1 Introduction

Crowdfunding emerged over the last decade and established itself as a substantial financing channel for entrepreneurs, artists and social projects. Patreon is the biggest crowdfunding platform that turns the one-off interaction between creator and crowd into a continuous relationship. It facilitates the presentation and communication of past/present/planned content and allows patrons to support creators of all kind (e.g., musicians, visual artists, podcasters, video creators, writers and journalists) with a monthly pledge. The patron can discontinue support at any time.

Our study uses data from Patreon in order to analyze crowdfunding for a monthly income, focusing on projects’ success determinants and funding dynamics. The previous research on conventional reward crowdfunding platforms, like Kickstarter or Indiegogo, shows that projects’ communication with funders is a key driver of success. For instance, Mollick (2014), Colombo et al. (2015), Crosetto and Regner (2018) find that Web site measures like videos, images and blog entries are determinants of project success.Footnote 1 This evidence suggests that effort put into communicating the project’s goal pays off as it is rewarded by the crowd. Studies analyzing the dynamics in reward crowdfunding show herding behavior among the crowd, very early in the funding period (van de Rijt et al. 2014; Colombo et al. 2015) but also later (Crosetto and Regner 2018).

Our data show that the income distribution at Patreon is very skewed. Hundreds of creators crowdfund a sizable income via Patreon—the top 1% of campaigns make at least $2500 monthly—while the majority of all campaigns attract only negligible amounts. Our analysis results in similar project-level success determinants as the existing research: Indicators of communication quality, like an image upload, the length of texts that describe the campaign, its creator or their goals, or a thank you video, are correlated with funding success. Funding dynamics are heterogeneous across campaigns. Some attract pledge boosts—in the beginning as well as later in their life—and others increase their pledge level at a slow and steady rate. Pledge deletions commonly occur only occasionally, but some campaigns experience substantial drops of crowd support.

Our study extends the analysis of reward crowdfunding (see Butticè et al. 2018, for a recent review) to a new, membership-based platform type, crowdfunding for a monthly income. We show that previous insights about the importance of communication quality for reward crowdfunding success and about funding dynamics also apply in this new context. These results are a stepping stone for a better understanding of a way of funding that can provide a monthly stream of income to creators. This is a relevant issue, because for creators without an established reputation generating revenues from their work is a serious challenge. Crowdfunding may offer an alternative to the standard approach of the Internet era. In the advertising-based business model, the creative class makes their content freely available on platforms like YouTube in return for a share of the revenue from advertisements shown in connection to their content. Online advertising comes with a series of question marks (ad blocking software, privacy regulations, biased content), though, see Peukert (2019) for an overview. Instead, crowdfunding gives creators an opportunity to receive transparent financial support directly from the people enjoying their content—if they convince the crowd of their potential and if they deliver on their promises.

2 Data

The online platform Patreon brings together content creators and people appreciating such content. It facilitates the financial support of creators on a monthly basis. Patreon refers to itself as a membership platform and is considered to be the world’s biggest of that type. As of November 2019, creators on Patreon have received over $1 billion from more than $4 million individuals (according to Patreon’s blog).

Our dataset consists of all transactions at Patreon from its launch in May 2013 until October 2015. This database dump is publicly available.Footnote 2 The data contain information at the following levels: users, campaigns started by users (then referred to as creators), goals set up within a campaign, and pledges made by users (then referred to as patrons) to a campaign. We operationalize our variables following the approach of related studies on reward crowdfunding success determinants.

For each creator, we know whether an image has been uploaded and a social media account has been linked (Facebook, YouTube or Twitter). In addition, we computed the word counts of what creators wrote about themselves. At the campaign level, we know whether a video, thanks video or thanks message are associated with the campaign and have accordingly created respective dummies. We have also computed word counts of the campaign summary and the thanks message. For each goal, we know the amount the creator set and we computed the word count of the goal’s title and text.

Table 1 provides summary statistics. Approximately, three out of four campaign creators upload an image and supply a link to their social media presence. About half of the creators write something about themselves. If they do, the average text of a creator consists of 49 words. Three out of four creators are male (according to the information that 23,840 creators provided about themselves). About 42% of all campaigns provide a video, and the average summary of a campaign consists of 157 words. After a pledge is made, about 40% of campaigns deliver a thanks message (with, on average, 27 words) and 24% show a thanks video. The majority of campaigns started getting pledges in 2015.

Table 1 Summary statistics
Full size table

Creators associated 188,128 goals with their campaigns. On average, they set 2.38 goals (minimum 1, median 1, maximum 60). Excluding outliers,Footnote 3 the average goal is $934.74 (minimum 1, median 200, maximum 78,000). The average goal has a title of four and comes with a text description of 29 words. Overall, there are 959,909 pledges. For each pledge, we know its size, which patron made it and which campaign received it. We also know the timestamp of the pledge’s start and—if applicable—the timestamp of its deletion. This allows us to aggregate pledges over time, by the patron who made them as well as by the campaign that received them. The average pledge is for $11.18 (minimum 0.01, median 5, maximum 25,000) per month.

