to comprehensively understand the determinants of crowdfunding success to promote capital resource allocation efficiency. To this end, following the guide- , we conduct a literature review on the determinants of crowdfunding success in the following steps. First, we plan the review by setting our research goals, search strategy, selection criteria, and the analysis procedure of the selected literature. Second, we review, report our findings, and construct an integrated framework for research on crowdfunding success. Figure 1 presents the specific stages and detailed steps of our study.

Research goals
Following Leidner (2018), we adopt an assessment review to overview the existing literature on crowdfunding success and identify the relationships between the determinants and crowdfunding success studied. We also assess the different research methods that different studies use and key determinants of crowdfunding success that different studies focus on to identify the gaps in need of more future attention. Specifically, we aim to answer the following four questions: (1) How can crowdfunding success be measured? (2) What determinants can influence the success of a crowdfunding project, and how can these determinants be classified? (3) Do different measurements of crowdfunding success and different crowdfunding models result in different findings on the relationships between the determinants and crowdfunding success (i.e., positive, negative, or nonsignificant effects)? (4) Do different definitions of a determinant result in different findings regarding the relationship between the determinant and crowdfunding success?

Search period
First, we identify the search period. Given that the first crowdfunding platform (Artist-Share) was launched in 2001, we set the search period from 2001 to 2021.

Search terms
According to Hossain et al. (2019), key phrases are essential for a literature review. For this study, we use crowdfunding-related words (i.e., crowdfund, crowd-fund, crowd fund, crowd funder, crowd-funder, crowdfunding, and crowd funding) combined with success-related words (i.e., success, succeed, successful, performance, and outcome) as search terms to search for crowdfunding success-related studies.

Search spaces
We choose web search engines, databases, journals, and authoritative conferences to search for relevant literature as follows: (1) we search for relevant papers in web search engines, such as Google Scholar and Baidu Scholar, and several digital databases, including ScienceDirect, Web of Science (SCI and SSCI databases), EBSCOhost, and INFORMS, for journals, conference proceedings, working papers, theses, reports, and books; (2) we search the Social Science Research Network (SSRN) to avoid omitting the latest research that has been disclosed online in advance; (3) we also manually search nine journals in which the related works are most likely to appear, namely, Administrative Science Quarterly, Decision Support Systems, Entrepreneurship Theory and Practice, Information Systems Research, Journal of Business Venturing, Journal of Management Information Systems, Management Science, MIS Quarterly, and Organization Science, to ensure that we do not miss important relevant papers, and (4) we search papers from the proceedings of authoritative conferences in the fields of information systems and entrepreneurship, such as International Conference on Information Systems (ICIS), Americas Conference on Information Systems (AMCIS), International Conference on Innovation and Entrepreneurship (ICIE), and European Conference on Innovation and Entrepreneurship (ECIE).
Notably, the search scope for the papers in each web search engine, database, journal, or conference depends on the search restrictions of the different search sources. For example, Google Scholar, Web of Science, and JSTOR allow us to search for the literature in all fields, while INFORMS and Taylor and Francis Online permit searching only by the title and abstract. Appendix 1 shows the sources of the studies and corresponding search scopes.

Inclusion and exclusion criteria
As we aim to provide a deep understanding of the significance and direction of the effects of the determinants on crowdfunding success, we set the inclusion and exclusion selection criteria to filter empirical papers and thus achieve the research goals of this study. Specifically, we set our inclusion and exclusion criteria according to Muller et al. (2019) and Qazi et al. (2017) as follows:

Inclusion criteria
The inclusion criteria are as follows: (1) the papers are related to the search items and keywords that we have described in "Search strategy" section; (2) peer-reviewed journals and conferences as well as working papers, theses, reports, and books, are all included; (3) only papers in English are included, and (4) full-text availability is essential. After screening the titles, abstracts, and keywords and deleting duplicates, we exclude papers that fail to meet the inclusion criteria and complete the first screening.

