Abstract
This article investigates multimodal elements—images, links, gifs, videos, and galleries—of crowdfunding campaigns on the platform Kickstarter to develop an understanding of characteristics of successful campaigns. The authors scraped 327,586 campaign pages, analyzing the multimodal elements of successful and unsuccessful campaigns. They found that successful campaigns featured more images, links, and gifs and more frequently included a project video than did unsuccessful campaigns. Images, links, and the presence of a project video had a positive impact on success while gifs and project galleries did not. These findings give business communicators practical guidance, develop theoretical aspects of Kickstarter research, and validate previous findings with a larger data set.
For many new businesses, seeking capital to fund the venture is a vital process, but financing through traditional or venture-capital loans are options that are not available to every new business. Businesses that seek to reach small niches, offer highly speculative ideas, or suggest atypical business methods may be shut out from those channels due to concerns over the businesses’ viability. Entrepreneurs of all types, but particularly artists and artisans in creative industries, may be shut out from such funding due to traditional funders’ reservations over size, scope, and product types. Crowdfunding is an emerging fundraising model that allows small, speculative, or atypical businesses to seek capital from organizations beyond banks and venture-capital firms. This emerging model requires entrepreneurs to create a multimodal pitch to present online: Entrepreneurs can use text, video, images, and more to explain why the online public should give money to their company. Much research has been conducted on the text of crowdfunding proposals—called campaigns—to determine how successful crowdfunding campaigns are written, but less research has focused on how successful campaigns employ videos, images, and other multimodal elements.
This article uses a large sample of crowdfunding campaigns from the flagship crowdfunding platform Kickstarter to investigate what sorts of multimodal element use are associated with successful and unsuccessful campaigns. After reviewing the literature on crowdfunding, we explain our statistical methods for analyzing multimedia use in 294,593 Kickstarter campaigns, present the results of analysis, and discuss these findings. We conclude by suggesting the need to attend to the subtle, meaningful transformations that occur when variants of traditional genres such as proposals are written online.
Literature Review
Crowdfunding is “an open call made through the internet to provide financial resources to support new ventures” (Belleflamme et al., 2016, p. 6). Crowdfunding represents a significant innovation in new venture funding, allowing a wide array of groups that have been unable to seek traditional venture capital to acquire early-stage funds (Bradford, 2012; Mollick, 2014). Entrepreneurs, artists, and independent journalists are among these groups (Davidson & Poor, 2014; Harrison, 2013; Hunter, 2014). With business communication expanding its boundaries from organizational writing to also include nonorganizational writing (Carradini, 2019, 2020), crowdfunding represents an emerging tool for a group of professionals that business communication increasingly considers part of its research remit.
Despite being an alternative to traditional funding methods, crowdfunding has gained legitimacy in professional spaces (Ahlers et al., 2015). The volume of money raised lends legitimacy to the endeavor: As of 2020, creators of crowdfunding campaigns have raised $34 billion across all crowdfunding platforms (Fundly, 2020). The U.S. government’s legal guidance also confers legitimacy to this fund-raising method: The U.S. Securities and Exchange Commission allows companies to “raise a maximum aggregate amount of $5 million through crowdfunding offerings in a 12-month period” (U.S. Securities and Exchange Commission, 2021). With crowdfunding as a socially accepted, financially viable, and legal alternative means of acquiring funding, the ability to write an effective crowdfunding campaign is an increasingly valuable skill for the business communicator (Pope, 2018).
