Byrnes 2014 Sci Fund P Los One
Byrnes 2014 Sci Fund P Los One
Byrnes 2014 Sci Fund P Los One
Abstract
OPEN ACCESS
Citation: Byrnes JEK, Ranganathan J, Walker
BLE, Faulkes Z (2014) To Crowdfund Research,
Scientists Must Build an Audience for Their
Work. PLoS ONE 9(12): e110329. doi:10.1371/
journal.pone.0110329
Editor: Rodrigo Huerta-Quintanilla, CinvestavMerida, Mexico
Received: June 22, 2012
Accepted: September 19, 2014
Published: December 10, 2014
Copyright: 2014 Byrnes et al. This is an openaccess article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and reproduction in any medium, provided the original author
and source are credited.
1 / 29
Introduction
Rise of Science Crowdfunding and the Decline of Public Research
Funding
Over the past five years, a new method of Internet-based fundraising known as
crowdfunding has exploded in popularity [1]. In the first six months of 2013
alone, almost US$200 million was raised for technology and arts-related projects
on just one leading crowdfunding website [2]. But what role can crowdfunding
play in the sciences? How must science adapt to take advantage of this growing
pool of available funding?
The rise of crowdfunding comes at a time when scientists are facing increasing
competition for declining sources of public funding [3]. Between 1992 and 2012,
state appropriations fell by 15% at the U.S. public research universities with the
largest research and development funding inflows [4]. Further, U.S. federal
funding for research in most physical sciences, mathematics, and engineering has
declined or remained relatively flat in inflation-adjusted purchasing power for
several decades [5]. A recent National Research Council report concluded that
federal funding for university research has been unstable overall, and is declining
in inflation adjusted dollars [6]. As one consequence, the average age of principal
investigators receiving their first major research grant (R01) from the National
Institutes of Health is 42 years old [7].
Interest in science crowdfunding is largely driven by this steady downturn in
government funding for science, particularly in the United States. Indeed, well
before crowdfunding began to catch on among scientists, Gaggioli and Riva [8]
suggested crowd-funding as a possible strategy to cope with the lack of
investments in research, as well as to increase democratization in the sciences.
Crowdfunding democratizes science funding by using a model for supporting
projects that charities have long used: aggregating small donations from many
people to achieve a successfully fund a project. The arrival of dedicated Internet
platforms truly democratized this fundraising model by removing the need for
substantial infrastructure and manpower traditionally needed for charity
fundraising. Crowdfunding now allows a wider range of potential users, including
scientists, to ask for and receive small donations. These users then become
involved in science by helping shape what projects get funded and by maintaining
their personal investment in new fields of scientific inquiry.
Crowdfunding serves a further need beyond merely funding science.
Crowdfunding provides a crucial conduit for communication between scientists
and the public. To create a crowdfunding proposal, scientists must talk about
their work in a way that appeals to people outside of the academy. They must be
good science communicators, and then are rewarded for their efforts with money
for their research.
2 / 29
3 / 29
4 / 29
demonstrate the link between online outreach and success in crowdfunding for
research dollars.
Methods
Structure of the #SciFund Challenge
#SciFund is a crowdfunding experiment for science. As part of #SciFund, we
organized scientists to run their own crowdfunding projects simultaneously for
their research under the #SciFund banner. #SciFund ran in a round-based
format, with three rounds occurring between July 2011 and December 2012. Each
round lasted several months and was divided into three phases: (1) soliciting
proposals, (2) training participants, and (3) executing proposal campaigns. In
the soliciting phase of each round, #SciFund organizers encouraged scientists
(across disciplines and countries) to participate in this crowdfunding exercise, via
e-mail lists, blog posts, and social media (e.g., Twitter and Facebook). This
soliciting phase lasted three months in the first round and one month each for the
next two rounds. To ensure scientific credibility, each scientist who signed up to
participate was vetted, via an application form that was evaluated by a science
advisory board consisting of experts in biology, physics, chemistry, and sociology
(at least two scientists who deemed their experience relevant to the project
evaluated every application). In the training phase of each round, organizers
trained the scientists to run a crowdfunding campaign via instructional blog posts
on our website (round 1: http://scifund.wordpress.com; afterwards: http://
scifundchallenge.org, all posts are still present and used regularly for new rounds),
an online discussion group, and by encouraging discussion and feedback on draft
projects and project videos within a private online space. This training phase
lasted one month in each round. By the end of the training phase, participants had
a fully formed crowdfunding proposal ready to be deployed.
In the executing phase of each round, the #SciFund crowdfunding projects and
any accompanying videos went live on the Internet. All projects within a round
launched simultaneously and ran for the same length of time. Although all
#SciFund projects were running under the same banner, each participating
scientist fundraised primarily for his or her own project (that is, there was no
collective fundraising, although during the campaign periods, the project
organizers advertised and promoted the #SciFund Challenge more broadly).
Most projects each had a single scientist behind them, but there were several
multi-researcher projects in each round. A wide range of scientific disciplines were
represented (Table 1), although most projects focused on ecology or conservation
biology, reflecting the professional networks of the #SciFund organizers. The
total number of projects and the number of days of fundraising varied with each
round (3345 days, see Table 2).