Exclusion criteria
After obtaining the relevant papers according to the inclusion criteria, we carefully read the full text of each article and apply the following exclusion criteria to select empirical papers that meet our research goals: (1) Studies without an empirical approach are excluded. Theoretical and review papers on crowdfunding and crowdfunding success are excluded. For example, we exclude the papers by Kaartemo (2017) and Popescul et al. (2020) based on this exclusion criterion because they are both review papers on the determinants of crowdfunding project success. (2) Papers that do not investigate the empirical effect (i.e., positive, negative, or nonsignificant effect) of a determinant on the success of crowdfunding projects are excluded. For example, we exclude the papers by Ryoba et al. (2020) and Huang et al. (2021) because they fail to study the empirical effect of the determinants of crowdfunding success. For example, Huang et al. (2021) is excluded because this study conducts sufficiency analysis to examine how multiple signals of entrepreneurs' credibility and project quality work together to produce crowdfunding success while failing to test each factor's significance. We also exclude Du et al. (2015) and Schraven et al. (2020). Although these two studies conduct empirical research and consider some determinants of crowdfunding success, they use the determinants only to predict crowdfunding performance rather than to investigate the influence of each determinant. (3) Cases, surveys, and experimental studies that explore the determinants of crowdfunding success in an offline context rather than based on a crowdfunding platform are excluded (Shneor and Vik 2020). This exclusion criterion is based mainly on three considerations. First, one purpose of our research is to develop a deep understanding of the crowdfunding platform involved in each study and survey the general trend of the platform on which crowdfunding projects and existing research are based. Second, we endeavor to ascertain how the platform as a determinant empirically influences the success of a crowdfunding project. Third, we also aim to understand the determinants of crowdfunding success across different crowdfunding models (i.e., reward-, equity-, loan-, and donation-based crowdfunding projects), which are associated with platform features. For example, we exclude Lacan and Desmet (2017) according to this criterion because they collect data through an online survey rather than utilizing data from a real platform.
To assess the reliability of our inclusion and exclusion criteria, we apply Cohen's kappa statistic to check the level of agreement between the inter-raters (Pérez et al. 2020;Viera and Garrett 2005). Specifically, two research assistants responsible for selecting the articles independently rate a randomly selected sample of 50 articles according to the selection criteria. Their judgments are then analyzed using Cohen's kappa statistic. Finally, we obtain a result of 0.797, which reflects an "almost perfect agreement" level (Pérez et al. 2020) between the judgments of the two assistants, confirming that our selection criteria are acceptable.
Based on the inclusion and exclusion criteria, we complete the literature screening and select 94 empirical articles in total for our analysis in this review.

Analysis of selected literature
At this stage, the selected articles are carefully read. Related data are manually extracted and coded into a database including the following elements: author, year, title, literature type (i.e., journal, conference, working paper, thesis, report, or book), literature source (e.g., journal name or conference name), data source (i.e., platform), platform model (i.e., "All-or-Nothing" or "Keep-it-All"), crowdfunding model (i.e., reward-, equity-, loan-, or donation-based crowdfunding projects), sample size, the measurement of the dependent variable, the measurement of independent variables, identified associations between the dependent and independent variables (i.e., the significance and the directions of the effects), and the research methods. After data extraction and coding, we conduct a conceptual aggregation for the measurements of crowdfunding success, determinants of crowdfunding success, and research methods. We also aggregate the determinants' effects on crowdfunding success by considering different measurements of crowdfunding success and different crowdfunding models. "Overview of selected studies" and "Findings" sections explain the details.

Literature identification, publication outlets, and trends
Our literature search results in a total of 94 empirical papers on the determinants of crowdfunding success for our literature review (Appendix 2 shows the list of 94 reviewed papers). Table 1 shows the main sources of the literature used in this study. The numbers of journal papers and conference papers are 80 and five, respectively. In addition, nine working papers, theses, reports, and books are included. The Journal of Business Venturing (13 articles) and Entrepreneurship Theory and Practice (10 articles) have published the most relevant papers in the search period. To understand the general research domains of the selected studies, we manually search for each journal involving the selected papers in Web of Science. Some journals cover more than one research domain. For example, MIS Quarterly covers computer science, information science and  Table 2 shows the temporal distribution of the selected papers, indicating that the papers analyzed in our study are from 2011 to 2021. The determinants of crowdfunding success have been prevalent in the academic community since 2014.

Measurements of crowdfunding success
Crowdfunding is defined as acquiring financial support from the crowd for various special tasks through the Internet and providing a product, equity, reward, or interest for the funders after the project's success (Belleflamme et al. 2014). Individuals or groups that need financial support from the crowd can create a project, post information about it on a crowdfunding platform, and receive backing. Accordingly, crowdfunding success is defined as success in fundraising for a project on a crowdfunding platform, which depends on the funding model (i.e., "All-or-Nothing" or "Keep-it-All") employed by the platform (Yuan et al. 2016). Numerous scholars have investigated the determinants of crowdfunding success by using various measurements to meet their research purposes. Table 3 shows the measurements of crowdfunding success used in the 94 selected empirical studies.
The 94 selected empirical studies use eight main ways to measure crowdfunding success. Most of them (57 papers) use Funding Success, which means achieving the funding goal, as the measurement of crowdfunding success. It is a dummy variable: if a project reaches its funding goal within the given time, it is coded as "1"; otherwise, it is coded as

Use of research methods
The definition and measurement of crowdfunding success mainly determine the choice of research method.