Research on Writing Crowdfunding Campaigns
Due to crowdfunding’s increasing value to businesses, much research has been conducted on how to write crowdfunding campaigns. This research has found that writing a crowdfunding campaign differs from traditional venture-capital funding pitches although some similarities exist in choosing how to structure the timing and economics of the request (Belleflamme et al., 2013; Meer, 2014; Mollick, 2014; Salahaldin et al., 2019). Authors of crowdfunding campaigns must write in a public, digital context for diverse audiences, such as existing social networks, community members, and geographically distant parties (Agrawal et al., 2015; Vealey & Gerding, 2016). Effective, successful crowdfunding writing requires
writing an in-depth proposal and providing updates to the project after initially publishing the campaign (Block et al., 2017; Koch & Siering, 2015; Xu et al., 2014) signaling credibility through disclosing experience level, economic planning, risk management, and endorsements (Ahlers et al., 2015; Courtney et al., 2017) including phrases that convey reciprocity, scarcity, social proof, social identity, likeability, and authority (Mitra & Gilbert, 2014) mentioning the name of the author in the campaign text repeatedly (Gafni et al., 2018) employing positive and inclusive language (Anglin et al., 2018, p. 484) avoiding language showing uncertainty (Kaminski & Hopp, 2020) using less language focused on monetary motivations in favor of more language describing prosocial reasons for the campaign (Kaminski & Hopp, 2020) employing image-focused language (Patel et al., 2021) using the linguistic style and persuasion tactics appropriate for the specific audience (Allison et al., 2017; Parhankangas & Renko, 2017)
Those findings focus on writing alphanumeric text for crowdfunding campaigns, which is indeed a primary component of writing the campaign. But in this article, we focus on Kickstarter, an early crowdfunding platform that is currently the industry standard. Kickstarter asks campaign creators to explain the project, note potential places where it might fail, and offer an unspecified amount and type of rewards to backers (Gerber & Hui, 2013). While Kickstarter offers a text box for this work, that box can include embedded media. As a result, crowdfunding campaigns on Kickstarter and similar platforms often contain more than just alphanumeric text.
Definitions: Multimodality and Multimodal Elements
The media included in the crowdfunding campaigns consist of multimodal elements that deliver meaning outside of alphanumeric text (Kress, 2003; Kress & Selander, 2012). Images, videos, and hyperlinks are elements that extend or bypass alphanumeric text to help communicate information to the reader (Koch & Siering, 2015).
Kickstarter provides the functionality to embed many types of media in the campaign: videos, images, gifs, links, and project galleries. It requires that entrepreneurs contribute a project image (Mollick, 2014)—an image that is placed directly under the title and subtitle of the campaign page (Garber, 2021). Although Kickstarter strongly encourages entrepreneurs to include a project video as part of the campaign, it does not require them to do so (Kickstarter, 2018). If the campaign creator chooses to make a video to promote the campaign, that project video will take the place of the project image (Singh, 2021). While other images and videos can be included in the body of the campaign, the project image and project video are part of the template that Kickstarter has created for a campaign. The campaign must include one—and only one—project image or project video.
Kickstarter has, at certain times, required campaigns about technological hardware to contribute a project gallery, Kickstarter’s name for an image carousel that allows users to scroll through multiple images depicting the campaign product (Mindset, 2017). These project galleries require creators to provide an extra level of detail regarding complex, potentially speculative technology projects. Only one gallery is allowed per campaign.
Gifs, links, more images, and more videos can accompany the required multimodal elements in the Kickstarter campaign. Gifs are short, looped moving images that can be embedded in text. They are created by clipping part of a video into a tiny segment or by creating an animation. They are often casual, humorous clips that rely on shared cultural touchstones, such as television shows or movies. Online writers can use them to express an emotion, send a message, or tell a joke, such as in the gif of Troy and Abed from the television show Community expressing that a person unexpectedly has a good point (Harmon, 2012). Links are hyperlinks to other webpages; they are multimodal elements because they allow the reader to go beyond the text—in this case, to the content of another webpage—to understand the meaning of the proposal.
Finding out what multimodal elements are being employed successfully in Kickstarter campaigns would continue to develop business communicators’ knowledge of what successful campaign writing looks like and provide guidance regarding best practices in Kickstarter campaigns.
Research on Multimodality in Kickstarter
Despite the ability to employ multimodal elements in Kickstarter campaigns and Kickstarter’s encouragement to use multimodality, research on the effects of multimodality on the successfulness of Kickstarter campaigns returns mixed results. Kaminski and Hopp (2020) found in their study of text and visuals in Kickstarter campaigns that “visual, potentially emotionally appealing cues are the most potent signals crowdfunding campaigns can provide” (p. 642). Yang et al. (2020) agreed, suggesting that “an increased number of images and videos improves fundraising performance” (p. 6) but that images and videos can “overshadow” the text and reduce the text’s effectiveness due to cognitive overload.