3 Results

3.1 Descriptive analysis

Most of the registered campaigns (115,183 out of 139,366) did not receive any pledges. A total of 24,257 campaigns getting pledges received, on average, 39.57 (minimum 1, median 5, maximum 27,547). Around 20% of all those "active” campaigns (4930) received one pledge, around a third got two pledges. A campaign’s average monthly peak pledge volume is $197.88, with the median being $31. Only 25% of the active campaigns get more than $120, 5% make more than $750 monthly and 1% more than $2500. Thus, most campaigns either do not get any pledges at all or they receive only negligible amounts. Nevertheless, the top 5%, about 1200 creators, receive a substantial income and around 250 creators (the top 1%) are funded to an extent that can be considered an income sufficient for making a living.

The majority of all campaigns (79,724) set at least one goal. As 20,110 of them received pledges, it seems that goal setting is conducive to funding success. However, it is not a must: 4147 creators who did not associate a goal with their campaign received pledges. On average, creators set 2.31 goals (minimum 1, median 1, maximum 60). Looking at the top 5% of creators with respect to funding received, for a quarter of them the monthly pledge level exceeds the maximum goal they have set. While goals surely provide some orientation about the financial needs/targets of creators, the crowd seems to continue providing support even beyond the goal.

Overall, 473,394 patrons pledged. The median patron pledged once, the average patron pledged twice, and the highest amount of pledges by one patron is 1074. The average patron spent $23 on their monthly pledges.

3.2 Success determinants

In contrast to conventional reward crowdfunding which features a funding target, it is not straightforward what actually constitutes the success of a Patreon campaign. Our analysis therefore employs a set of success proxies. The most basic one is whether a campaign attracted any pledges. Moreover, we consider the maximum monthly amount pledged to a campaign over its duration. Our empirical approach connects these two measures in a Heckman sample selection model with robust standard errors. Table 2 presents the regression results. The explanatory variables in the selection equation (column 1) are dummies for an image, the about text, a social media link, the summary text, a video, the thanks text/video, whether at least one goal has been set, and dummies for the creator being male/female. In addition, the amount equation (column 2) employs the number of goals set, the average goal amount and text and a control for the year the campaign started to get pledges. Its dependent variable is the logarithm of the peak pledged amount.

Table 2 Determinants of campaign success
Full size table

The results show that all our proxies of communication quality are highly correlated with whether a campaign managed to attract any pledges at all. Similarly, the peak pledge level of a campaign is correlated with whether an image is associated, the length of the about/summary/thanks/goal text, and whether a thank you video exists. While female creators seem to be more successful in attracting any pledges at all, there is apparently no gender difference in the pledged amount.Footnote 4 The pledged amount is negatively correlated (at the 5%-level) with the number of goals. The correlation of the error terms (\(\rho =0.63\)) is significant at the 1% level, justifying the sample selection approach.

Finally, our analysis considers a subjective success measure, taking the perspective of the creator: whether a goal has been reached. Table 2, last column, presents the results of a panel regression using all goals, controlling for creator heterogeneity. The specification is a linear probability model with robust standard errors. In addition to the explanatory variables familiar from the previous regressions, we use the goal’s amount and the word count of the goal’s text as regressors.

Whether a goal has been reached is highly positively correlated with an image upload, a social media link, the length of the summary/thanks/goal text and the thank you video. The goal’s amount is a negative determinant. The dummy for male creators is negative and significant at the 5% level.

3.3 Dynamics of pledges and deletions

Pledges at Patreon are recurring on a monthly basis, and patrons can cancel them at any time. This feature distinguishes Patreon from other reward crowdfunding platforms on which campaigns run for a preset period (usually not more than a month or two). These differences—no campaign deadline and the possibility to cancel the monthly pledge—will be the focus of our analysis of dynamics.

Overall, about 29% of all pledges have been stopped. One out of four pledge deletions takes place within the same day and one out of two during the first month. In order to get a first glimpse of the dynamics during campaigns, we relate the campaign’s monthly amount of pledges to the maximum monthly pledge level. For this purpose, we use the end of our dataset, considering the timing of the cutoff as random, in order to analyze how far, or close, campaigns tend to be off their peak pledge level. At the time our dataset ends, the median of the ratio to peak is 1 and the average 0.9. Three out of four campaigns are within 5% of their peak level, 90% are within a third of the peak. On average, this patterns holds across the distribution of a campaign’s peak pledge level. Campaigns in the bottom half of the income distribution suffer from a similar average rate of cancellations as campaigns in the top half or the top 5% earning campaigns.

We proceed to analyze pledges and their deletions over time. Figure 1 shows scatter plots of the cumulative monthly amount pledged to a campaign. In order to optimize visualization, we focus on a subset of campaigns. We limit the peak pledge level to the top 5% (more than $750) and the length of campaigns to the interval of 120 to 150 days. This results in 61 campaigns. Pledges are shown in dark gray, while deletions appear in light gray.