Platforms involved
As mentioned above, different platforms adopt different platform models and crowdfunding models. A platform may follow an "All-or-Nothing" or "Keep-it-All" model or even apply a rule that mixes "All-or-Nothing" with "Keep-it-All". In addition, crowdfunding projects can follow one of four crowdfunding models: reward-, equity-, loan-, and donation-based projects (Burtch et al. 2013), determined by the platform rule. Table 5 lists the platforms used in the 94 selected studies. 1 As shown, more than 20 platforms are explored in the selected papers. Eighty-six out of 94 studies collect data from "Allor-Nothing" platforms, and only 3 collect data from "Keep-it-All" platforms. Those platforms mixing the "All-or-Nothing" rule with the "Keep-it-All" rule are investigated in six papers. Most of the studies considered in our review focus on reward-based crowdfunding model (79 papers) and 19 studies examine projects from equity-based platforms. Notably, the most widely studied reward-based platform is Kickstarter (53 papers), and the most widely studied equity-based platform is Crowdcube (eight papers), both following the "All-or-Nothing" rule. The project creators on these two platforms must follow the "All-or-Nothing" rule, under which the failure to reach the funding goal means that the creator cannot obtain the funds raised, thereby rendering the project unsuccessful (Parhankangas and Renko 2016). IndieGoGo, studied by five papers, is a platform mixing the "All-or-Nothing" rule with the "Keep-it-All" rule. The creators on this platform can choose to follow one of the rules and can keep the funding raised even if the project fails to reach its funding goal if they choose to follow the "Keep-it-All" rule ). Some studies have investigated crowdfunding projects using more than one platform.  and GoFundMe to explore whether the platform type (i.e., under the "All-or-Nothing" or "Keep-it-All" rule) can influence the success of a crowdfunding project.

Sample size
The sample size of each selected article ranges from 50 to 403,445 observations, which could be related to the research period.

Definitions and measurements of determinants of crowdfunding success
According to the extant literature, we identify four types of determinants of crowdfunding success according to , Kaartemo (2017) and : project-, creator-, backer-, and platform-related factors. Table 6 lists the definitions and measurements of the determinants. We also display the direction of influence of each determinant involved in the selected articles.

Project-related factors
Project-related factors reflect project characteristics and the soft information associated with the project . Specifically, project characteristics are the characteristics of the funding campaigns and products. As shown in Table 6, the commonly investigated project characteristics include the funding goal (Goal), the amount of early funds that the project has received before an additional backer's investment (Early Funds), the number of early backers before an additional backer's investment (Early Backers), the number of total backers at the end of the funding period (Total Backers), the total funds raised divided by the total number of backers (Pledge/Backer Ratio), the total funding duration (Duration), the reward promised (Reward), the creator's team size (Team Size), the launch date-related variables (Launch Date Related), the need similarity (Need Similarity), the innovativeness of the product being funded (Innovativeness), the number of pledge levels (Pledge Level), the level of risk (Risk Level), the use of the funds raised (Use of Funds Raised), the percentage of equity offered in the campaign (Equity Offered), the expected outcome for the projects (Expected Outcome), and the category to which the project belongs (Category). Among the 94 selected papers, the most commonly examined project characteristics are Goal (78 papers), Duration (56 papers), Category (56 papers), Reward (27 papers), and Team Size (19 papers). Goal and Duration generally negatively impact crowdfunding success, while Reward and Team Size could positively influence crowdfunding success. Studies investigating the Category effect usually consider it as a control variable by adding a set of industry category dummies to the models (41 papers). However, some studies also explore the main effect of Category on crowdfunding success (17 papers). For example, Greenberg and Mollick (2017) investigate the effect of different industry categories (i.e., publishing, fashion, games, and technology) on the success of crowdfunding projects and find that publishing and fashion-related projects are more likely to succeed in fundraising. According to , soft information about fundraising projects should also be considered. It is the information provided to inform the crowd about the project that visitors and backers can access during and after the project and the knowledge that creators commonly update on the project page to attract more backers. The description of a project is the most critical soft information that potential backers need to understand a project and make their final decisions ), which includes text-related factors (Text) and visual-related factors (Visual). The text's narrative characteristics, such as text quality, readability, sentiment, word count, spelling errors, linguistic style, and specific terms, can influence backers' understanding of a project, and 43 papers investigate this factor. Forty-seven papers explore visual-related factors such as video-and image-related factors. Some studies examine videos or images separately  The level of risk [75] 1 Negative [75] 1 Non-significant [75] 1 Use of funds raised RandD, working capital, marketing and market expansion [80] 1 Positive [80] 1 Non-significant   1 Sexual orientation A dummy variable coded "1" if a creator identified as lesbian, gay, bisexual, transgender, or queer; otherwise, it is coded "0" [53] 1 Negative [53] 1 Non-significant [53] 1 User entrepreneurship The entrepreneurs who create and commercialize the new product or service are also those who use that product or service [46] 1 Positive [46] 1  1 Non-significant [7] 1 Platform tenure A backer's platform tenure on a given platform [7] 1 Non-significant  (27 papers), which can inform the crowd about the newest information and process of the projects. Social Network refers to external links to social networks (e.g., Facebook, Twitter, or other community websites). Updates reflect the number of updates made by the project creators during and after the funding period. Comment measurement varies across papers, and comment quantity is the most widely used measurement (25 papers). The extant literature has also investigated comment length, comment sentiment, comment replies, the availability of comments, and previously created/ backed comments. These determinants related to soft information are generally found to positively affect the success of crowdfunding projects. In addition, soft information such as the presence of "staff pick" quality tags provided by Kickstarter (Staff_pick), the number of shares on Facebook (Shares), the recommendations by the platform (Recommend), the "like" count of a project (Likes), media coverage on the project (Media Coverage), the signals of projects and products (Signals), the total number of followers of a project (Followers), and the questions asked and answered on campaign pages (FAQs) are also found to influence crowdfunding success. These factors, which are triggered by either the platform or the backers, reflect a project's popularity or importance and are thus considered as soft information about a project.