Some research has looked at the effects of individual multimodal elements, such as videos. Koch and Siering (2015) found that videos were highly significant in a Kickstarter campaign’s success. Yet Grebelsky-Lichtman and Avnimelech (2018) found that not all videos are equally helpful. Immediacy behaviors varied in Kickstarter videos. Immediacy behaviors are ways that “an individual signals closeness, willingness to communicate, and positive feelings for a person or idea.” They found that videos featuring verbal immediacy behaviors (e.g., plural pronouns, informal language, complimenting others, disclosing information) while avoiding nonimmediacy behaviors—both verbal (e.g., formal expressions, individual pronouns) and nonverbal (e.g., “speaking in a monotone, looking away from the person receiving the message, frowning while talking, tense body posture, and avoiding gestures”)—were likely to be associated with successful campaigns. Korzynski et al. (2021) found that for videos, “self-presentation and exemplification techniques are positively associated with crowdfunding success, while intimidation is negatively related to crowdfunding success” (p. 675).
Research on images in Kickstarter has also resulted in mixed findings. Hou et al. (2019) found that only certain types of images were effective in Kickstarter campaigns. That is, images eliciting sadness, contentment, and amusement were positively correlated with the number of backers whereas images eliciting fear were negatively associated with the number of backers.
Ultimately, then, research on multimodal elements’ effects on the efficacy of Kickstarter campaigns reveals primarily (but not exclusively) positive relationships between multimodality and success in limited categories and amounts of data.
Generalizability of Findings
Although many crowdfunding platforms exist and allow multimodal elements in their campaigns (Lagazio & Querci, 2018), Dushnitsky and Fitza (2018) warned that “factors associated with success in a given platform do not replicate to the other platforms” (p. 1). But findings concerning individual platforms may be generalizable within that platform. For this reason, this study considers a large amount of data from a specific platform (Kickstarter) in order to help business communicators communicate more effectively in the most well-known of the crowdfunding platforms.
While cross-platform generalizability of findings regarding multimodality is difficult (Dushnitsky & Fitza, 2018), several factors confound generalizability in previous research, even within the single platform of Kickstarter. A focus on exclusive categories of campaigns hampers a generalizable analysis; for instance, Kaminski and Hopp (2020) investigated 20,188 campaigns in the technology and product design categories of Kickstarter, leaving out all other categories. The size of the analysis can also confound generalizability: Koch and Siering (2015) studied campaigns of all categories on Kickstarter but only studied 1,000 campaigns. Sometimes these factors appear together due to a novel or specific methodological concern. Grebelsky-Lichtman and Avnimelech’s (2018) in-depth, groundbreaking analysis of verbal and nonverbal patterns in 120 Kickstarter campaigns focused on campaigns surrounding 3-D printers, mobile applications, iPhone stands, and organic food. This valuable research would be difficult to scale. And Hou et al. (2019) studied 840 campaigns in the public benefit category of Kickstarter in order to test a novel methodological concern. Our study uses a large data set from all categories of Kickstarter to work toward generalizability of the findings for the platform.
Ultimately, this study investigates the effects of multimodality in Kickstarter campaigns with an effort toward generalizability by analyzing multimodal elements in 294,593 Kickstarter campaigns via two research questions.
Research Questions
To answer our first research question (RQ1)—Do successful campaigns feature more multimodality (in their number and complexity of multimodal elements) than do unsuccessful ones?—we considered descriptive statistics such as the average number of images, links, and gifs in successful and unsuccessful campaigns. We also assessed whether successful campaigns included a project video or a project gallery more often than did unsuccessful campaigns. Finally, we investigated whether choosing a less complex multimodal piece (a project image) instead of a more complex one (a project video) was associated with unsuccessful campaigns.
To answer our second research question (RQ2)—Does an increase in multimodality (in the number and complexity of multimodal elements) increase the odds of succeeding in a campaign?—we drew on inferential statistics (logistic regression analysis) to determine whether increasing the number of images, links, and gifs in a campaign would increase the odds of succeeding in a campaign and whether including a project video (instead of a project image) or project gallery would increase the odds of succeeding in a campaign.