These projects were hosted on a special section of the crowdfunding platform
RocketHub (http://scifund.rockethub.com). Resulting funds were directly disbursed by RocketHub to the recipients designated by the participants (generally
5 / 29
100
Psychology
Biomedical research
Organic chemistry
Human development
Evolution
STEM education
Climate science
Computer science
Genetics
Anthropology
Applied math
Open science
Astronomy
Business research
Cancer biology
Engineering
Neuroscience
Paleontology
Political science
Seismology
Toxicology
doi:10.1371/journal.pone.0110329.t001
the participants home institution or affiliated nonprofits). The only charges that
#SciFund participants incurred were RocketHubs customary fees for crowdfunding projects running on their site (812% of the total raised, depending on
whether they achieved their funding goal). #SciFund participants received funds
even if they did not reach their financial targets, unlike the funding model for
some crowdfunding platforms, where funds are disbursed only if the project is
fully funded. It should be noted that several of this papers authors (Walker,
Table 2. Descriptive summary statistics about duration and project performance from all three rounds of the #SciFund Challenge.
Round
Dates
Projects
Projects
funded at
100%
Total raised
Project Average
Project Median
Nov. 1Dec. 45
15, 2011
49
10
20.40%
$76,230
$1,555.71
$1,104.00
May 1May 31
31, 2012
75
33
44.00%
$100,345
$1,341.37
$1,046.00
Nov 11
Dec. 15,
2012
35
16
45.70%
$75,978
$2,170.80
$1,440.00
159
59
37.10%
$252,811
Overall
Days
33
Percent
funded at
100%
doi:10.1371/journal.pone.0110329.t002
6 / 29
Byrnes, and Faulkes) ran individual crowdfunding projects under the #SciFund
banner in round one. The organizers of #SciFund were not paid by RocketHub
nor did they receive any funds, either directly or indirectly, from #SciFund
participants or donors (other than the donor funds Walker, Byrnes, and Faulkes
received from their individual projects).
Data Sources
After each of the three #SciFund rounds, we compiled data from three sources to
analyze the factors that led to successful crowdfunded projects. First, we acquired
the web visit and donation logs of each project from RocketHub. Second, we
collected publicly available information from the Internet. Each RocketHub
project page included buttons allowing visitors to tweet about the project on
Twitter (http://twitter.com), or Like the project on Facebook (http://facebook.
com). The number of tweets and Likes were publically displayed on the project
page, were updated dynamically, in real time. We recorded the number of tweets
and Likes from each #SciFund project page within hours of the campaign
ending. Thus, these are conservative measures of project promotion for these two
social media sites, because they only include button clicks on the RocketHub page,
and not tweets or Likes created by other means (e.g., copying the project URL
directly). Similarly, project videos were embedded on the RocketHub project
pages, but hosted by other websites (e.g., http://youtube.com, http://vimeo.com),
which also displayed the number of video views publically, and updated the
numbers in real time. The number of times project videos were viewed was also
collected within hours of the campaigns ending [52, 53, 54].
Last, we designed a survey for all #SciFund participants to measure: (1)
strategies used to create crowdfunding materials, (2) strategies used to promote
crowdfunding campaigns, (3) social network size (i.e., number of Facebook
friends and Twitter followers), and (4) various aspects of ongoing online outreach
activities (e.g., Do they have a blog?); see Table S2 for a complete list of questions.
This survey was completed by #SciFund participants in the first few weeks after
their crowdfunding project finished. The survey was answered by 47 of the 49
#SciFund round one participants, 48 of 75 round two participants, and 22 of 35
round three participants. The survey instrument for rounds two and three differed
in some ways from the instrument we used for round one. Specifically, we
changed the requested response for several questions from a Likert scale selection
to a specific quantitative answer (see Table S2 for complete list of changes). For
example, questions regarding the number of tweets, Facebook posts, Google+
posts, and e-mails made by participants required a numerical response in the
survey instruments for rounds two and three (where they had required a Likert
scale selection in the round one survey). We asked about number of hours spent
promoting a project, but found that these self-reported numbers proved
unreliable and were often answered qualitatively rather than quantitatively in the
survey.
7 / 29
8 / 29
For this and other analyses incorporating project page views, we excluded a
single outlier. One project had an enormous number of project page views:
38,131, compared to the mean of 2,217.75 and median of 1,070. The next highest
number of page views was 6,702. The number of page views in the most viewed
project was due to promotion on two highly popular web sites that other projects
did not have. This outlier exerted an enormous leverage on the analysis and was
therefore excluded. Analyses with this outlier project were qualitatively the same,
but quantitative results and amount of variance retained were quite different. In
analyses of future rounds, should there be a larger sample size in the 7,000
30,000-page-view range, we would be better able to detect linear or nonlinear
relationships involving this data point. For this round, the 38,131 data point was
excluded for analyses involving page views.