Creator-related factors
Creator-related factors are those associated with the individual, entrepreneur, or firm that creates the project. Forty-three papers examine project creators' previous (or previous successful) creation and backing experience or entrepreneurial experience (Experience). Other common factors in most of the empirical studies include geographic distance or dummies (Geography, 24 papers), creators' gender (Gender, 22 papers), number of creators' Facebook friends (Facebook Friends, 20 papers), whether a creator is a group or an individual (Creator Type, 12 papers), the level of creators' education (Education, nine papers), and creators' race (Race, seven papers). Most studies investigating the effect of Race explore whether being Caucasian can influence crowdfunding performance. The culture, language, patent ownership, sexual orientation, credibility, and the picture/logo of creators can also affect their fundraising outcomes. In addition, creators' Preparedness, Passion, Innovativeness, and User Entrepreneurship toward the project or product also determine crowdfunding success. Researchers rarely investigate Entrepreneur Aspect, Brand Prominence, Firm Age, Prior Funding, and Diversification related to the entrepreneur or the firm that creates the projects. Most of these factors are measured using different methods, and the findings regarding their relationships with crowdfunding success are inconsistent across studies. In contrast, some of these factors have only been examined by a few researchers. In particular, models include some of these factors (i.e., Culture, Geography, Race, Language, Creator Type, and Gender) as control variables rather than as independent variables. Furthermore, some studies investigating the main effects of these factors tend to conduct group tests or use different measurements in one model.

Backer-related factors
Backer-related factors are those associated with the people who back the projects, including Funder's Positive Affective Reactions, Motive, a backer's previous backing experience (Experience), a backer's tenure on the platform (Platform Tenure), and geography (Geography). Among the 94 studies, only 6 investigate these factors.

Platform-related factors
Platform-related factors investigated in the selected articles include the number of projects being funded on the platform (Competition), the type of platform (Platform Type), and the number of years since the platform's establishment (Platform Age). Seven papers explore the effect of platform type on crowdfunding success. Different crowdfunding platforms have additional requirements for crowdfunding projects (e.g., the standard practice of retaining funds received from the crowd at the end of a project, i.e., "All-or-Nothing" or "Keep-it-All"), which to some extent also affect the projects' success

Effects of the determinants on crowdfunding success based on different measurements of crowdfunding success
Different measurements of crowdfunding success may lead to different findings regarding the effects of its determinants. We consider each measurement of success separately to identify how these factors influence project fundraising performance. We also subdivide some determinants based on the reviewed papers and investigate whether different definitions of a determinant can induce different effects on crowdfunding success. Specifically, we collate and classify the empirical results (i.e., positive, negative, or nonsignificant effects) of each determinant of crowdfunding success for each paper. We mainly focus on Funding Success, Funds Raised, Success Ratio, and Number of Backers, which are widely used in the selected papers. Some studies yield inconsistent findings regarding the relationship between the same determinant and crowdfunding success (see Table 7). In particular, using different measurements within the same study for determinants or crowdfunding success can yield different results for the same determinant.