Method
To analyze the multimodal elements of Kickstarter crowdfunding campaigns, we scraped data from Kickstarter. Using a scraper developed for this purpose from a starting point of the software tool Quickscrape, we gathered data on the multimodal elements in each campaign (Shuttleworth Foundation, 2014). The terms of service at the time of scraping did not outlaw scraping data from the service (Kickstarter PBC, 2019). Using a Ruby-based script, a hired coder searched through the pages of Kickstarter’s discovery feature for campaign URLs. Kickstarter organizes campaigns by location and general category of the project type, making possible searches such as “all music projects in Montana” or “all technology projects in Europe.” We sought as comprehensive a data set as possible, seeking campaigns of every location and type. Using this categorical system of project type and location, the coder used a scraper to gather all URLs of campaigns listed on the page for each combination of location and project type. The scraper searched for pages of each type–location combination and scraped all campaign URLs until a 404 (Not Found) status code appeared. This scraping process proceeded smoothly from one page of campaign listings to the next. Some URLs could not be scraped because Kickstarter’s website only allows each category to display 200 pages of campaign listings per search. Thus, while rare, type–location pairs containing more than 200 pages of campaign listings could not be scraped in their entirety.
After collecting campaign URLs from each of the listing pages, the coder used a scraping tool to harvest references to the content elements of the individual campaign pages. The tool scraped all the text, html tags, and embedded URLs of the content, allowing us to see the types of content associated with each campaign. Html tags (e.g., header tags, video buckets) and file types (e.g., .gif) provided content information. The scraper could not access some data due to changed page structures, variable types of included data, and inconsistent layouts over the site history. In particular, the scraper could not access the target amount of money for a campaign from successful pages due to Kickstarter’s reformatting of successful campaign pages to remove the target amount of money from the page.
The scraper ran for 24 days (May 31, 2018, to June 23, 2018). After eliminating broken or empty records, this process resulted in usable data from 327,586 scraped URLs. We removed from the data 1,936 campaigns that were ongoing at the time of the search because these active campaigns could not be slotted into successful or unsuccessful categories for analysis. Their removal resulted in 325,650 usable records with information about multimodal elements for analysis. This number included data on 190,233 unsuccessful campaigns and 135,417 successful campaigns. This scraper is available in a public GitHub repository (Carradini, 2022).
To prepare the data for analysis, we deleted duplicate projects and projects displaying scraping errors. For example, the data indicated that some projects featured neither a project video nor a project image, but Kickstarter does not allow a campaign to run without one of these. We eliminated faulty data from the analysis to ensure reliable results. For the rare randomly missing values in certain items, we imputed the missing values via predictive mean matching (Little & Rubin, 2002). This statistical method relies on a combination of techniques (linear regression, random selection from posterior predictive distribution) to develop likely values for missing numbers in a data set based on the preexisting data set (Schork, n.d.). After this data-cleaning process, we had 294,593 campaigns to use for our analysis.
We then processed this data into multiple Excel spreadsheets for analysis. Among other information, the scraped data provided information on the number of static images, the presence or absence of a video, the number of gifs, and the number of hyperlinks. The data also told us if the campaign reached its monetary goal and allowed us to report descriptive statistics and proceed to inferential statistics.
Analysis
To begin studying the data with inferential statistics, we dummy-coded the Kickstarter campaigns to represent their successful (1) or unsuccessful (0) status, which was our dependent variable. Campaigns coded with success met or exceeded the goal amount at the end of the assigned chronological period. We coded three types of campaigns as unsuccessful: projects that failed to receive enough donations to meet the goal amount in the allotted time (failed), projects terminated by Kickstarter for violation of rules (suspended), and projects terminated by the creator before the assigned chronological period ended (canceled). We included suspended and canceled campaigns under the category of unsuccessful because aspects of the campaign that resulted in its suspension or cancelation were communicated in alphanumerical or multimodal communication of the campaign. In the overall sample, there were only 1,109 suspended campaigns—these were campaigns that might have been succeeding, but Kickstarter removed them for content violations—and 22,512 canceled campaigns, which is a small percentage of the overall sample. Professional sources often suggest canceling a project that is in severe potential of failing before it reaches its conclusion (Stonemaier, 2013). Thus, we considered canceled campaigns as part of the unsuccessful category.
We also assigned numerical codes to our five independent variables: number of images, number of links, number of gifs, presence or absence of a project gallery (present = 1, absent = 0), and project video or project image (project video = 1, project image = 0). Kickstarter allows any number of images, links, and gifs to be included in the text of a project. We treated these as continuous data. Kickstarter allows only one gallery per project. Thus, we treated this variable as a binary. Project videos and project images are mutually exclusive, so we treated them as one variable.