Fig. 1
figure 1

Cumulative monthly amount pledged over time (pledges in dark and deletions in light gray)

Full size image

The plot indicates sharp increases in the cumulative pledged amount, in the beginning of campaigns but as well at later points of time. This is in line with existing evidence from conventional reward crowdfunding sites (van de Rijt et al. 2014; Colombo et al. 2015; Crosetto and Regner 2018). Besides the pledge boosts, several campaigns also increase their pledge level at a fairly steady pace, moving along a somewhat linear trajectory. The dynamics of deletions are also quite heterogeneous. One campaign passed the level of $5000 in monthly pledges within its first ten days, peaked after one month of its launch, and then gradually lost patrons during the following year (for a total decrease of about 20% off its peak). Another campaign suffered a sharp drop of its pledges at around day 130 and essentially discontinued. However, most of the campaigns receive only occasional pledge cancellations. They tend to get more new pledges than deletions.

These behavioral patterns suggest that the crowd reacts to the creators’ work. Sharp or gradual decreases in crowd support are likely to be rooted in a (partly) failure of the creator to deliver the content promised at the outset of the campaign. Thus, the possibility to cancel the monthly pledge can be seen as a feedback system. Patrons have a way to stop support at any time, if they feel the creator failed to deliver or provided only disappointing content. An important caveat is that our analysis does not allow us to causally connect deletions to a lack of content delivery. Another potential motivation for pledge deletions can be on the patrons’ side. They may not be able to afford their pledges any more, or have set a certain budget to support creators at the platform and decide to shift their pledge to someone they like more. Such patron-driven motivations to cancel pledges would tend to be evenly distributed across campaigns, though. Nevertheless, a combined analysis of the dynamics of pledging and communication activity (e.g., uploads of new content) would be a promising path for future research to identify causes behind consistent campaign success.

4 Conclusions

Our study provides a starting step analysis of crowdfunding for a monthly income, based on data from Patreon, the biggest platform of that kind. We find that the top 1% of creators (about 250) crowdfund a monthly income of at least $2500. The top 5%, or about 1200 creators, receive more than $750 monthly. Thus, a substantial number of creators receive a steady revenue stream via Patreon, presumably big enough to focus their time and energy, at least partly, on their creative work. As Patreon reports creator revenues of $150 million in 2017, 1.5 times more than in its first 3.5 years of business (Church 2017), these numbers have likely increased significantly by now.

Overall, our data suggest that identifying quality content functions well at Patreon. The entry barriers to start a campaign at the platform are negligible (i.e., merely setting up an account). The lack of basic features of a serious campaign (creator image, descriptions) indicates that many would-be creators just try whether they can somehow attract pledges, or rather donations. However, our analysis shows that the crowd tends to support campaigns that communicate why they deserve to get funded. Measures of communication quality, like an image upload, the text length of content provided about the campaign, its creator or goals or a thank you video, are correlated with funding success. Finally, the funding dynamics indicate a substantial amount of deletions at some campaigns. The option to delete the monthly pledge to a creator at any time can be regarded as a feedback mechanism. Creators’ promises, as laid out in the campaign’s description, seem to have to be kept. If not, patrons stop support by canceling their pledge.

Our analysis of success determinants at the membership platform Patreon is consistent with the results from conventional reward crowdfunding portals (e.g., Mollick 2014; Colombo et al. 2015; Crosetto and Regner 2018). A more in-depth analysis of the campaigns’ content promises to be an intriguing path for future research. This includes the textual features of creators’ content connected to their campaigns as well as the effect of creators’ activities (e.g., the upload of a new video) on the funding dynamics.

How to generate revenues from their work is a serious challenge for creators, especially in the era of the Internet. For emerging creators, those without an established reputation, the standard approach is to make content available for free on advertising-based platforms. However, as argued by Peukert (2019), the repercussions of the ad-based business model on the provision of cultural goods are not well explored yet. Generally, there is a lot of uncertainty about the actual returns to online advertising (Lewis and Rao 2015; Gordon et al. 2019). Moreover, the use of ad blocking tools can undermine generated revenues (Shiller et al. 2018). Concerns about consumer privacy may lead to regulation that restricts the use of data (Acquisti et al. 2016; Greenstein et al. 2016). Finally, provided content may be biased toward more popular topics in order to increase ad revenue (Sun and Zhu 2013).

The crowdfunding model seems to offer a viable alternative to the creative class, one that allows a more direct interaction between creator and crowd. Given that in a related industry, media, the duality of revenue streams from advertising as well as subscriptions is well-established, it might be complementary for creators to pursue both approaches.

A crowdfunded income via a membership platform like Patreon can be regarded as a crowd-based implementation of a subscription service, albeit at an individual level. The crowd provides a monthly stream of funds to creators who it deems worthy of support and it continues to do so—presumably—as long as the creator delivers what the crowd expects. A further avenue for future research will be to study whether the existing platform design provides sufficient checks and balances to motivate the creation of quality content and to discourage opportunism.