Funding success as the measurement of crowdfunding success
As outlined in the previous section, more than half of the selected studies use Funding Success as a proxy for crowdfunding success; thus, the list of determinants involved in these studies is comprehensive and diverse. As shown in Table 7 . Thirty-seven papers examine the impact of Category on crowdfunding success, with 11 papers investigating its main effect and 28 papers considering it a control variable. We do not compile the empirical impact of Category, as each paper has a different research orientation. For example,  explores the difference between environmentally oriented projects and projects focusing on other aspects and finds that environmentally oriented projects are more likely to succeed.
Second, we focus on the different definitions of the determinants that have conflicting effects on crowdfunding success. Commonly used reward-related proxy variables (Reward) include reward quality, number of reward levels, reward type, reward quantity, and reward availability. The number of reward levels and reward availability both yield consistent results that projects providing more reward levels are more likely to succeed (seven papers), while whether the project offers a reward is found to have no relationship with crowdfunding success (two papers).  use two methods to investigate the impact of Reward and find that reward quality positively affects crowdfunding success, while reward quantity has a negative effect. Butticè et al. (2017) use a set of reward dummies to test the effect of reward type on the success of crowdfunding projects and find that community-belonging rewards do not influence crowdfunding success, while rewards trigger backers' motivation and offer customized products can attract individuals' funds. For text-related factors (Text), word count is the most commonly studied factor (18 papers), more than 70% of which have a positive effect. In addition, studies testing the effects of text readability, text sentiment, spelling errors, linguistic styles, and specific terms have also yielded inconsistent results.              Keep-it-all or All-ornothing control [13] 1 Platform age Positive [16] 1 Negative [16] 1 Non-significant [16] 1 [16] and only two papers find that it has a negative effect. However,  use two different approaches to measure creators' Experience and obtain consistent results that the ratio of previous successful backed projects and the number of previously created and backed projects both positively affect the success of crowdfunding. Some categorical variables (i.e., Launch Date Related, Geography, Race, Language, Creator Type, Gender, Firm Age, and Platform Type) yield inconsistent results due to different categorical methods, while some have a clear impact. For example, Caucasian and female creators are more likely to attract crowdfunding. In contrast, black and male creators are less likely to succeed in crowdfunding, and individual creators are less likely to succeed than their company or group peers. The Platform Type, namely, the rule of "Keep-it-all" or "All-or-nothing" seems to have no effect on crowdfunding success. Third, the rest of the determinants are factors on which the results yielded inconsistency by different studies, even with the same definition criteria. However, we conclude that, in general, some determinants have a consistent effect on crowdfunding success. For example, most studies find that Goal and Duration negatively influence crowdfunding success. Fifty-one papers explore the impact of Goal, among which 42 find a negative relationship between Goal and crowdfunding success. Twenty-three out of the 38 papers investigating Duration have a negative effect. In addition, previous studies have found that Total Backers (3 papers), Team Size (11 papers), Social Network (10 papers), Updates (18 papers), Staff_pick (6 papers), and Facebook Friends (12 papers) commonly have a positive impact on crowdfunding success.

Funds raised as the measurement of crowdfunding success
Studies using Funds Raised to measure crowdfunding success consider all four types of determinants (see Table 7). Backers' motive (Motive) and the competition level on the platform (Competition) are only explored by studies using Funds Raised as the measurement. We first focus on the determinants examined in only one paper. Text sentiment and word count could positively affect crowdfunding success, while text readability, spelling errors, and specific terms negatively correlate with success. Therefore, we conclude that when Funds Raised is used as the measurement, Visual and Text tend to positively affect the outcome of crowdfunding success (which is consistent with the findings based on Funding Success), and Reward is found to have a mixed effect. In addition, the creators' Experience and Geography reflect conflicting results. Boeuf et al. In summary, some factors yield inconsistent results between the studies based on Funding Success and those based on Funds Raised, indicating that the measurement of crowdfunding success is a critical contextual factor for the different research results. For example, Goal and Duration are the most common factors investigated in the literature on crowdfunding success. Studies based on Funding Success tend to have a negative effect, while those based on Funds Raised tend to find a positive or nonsignificant impact.

Other measurements of crowdfunding success
As shown in Table 7, for the studies based on the measurements of Success Ratio and Number of Backers, we find that most of the determinants have mixed effects or are examined by only a few researchers. However, we can still conclude several rules from these studies for some widely studied determinants. Pledge Level, Social Network, Updates, Comment, Staff_pick, Shares, Likes, Facebook Friends, and Credibility tend to have a positive effect, consistent with the findings of Funding Success and Funds Raised. The studies examining Goal on Success Ratio find the same negative effect on Funding Success. In contrast, those on Number of Backers find a positive effect inconsistent with those for the other measurements of crowdfunding success. For Duration, unlike the studies using Funding Success that find a negative effect, mixed-effects are found in studies using the Success Ratio and Number of Backers, which is consistent with the findings from the studies using Funds Raised. Different definitions of some determinants (i.e., Reward, Text, Visual, and creators' Experience) have conflicting effects on crowdfunding success, inducing difficulty in identifying consistent rules for them. Taking the factor, Text, as an example,  document the role of narrative styles (i.e., "Results in progress" and "Ongoing journey") in explaining the success ratio of crowdfunding and finding a positive relationship between them. Duan et al. (2020) explore the effect of narrative styles in terms of readability, length, tone, and uncertainty on Success Ratio and Number of Backers, and demonstrate a positive effect of length and tone, whereas readability and uncertainty have a negative effect.
Studies using Time to Funding, Decision to Invest, Pledge/Backer Ratio, and Overfunding as the measurement of crowdfunding success account for a minority. Except for Time to Funding, which is negatively related to crowdfunding success, all measurements refer to success. Most of the determinants in these studies are examined by only one paper, while the others are found to have mixed effects. However, we analyze the empirical results for the effects of these factors on crowdfunding success. It is challenging to conclude consistent rules from these studies. Most of them yield inconsistent results for the same factors under different or even the exact measurements of crowdfunding success. For example, Chan and Parhankangas (2017) and  both find that Goal positively affect Pledge/Backer Ratio and Decision to Invest, respectively. However, one study find that Goal can negatively affect Overfunding, while the other finds that Goal does not influence Overfunding. Moreover, three papers document a positive relationship between Goal and Time to Funding, while two papers demonstrate a negative and nonsignificant effect of Goal on Time to Funding, respectively. These studies find no evidence for exactly how Goal affects crowdfunding success, as the results are inconsistent. However, in the above section, we conclude that Goal tends to negatively affect Funding Success and Success Ratio but is less likely to positively affect Funds Raised and Number of Backers.