We performed logistic regression analysis to assess the effects of multimodal campaign elements on campaign success. All the conditions for running logistic regression were checked and met. The multimodal campaign elements—number of images, number of links, project image or project video, project gallery, and number of gifs—were included as predictors for successful (1) or unsuccessful (0) campaigns. Due to the large sample size, we used the open-source data analytics software KNIME. Data science with KNIME follows a workflow structure of preparing data for analysis, developing prediction models, and testing the prediction strength of the final model. To be able to conduct logistic regression analysis, all data had to be standardized. We conducted further analysis exclusively with z-scores.
To build the prediction model, KNIME offers the node logistic regression learner. We defined the dependent variable status as the target prediction column. We chose the stochastic average gradient (SAG) method to build the algorithm because SAG works well with large samples and converges relatively fast (Schmidt et al., 2017). After defining the parameters of the model, we used a random approach to split the data into training data and testing data. KNIME uses 80% of the data to train the algorithm and the remaining 20% for testing the prediction model.
With the combination of the learning model and the training data, we could predict the logistic regression model. This step includes the actual prediction of status (the dependent variable) as a function of images, links, project video/image, gallery, and gifs (the independent variables). Finally, the node Scorer (JavaScript) compares the actual status and the predicted status of the testing data. This step assesses the prediction performance of the model.
When running the prediction model with the testing data (n = 58,919) and comparing the predicted status with the actual status, we find an overall prediction accuracy of 68.10%; that is, 40,123 projects were correctly classified by the model. The confusion matrix in Table 1 shows that the status was predicted correctly in 84.73% of all the unsuccessful cases tested whereas the status was correctly predicted in 41.39% of all the successful cases tested. While prediction quality differs between the two statuses, Cohen’s κ = 0.279 indicates a good prediction accuracy of the model, given the slightly unequal distribution of statuses. That is, our data set reflects the reality that Kickstarter contains more unsuccessful campaigns than successful campaigns, leading to an unequal distribution of statuses.
Confusion Matrix for Prediction Accuracy of Campaign Status (n = 58,919).
Limitations and Future Questions
This study leaves some questions unanswered for future research. While this study quantitatively analyzes what multimodal elements already exist in a large-scale data set with no distinctions in campaign category (visual art, theater, music, technology, design, etc.), comparative analysis of data from multiple project categories could reveal if variances in multimodal usage exist within categories. This large-scale study analyzed the value of the existence of multimodal elements but did not analyze their content; research methods such as those employed by Hou et al. (2019) could be expanded to larger data sets to focus on detailed analysis of the content of multimodal elements. Collection methods that result in a comprehensive set of campaign records from Kickstarter (consider Mollick, 2014) could further expand the data size of this project to check the findings’ veracity and generalizability. A qualitative analysis that assesses user response to multimodal elements would help validate quantitative findings. Research that focuses on how creators choose what multimodal elements to employ could address why certain multimodal elements appear and the reasoning for multimodal choices. Research that focuses on other platforms or compares multiple platforms at once could help generalize these findings even further.
Results
We drew on descriptive statistics and inferential statistics, respectively, to answer our two research questions. We present these results here.
Descriptive Statistics
Before analyzing the descriptive statistics of successful and unsuccessful campaigns, we analyzed the overall descriptive statistics for the whole data set. Of the campaigns in our sample, 38.3% were successful. Campaigns include an average of 6.241 images, 2.264 links, and 0.245 gifs. The majority (75.8%) of campaigns have a project video rather than a project image. Only 1.8% of campaigns include a project gallery. Table 2 lists the descriptive statistics for all the independent variables.
Statistics Regarding Successful and Unsuccessful Campaigns.
While the continuous variables (images, links, and gifs) show that some campaigns include a large number of images or links, a vast number of campaigns feature zero or few images or links. The box plot for images shows that 50% of all campaigns have between 0 and 8 images, with a median of 1 (consider Figure 1). Projects with more than 20 images can be classified as extreme outliers. Similarly, 50% of all campaigns include between 0 and 2 links, with a median of 0. Projects with more than five links are considered outliers.
Box plots for number of images and number of links in the campaigns.
To answer our RQ1—Do successful campaigns feature more multimodality (in their numbers and complexity of multimodal elements) than do unsuccessful ones?—we performed several different analyzes of the descriptive statistics.
We found that the answer to this question was mostly yes. Independent t-tests showed that the successful and unsuccessful campaigns differed significantly in their use of multimedia elements (consider Table 2). The successful campaigns averaged more images, links, and gifs per campaign than did the unsuccessful ones. Successful campaigns averaged 9.78 images, 3.74 links, and .41 gifs whereas unsuccessful campaigns averaged 4.04 images, 1.35 links, and 0.14 gifs. But both successful and unsuccessful campaigns averaged less than 1 gif, suggesting infrequent use across both successful and unsuccessful campaigns.
Project videos are usually more complex multimodal elements than are project images. Almost all successful campaigns employed a more multimodal project video instead of a less multimodal project image, with 99.3% of successful campaigns featuring a project video. The majority of unsuccessful campaigns also employed a project video instead of a project image: Unsuccessful campaigns featured a project video 61.3% of the time and a project image 38.7% of the time.
Project galleries, however, appeared slightly more often in unsuccessful campaigns (2%) than in successful ones (1.4%); this result goes against our primary finding that more multimodal elements (in number and type) appear in successful campaigns than in unsuccessful campaigns.
Inferential Statistics
To answer our RQ2—Does an increase in multimodality (in the number and complexity of multimodal elements) increase the odds of succeeding in a campaign?—we performed several analyses of inferential statistics. Again, we found that the answer to this question was mostly yes. All the tested multimodal elements had a significant effect on campaign success (consider Table 3). Both number of images (β = 0.319, p = .000) and number of links (β = 0.248, p = .000) had a positive effect on campaign success, and campaigns with a project video rather than a project image had a higher likelihood of success (β = 1.807, p = .000). But two multimodal elements did not increase the odds of success, tempering an unqualified yes to the research question. The existence of a project gallery (β = −0.106, p = .000) affected campaign success slightly negatively, and the number of gifs (β = −0.013, p = .003) did not affect campaign success.
Logistic Regression Analysis Results for the Effect of Multimodal Elements on Campaign Success.
Our correlation matrix (consider Table 4) shows that multimodal elements of the Kickstarter campaigns are significantly and positively correlated. For example, a campaign with many images is likely to also feature many links. Campaigns with project videos rather than project images tend to feature more of all other elements. Gifs and a gallery are only mildly correlated with other elements. Links and gallery do not correlate significantly. Images, links, and project video show the highest correlations. We suggest in the following discussion potential reasons why these variables correlate and what that means for Kickstarter campaign writing.
Correlation Matrix of the Multimodal Elements and Campaign Success.
Note. **p < .01.
Discussion
Both research questions resulted in similar findings: Descriptive and inferential statistics showed that successful campaigns were associated with high numbers of images and links and high rates of project video inclusion whereas unsuccessful campaigns were associated with low numbers of images and links and lower rates of project video inclusion.
Successful campaigns, on average, featured more than double the amount of images and links than did unsuccessful campaigns. The logistic regression analysis conducted with number of images and links suggested that a higher number of images and links corresponds to a higher probability of success than does a lower number of images and links, adding evidence to the claim that certain types of multimedia correlate with successful campaigns (Kaminski & Hopp, 2020; Koch & Siering, 2015). Yet images showed a stronger positive impact than did links. Thus, images would be expected to be more valuable than links in terms of elements to consider when writing a campaign. This finding supports and extends the findings of Kaminski and Hopp and Yang et al. (2020) about the high value of visuals in successful Kickstarter campaigns. Links are less valuable but still meaningful when writing a campaign. This finding expands the positive aspects of multimodality beyond visuals to the web of connections that hyperlinks make possible.
Successful campaigns more often included project videos than did unsuccessful campaigns (99.3% to 61.3%), and the existence of a project video increased the probability of a successful campaign. Conversely, only .7% of successful campaigns did not create a video whereas 38.7% of unsuccessful campaigns did not create a video. Unsuccessful campaigns, then, included a project image instead of a project video more often than did successful campaigns. These findings confirm the findings of Koch and Siering (2015): Creators should strongly consider making a project video when developing a Kickstarter campaign. This finding contrasts with the finding of Lagazio and Querci (2018) that videos on IndieGoGo are not correlated with success. Thus, we confirm Dushnitsky and Fitza’s (2018) analysis that findings for one platform do not hold across multiple platforms.