The findings from studies with multiple measurements of crowdfunding success
Among the 31 papers mentioned in "Use of research methods" section and Appendix 3, use Funding Success and Funds Raised as proxies for crowdfunding success and find that Goal, Duration, and Text have different relationships with Funding Success and Funds Raised. More specifically, Goal and Duration can negatively affect Funding Success but have no relationship with Funds Raised. In contrast, word count positively affects Funds Raised but does not affect Funding Success. Therefore, we conclude that different measurements of crowdfunding success can lead to different findings regarding the impact of a determinant.

Effects of the determinants on crowdfunding success based on different crowdfunding models
As mentioned in "Platforms involved" section and the involved platforms shown in Table 5, the studies exploring loan-based (eight papers) and donation-based (two papers) crowdfunding models account for a minority. It is challenging to conclude consistent rules. Therefore, in this section, we mainly focus on research using reward-and equity-based samples to review the effects of the determinants on crowdfunding success, respectively (see Table 8). As shown in Table 8, the determinants studied using reward-or equity-based samples exhibit a large difference. Except for the determinants that are commonly studied with both samples, for example, Goal, Early Funds, Early Backers, Total Backers, Duration, Team Size, Category (i.e., project characteristics); creators' Experience, Education, Patent Ownership, and Creator Type (i.e., creator-related factors); backers' Experience (i.e., backer-related factors); and Platform Type (i.e., platform-related factors), other determinants are only investigated with reward-or equitybased samples.
First, different effects are still found for the determinants studied by both samples. Goal is concluded to have a negative effect in reward-based studies, while an approximately nonsignificant effect is found in equity-based studies. Creators' Experience is commonly found to positively affect the success of crowdfunding projects in rewardbased studies, while in equity-based studies, a mixed effect is found. Education has no significant effect in reward-based studies, while a mixed effect is found in equity-based studies.
Second, we focus on the determinants that have only been explored in equity-based studies. We find that Use of Funds Raised, Equity Offered, and Expected Outcome related to a project, and Firm Age, Prior Funding, and Diversification related to the fundraising firm have mixed effects on crowdfunding success. Only three studies examine the effects of the determinants related to soft information about fundraising projects. Specifically, Mamonov Table 6, are ignored in equity-based studies.
Third, except for the determinants examined only in equity-based studies, all the other determinants displayed in Table 6 are investigated in reward-based studies. The significance and direction of the determinants' effects on crowdfunding success in these studies are roughly similar to those shown in "Definitions and measurements of determinants of crowdfunding success" section and Table 6.
In summary, we conclude that most of the research among the 94 selected papers focuses on exploring the effects of the determinants on the success of reward-based crowdfunding projects. The determinants considered in reward-and equity-based studies are roughly different, and even those determinants considered in both crowdfunding models are found to have different effects. Different crowdfunding models of projects can lead to additional findings regarding the impact of a determinant on crowdfunding success.