Our finding that the project video was more successful an option than the project image also suggests that more complex multimodal elements instead of less complex ones should be included in campaigns. Thus, Kickstarter campaign success correlates with the use of a large amount of multimodal aspects and the inclusion of complex multimodal components. The digital nature of the platform allows and even expects multimodal activity, and those who engage in multimodal writing have increased odds of success.
Two types of multimodal activities are not necessarily related to success. We found that neither gifs nor galleries seemed to influence a campaign’s success. The difference in the average number of gifs between successful and unsuccessful campaigns was only .27 per campaign. Neither successful nor unsuccessful campaigns reached an average amount of 1 gif per campaign, suggesting that creators rarely include gifs in campaigns and that gifs do not necessarily help campaigns succeed. This finding complicates the story that all types of complex multimodal elements are valuable; certain types of multimodality, such as gifs, were not as valuable in this data set as were images, links, and videos.
The infrequent project gallery related negatively to success, appearing more often in unsuccessful campaigns (M = .020) than successful campaigns (M = .014). This finding might suggest that the gallery is a particularly unconvincing multimodal element, or it might simply reflect certain aspects of Kickstarter’s specifications. At the time of scraping, only campaigns listed in the technology, design, or gaming hardware categories could include a project gallery to show the prototypes of the project (Gallagher, 2016). 1 Technology projects tend to fail at a higher rate than do other categories of Kickstarter campaigns. For example, on May 13, 2021, Kickstarter reported 9,765 successful technology projects versus 36,317 unsuccessful ones whereas it reported 32,277 successful music projects versus 32,002 unsuccessful ones; music is the category with the most successful campaigns (Kickstarter PBC, 2021). Design and gaming categories similarly tend to have more unsuccessful than successful campaigns. Thus, completing a successful campaign in these areas may be more difficult, no matter what multimodal elements are included.
Thus, these findings suggest that not all types of multimodality are immediately successful strategies for success in crowdfunding. While project videos, images, and links correlate positively with success, not every form of multimodality is a net positive in Kickstarter. Grebelsky-Lichtman and Avnimelech (2018), Hou et al. (2019), and Korzynski et al. (2021) found that videos and images are not necessarily all associated with success. Further, distinctive strategies for certain types of low-performing multimodal elements (e.g., gifs or galleries) could influence whether they are effective for individual campaigns despite being statistically ineffective. Table 5 gives further guidance for business communicators regarding each of our findings. We did not study the content of the images to determine the creators’ goals or strategies for their multimodal elements, which could be a further area of study.
Guidance for Business Communicators Regarding Each Study Finding.
We also found that variables were correlated with each other. Correlation findings generally suggest that the effectiveness of the variables is not directly traceable to a single variable. Although this type of finding confounds some types of research, it is valuable for practical research on multimodality: Instead of suggesting that one variable is “the answer,” we show that multimodality is not fully reducible to individual multimodal elements. The multimodal elements of Kickstarter campaigns work together to produce meaning, with elements reinforcing each other in doing so. Thus, correlation findings suggest that any individual multimodal aspect of a campaign may not be as effective on its own as it is in concert with others; as business communication scholars, we should expect this outcome to be true instead of expect individual aspects of communication to be divorced from the whole in their unique meaning making.
Conclusions
These findings have practical, methodological, and theoretical outcomes.
Practical Outcomes
These findings suggest some ways forward for entrepreneurs when creating Kickstarter campaigns:
Include a large number of strong images and a project video instead of a project image. Provide links to relevant texts although links are lesser indicators of potential success than are images and videos. Do not expect gifs or project galleries (now an extinct feature) to help the campaign.
Entrepreneurs who create a project video, include multiple images in the body of the text, and provide links to outside sources will likely spend more time in preparing their campaign than would those who use a project image and no other multimodal elements. But this expenditure of time is in turn likely to produce better results than would a campaign that includes only a project image in its header and text in the body of the campaign.
Furthermore, entrepreneurs should begin planning their campaign with all of the multimodal elements in mind. The finding of statistical correlations between multimodal elements further suggests that all multimodal elements work together to help the campaigns succeed. The success of a campaign, then, does not depend on just including a project video. Instead, writers would profit from conceiving of their Kickstarter campaign as an overall, integrated pitch that includes a variety of multimodal aspects to convey the meaning.
Methodological Outcomes
These findings have methodological outcomes. Confirming previous findings with newer and larger data sets develops and validates knowledge in the field (Cordova et al., 2015; Graham, 2017). Our large data set mostly confirmed previous findings about videos and images and thus strongly validates previous research (Kaminski & Hopp, 2020; Koch & Siering, 2015). These confirmations give weight and merit to the practical suggestions that arise from such research.
Theoretical Implications
These findings also have theoretical implications for campaign communication. While we found that certain multimodal elements (and multimodal elements collectively) correlate with success in Kickstarter crowdfunding, the nature of multimodal elements’ effects is still to be fully understood. Both multimodal elements and alphanumeric text convey information, but in different ways. The inclusion of multiple multimodal elements may reflect a well-prepared and detailed idea, and well-developed ideas may be more likely to convince a funder than would less-developed, more speculative ideas (Koch & Siering, 2015). Well-developed ideas may even require multimodal elements for clear explanation.
Multimodal elements, however, may do more than just convey information. Links to other websites can reflect credibility, as links may reference related people or ventures to bolster the credibility of the creator by association (Ahlers et al., 2015; Courtney et al., 2017). Images and videos that show the product or explain the concept may reflect expertize and authority, as well as diminish uncertainty (Kaminski & Hopp, 2020; Mitra & Gilbert, 2014). In these ways, multimodality reflects not an independent concern of crowdfunding but an extension of previous findings on how to write successful Kickstarter campaigns. Multimodality is a part of writing, and thus our findings about how to use multimodal communication elements effectively in crowdfunding campaigns align with research on how to use alphanumeric text effectively in such campaigns.
But not all multimodality is equal. While project galleries’ correlation with failure may be reflective of unusually difficult conditions in which to succeed, gifs were not reflective of such complicated conditions, and yet they did not help the creators succeed. A reason that gifs did not contribute to the campaigns’ success could be that their content clashed with the goals of the platform: Gifs are often casual, humorous clips that rely on shared cultural touchstones for effect ([Reversed Gif], 2013). Including these types of gifs may not be an effective strategy for developing successful campaigns. Gifs, however, can be used for purposes beyond humor, such as those conveying moving models or animated diagrams. Without further content analysis of the gifs—such as Korzynski et al. (2021) and Grebelsky-Lichtman and Avnimelech (2018) did with videos—we do not know what percentage of gifs are intended to be humorous (Emerson, 2016) or entertaining, to convey information (Fowler, 2020), and to do both at once (Singh, 2021). More research is needed into why gifs in general are not positively related to success.
Multimodality and Online Genres: Kickstarter and Beyond
While these findings have theoretical implications for Kickstarter research, they have implications for scholars of business communication, multimodality, and online genres in general.
As more and more business communication becomes digital, care should be taken to investigate how digital-only elements of communication change and shape new versions of old genres. Kickstarter campaigns are a unique form of business communication that has similarities to traditional business pitches. Yet the online space where the campaigns are posted allows the inclusion of multimodal elements that change the process of giving and receiving the pitch. Multimodal communication as a whole, and some individual multimodal communication elements, correlate with success in Kickstarter campaign funding. These distinctive elements make a difference in this emerging type of online pitching.
Not only is the traditional pitch genre changed as it enters the digital space, but elements that the digital space allow and encourage work together with traditional elements such as alphanumeric text and images to transform how the pitch creates meaning. We argue that adding videos, gifs, links, and galleries does not simply duplicate the meaning made by the alphanumeric text, but instead these elements work with the text and the other multimodal elements to change the way that the audience receives and responds to the information. These sorts of transformations are obvious and yet subtle: A big video at the top of a pitch announces itself loudly (especially if autoplay is on). But these digital multimodal elements are not mere attachments to a piece of writing. They contribute to the meaning that the reader receives in complex ways that should continue to be investigated by business communication scholars in this genre, other genres being transferred from print to online, and born-digital genres.
Footnotes
Acknowledgments
A 2018 C.R. Anderson Grant from the Association for Business Communication supported the research in this manuscript. The authors would like to thank Barbara Carradini for her assistance in scraping the data.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the C.R. Anderson Research Fund (grant number 2018 Grant).