An integrated framework for the determinants of crowdfunding success
According to our review of its determinants, Fig. 3 depicts an integrated framework that reflects current and future research on crowdfunding success. On the one hand, the integrated framework comprises the platform and crowdfunding models, the classification of determinants, the main measurements of crowdfunding success, the research methods, and the gaps in need of future attention. On the other hand, it shows how different determinants affect crowdfunding success and how the measurement of crowdfunding success determines the research methods.    1 Non-significant [13,66]          1 Non-significant [16] As displayed in Fig. 3, the main measurements of crowdfunding success in the extant literature are Funding success, Funding Raised, Success Ratio, Number of Backers, Time to Funding, Pledge/Backer Ratio, Decision to Invest, and Overfunding, and the concept of each measurement is shown in detail. Among these measurements, Time to Funding is in contrast to crowdfunding success, as the longer the fundraising time, the less successful a project is. Therefore, the determinants' impacts may differ when adopting this measurement. In particular, Funding Success, Funding Raised, Success Ratio, and Number of Backers are the most widely adopted measurements in current research, while other measurements are the minority. Moreover, Funding Success is adopted in more than 60% of the 94 selected papers. A possible explanation is that 86 out of 94 papers use data from "All-or-Nothing" platforms, which, to a large extent, enable researchers to set a dummy variable to reflect success. Considering that other measurements can also reflect the success of crowdfunding projects but are rarely studied by extant research and that different measurements of crowdfunding success can lead to different findings; we argue that future work could adopt these rarely used measurements of crowdfunding success to obtain new insights. As mentioned above, more than 90% of the papers use data from platforms following the "All-or-Nothing" rule. Therefore, ample possibilities exist to obtain more novel findings by using the data from some niche crowdfunding platforms, especially the platforms following the "Keep-it-All" rule or mixed rules that combine "All-or-Nothing" and "Keep-it-All". In addition, we also find that a handful of papers conduct surveys or experimental studies, which can create a deep understanding of individuals' funding and backing behaviors and even the operation mode of crowdfunding platforms. Hence, we propose that future research should conduct surveys and Most studies adopt a linear regression model to test the effects of the determinants on crowdfunding success, except for the measurement of Funding Success. In particular, the dichotomous variable Funding Success is widely used in logistic/logit regression and probit regression models. Notably, survival analysis is mostly conducted in the research for Time to Funding, tobit regression for Success Ratio, and negative binomial regression for Number of Backers, reflecting that the measurement of crowdfunding success determines the research method. As presented in the framework in Fig. 3, there is a need to conduct studies using different measurements of crowdfunding success and different research methods. According to our review, 31 out of 94 papers use multiple measurements of crowdfunding success, which clearly shows the influencing mechanism with different measurements and research methods (see Table A3 in Appendix 3).
The framework classifies the factors that are determinants of crowdfunding success into four categories: project-related factors (associated with project characteristics and soft information), creator-related factors, backer-related factors, and platform-related factors. Among these determinants, only five and three are considered in the studies investigating backer-related (six papers) and platform-related (ten papers) factors, respectively. Thus, the determinants of backers and platforms may require more attention. Moreover, many determinants related to projects or creators are still required further investigation. According to our review, there are both consistent and inconsistent findings regarding the impact of crowdfunding success determinants. We find that the same factor can yield inconsistent results due to different measurements of crowdfunding success (e.g., Goal and Duration have roughly negative effects in studies focusing on Funding Success, but have mixed effects in studies focusing on Funds Raised). Those determinants with subdivided definitions such as Reward, Text, Visual, Comment, and creators' Experience can also have conflicting effects on crowdfunding success. We argue that the determinants with inconsistent effects warrant future research attention, especially from the perspective of the measurements of crowdfunding success and the definitions of the determinants, rather than merely investigating their significance and the directions of the effects.
In addition, we discover that the selected papers rarely use text, image, and video mining techniques. These widely used techniques in the business field can be applied to examine the impacts of determinants such as Text, Visual, and Comment on crowdfunding success, thereby extending the research to behavior and psychology. As can be concluded from the 94 selected papers, the widely studied determinants related to Text, Visual, and Comment include word count, visual quantity and availability, and comment quantity and availability. In contrast, the determinants that need mining techniques (e.g., readability, sentiment, linguistic styles, and specific terms of text and comments that require textual analysis, and the quality and valence of visual analysis that require recognition technology) account for a minority. Moreover, with the development of the Internet and networks, the content on social networks is increasing dynamically, making it difficult to identify potential determinants of crowdfunding success by utilizing traditional analytical methods. Therefore, for future research, neural networks and machine learning methods can be used to investigate more determinants related to social media (e.g., Social Network, Shares, Likes, Media Coverage, and Facebook Friends) and to learn how social network relationships among individuals or groups play a role in the success of crowdfunding projects. For example, artificial neural networks can be used to build project-, fundraiser-, and funder-oriented social network graphs by extracting relevant content from projects as well as unique social features of fundraisers and funders, which can provide insights into the deep links between projects and funders, and further identify the possibility of success in crowdfunding projects (Rivas et al. 2020).
Crowdfunding models of projects also need attention in future work. First, most of the reviewed papers investigate the success of reward-based crowdfunding projects (79 out of 94 papers), and fewer than 20 papers focus on equity-based projects. In comparison, only ten papers analyze loan-and donation-based projects. Importantly, it is challenging to summarize the overall significance and directions of the determinants related to the success of loan-and donation-based projects based on the limited literature. Therefore, researchers are encouraged to devote more attention to equity-, loan-, and donationbased projects to find new insights into the determinants of crowdfunding success in their future endeavors. Second, we find that the determinants considered in rewardand equity-based studies are roughly different. Several determinants considered in both crowdfunding models are paradoxically found to have different effects. In particular, for equity-based projects, we find that the determinants related to soft information, creators, backers, and platforms are rarely examined in current studies. Therefore, they should receive more research efforts in future research.
To conclude, platform models, crowdfunding models, and measurements of crowdfunding success should be considered when analyzing the determinants of crowdfunding success. Opportunities still exist in future research for projects that belong to rarely examined platforms or crowdfunding models. The determinants investigated by a few studies also need more attention, which requires a combination of some more recent methods and techniques across multiple domains.

Conclusion
We conduct a review of extant research on the determinants of crowdfunding success. Our review is based on the approach of an assessment review to assess different studies and identify the aspects that need more attention in future research. Following the guidelines for the literature search and review advocated by Hossain et al. (2019) and Leidner (2018), we select 94 empirical studies from 2011 to 2021 from 37 journals, three conference proceedings, and other resources (i.e., working papers, theses, reports, and books) with a multistage search strategy. We then collate and analyze them based on different measurements of crowdfunding success and different crowdfunding models to separately list and assess the determinants. We assess the empirical impacts of various determinants on the success of crowdfunding projects and summarize several influencing rules to provide multiple potential dimensions of theory and practice for future work on crowdfunding success. Finally, we construct an integrated framework for the determinants of crowdfunding success and highlight several research gaps in the need of more attention. We identify eight main ways to measure crowdfunding success and find that the dichotomous variable Funding Success and the continuous variables Funds Raised, Success Ratio, and Number of Backers are the most common measurements of crowdfunding success. We document the definitions and measurements of crowdfunding success, the different crowdfunding models, and various determinants that can affect the empirical findings. We also suggest that the platforms that follow a "Keep-it-All" or mixed model, the projects that belong to equity-, loan-, and donation-based models, the determinants related to backers and platforms, and the determinants with inconsistent findings or those that are rarely studied further merit exploration. We also call for popular techniques (e.g., text, image, and video mining) and methods (e.g., neural networks and machine learning), as well as multiple domains, such as behavior and psychology, to be more intensively considered in future work.

Discussion
We identify 94 empirical studies that examine the empirical effects of the determinants on crowdfunding success, most of which are from Entrepreneurship Theory and Practice and the Journal of Business Venturing. We find that the extant literature has generally been published in the fields of business and economics and computer science since 2011. Through our collation and analysis of the selected papers, we conclude eight main ways to measure crowdfunding success, which responds to Question (1) proposed in our Research Goals section. The dichotomous variable Funding Success and the continuous variable Funds Raised, Success Ratio, and Number of Backers are the most widely used measurements of crowdfunding success. Furthermore, the measurement of crowdfunding success determines the research method. Papers focusing on Funding Success widely adopt logistic/logit regression or probit regression; papers on Funds Raised and Success Ratio widely use linear regression, while papers on Number of Backers commonly use negative binomial regression. Papers using other measurements of crowdfunding success are the minority, which reveals a gap in the need for attention in future research. It is worth noting that the choice of research method is appropriate for measuring crowdfunding success. In addition, there is also a need to conduct surveys and experimental studies based on crowdfunding platforms to study individuals' or groups' behavior, as the majority of the extant literature merely uses secondary data collected from crowdfunding platforms. In this case, the measurements of crowdfunding success rarely studied in the extant literature (e.g., Decision to Invest) can be considered. We list the studies that use multiple measurements of crowdfunding success and the research methods adopted in Appendix 3, which can be used as a reference.
To address Question (2) proposed in the Research Goals section, we list the measurements and definitions of the determinants in Table 6. The studies that examine a certain determinant and the direction of influence for each determinant are listed in Table 6. We classify the determinants into four categories based on , Kaartemo (2017) and . Although most selected studies have widely studied project-and creator-related factors, backer-and platform-related factors are rarely considered. The most widely examined are Goal, Duration, Reward, Team Size, and Category related to project characteristics. Many of the selected studies investigate Text, Visual, Social Network, Updates, and Comments, which are factors associated with project soft information. In addition, creators' Experience, Facebook Friends, geographic differences, and gender differences are also examined by studies on creator-related factors. According to our review of the determinants, researchers can focus on relatively new or unique variables that have rarely been studied in the extant literature. Accordingly, the backer-and platform-related factors, the factors that yield inconsistent findings, or those rarely studied merit further exploration.
Furthermore, we assess the empirical effects of the determinants on crowdfunding success by considering each measurement of success and each crowdfunding model separately and subdividing some determinants in response to Questions (3) and (4). We document that the measurement of crowdfunding success is one of the important reasons for the differences in the research results, and different definitions for the same types of determinants can also display conflicting effects on crowdfunding success. Moreover, different measurements for crowdfunding success or a determinant can also yield different results even within the same paper. For example, Anglin et al. (2018a) find that Goal and Duration negatively affect Funding Success but have no relationship with Funds Raised, and creators' Experience, which refers to the number of previous projects created, has no connection with Funding Success and Funds Raised. In contrast, entrepreneurial experience is found to have a positive effect. Therefore, it is important to identify each determinant's definition and measurement of crowdfunding success. This will affect the ability to accurately assess the success of crowdfunding projects and influence researchers' degree of attention to various factors. In addition, we also underline that the determinants with inconsistent findings and those with subdivided definitions such as Reward, Text, Visual, Comment, and creators' Experience need more attention in the future. Combining popular techniques that can handle text, image, and video content (e.g., text, image, and video mining), popular research methods in the field of social networks (e.g., neural networks and machine learning), and knowledge from other domains, such as psychological and behavioral sciences, can yield more interesting findings. Moreover, we note that the crowdfunding projects involved in a majority of the selected papers (more than half ) belong to the reward-based model and are from "All-or-Nothing" platforms. There is significant variation among the determinants considered in studies focusing on different crowdfunding models, and several determinants are found to have different effects in different crowdfunding models; therefore, we conjecture that the data collected from "All-or-Nothing" reward-based crowdfunding platforms may be more comprehensive, larger scale, and more suitable for conducting research. Nevertheless, other niche platforms and equity-, loan-, and donation-based projects may also contain valuable information discovered in the related literature.