We next evaluated the relationship between page views and three predictors of
project popularity: the size of ones social network (Facebook friends), the size of
their outreach generated social network (Twitter followers), and the ability of a
scientist to cultivate interest in a project as measured by the number of people
who had clicked the Like on Facebook button on a projects web page. Again,
we split pre- and post-goal views. For pre-goal project page views, we fit a model
as above. For post-goal project page views, we only analyzed the subset of projects
that met their goal. Additionally, a number of projects met their goal during the
final days of #SciFund. Most of these projects had no post-goal project page
views. We therefore fit a model with a log rather than linear link function.
Last, to explore whether ongoing online outreach efforts by scientists increased
their Twitter followers, we looked at the relationship between Twitter followers
and the average number of monthly blog posts by #SciFund scientists who had
established blogs. We assumed the direction of causality went from monthly blog
posts to number of Twitter followers, because it seemed unlikely that researchers
would blog more often because they had more Twitter followers. Rather, we
hypothesized that the more frequently a researcher posted to their blog, the more
likely they would be to attract a larger following on Twitter. For participants who
did not have a blog, we set their number of monthly posts to 0. The age of these
blogs ranged from a few months to nearly ten years. As blog age and posting
frequency were highly correlated (r50.68), we did not include them as
independent measures of online outreach.
9 / 29
well as how the two interacted. We also estimated parameters for the effect of
number of people contacted via email and the number of people contacted by
project scientists in the press. We fit models with a Gaussian error term, as the
data did not meet the assumption of a mean-variance scaling relationship from a
Poisson or quasi-Poisson error distribution. We removed one outlier data point,
as its number of press contacted was two orders of magnitude larger than any
other data point, and was likely a typo on the form or a misunderstanding of the
question (post-hoc requests for verification from the participant yielded no
response). We fit this model both for total page views and pre- and post- goal
page views. However, due to the smaller sample size for post-goal page views (27)
and the high number of parameters for the model (k510), we elected to drop the
parameters assessing the impact of Google+, as they were not different from 0 in
the initial model and contributed to an exceedingly high variance inflation factor
in the post-goal page views model. Last, we fit a simple model examining to what
extent post-goal page views were merely explained by pre-goal page views, as none
of our predictors appeared to explain variability adequately. After analysis of our
increased sample size, we also recognized that Facebook Likes are often an
epiphenomenon of people visiting projects, not a causal driver. Indeed, they were
highly correlated with variables that were more causally related to effort, such as
number of press contacted (r50.76), number of Tweets (r50.61) or number of
Facebook posts (r50.81).
Results
Money Raised through the #SciFund Challenge
Over three rounds, #SciFund raised US$252,811 from 3,904 donors funding 159
projects (see Table 2 for summary statistics). The timing of donations was
relatively similar for all three rounds and conformed to what has been observed in
other crowdfunding campaigns [59]: a large amount of funds raised early in the
campaign, a gradual leveling out, and then a sudden burst of funding activity at
the end (Fig. 1).
The first round of #SciFund raised US$76,230 over 45 days from at least 1,195
donors (donor counts for rounds one and two are likely to be underestimates, as
donor names in those rounds were used to identify unique donors and multiple
donors may have had the same name). There was a large range in the financial
targets of the 49 #SciFund projects (range: US$50020,000; median: US$3,500;
average: US$4,601). Similarly, there was a large range in the amount received by
the projects, as measured by total dollars (range: US$12210,171; median:
US$1,104; average: US$1,556). The project that raised the most, both in terms of
dollars raised and percentage of goal (US$10,171 raised on a US$6,000 goal, 170%
of target fundraised), was an outlier, as the second-highest amount fundraised was
less than half of the first-place take (US$5,085). Ten projects matched or exceeded
their targets (20% of projects); all six projects that asked for US$1,200 dollars or
less met or exceeded their target.
10 / 29
Figure 1. Crowdfunding donation patterns. The daily time series of donations during the firth three rounds of #SciFund.
doi:10.1371/journal.pone.0110329.g001
Round twos 75 projects raised US$100,345 over the course of 31 days with 44%
of participants achieving or exceeding their funding goal. At least 1,579 donors
contributed to round two (likely an underestimate, as with round one, due to
shared donor names). The financial targets of round two projects tended to be
much lower than for round one and the range of dollar targets was also narrowed
(range: US$33312,000; median: US$2,000; average: US$2,215). A major reason
for these lower funding goals was that #SciFund organizers, based on round one
experience, strongly recommended that round two participants lower their
financial targets. The amounts raised in round two were within a tighter band
than in round one, but the median amount raised remained relatively steady
(range: US$305,688; median: US$1,046; average: US$1,341).
Round threes 35 projects raised US$75,978 over 33 days with 46% of projects
achieving or exceeding their goal. Round three had contributions from 1,130
donors (an exact count, unlike with rounds one and two). The financial targets of
round three projects generally rose from the levels found for round two, though
they were still lower than the targets for round one (range: US$38010,000;
median: US$2,500; average: US$3,083). In terms of the amounts actually raised,
11 / 29
round three projects were on average the most successful of the three rounds
(range: US$08,645; median: US$1,476; average: US$2,177). This is likely because
the training that the Round 3 participants received was refined based on Rounds 1
and 2, and thus more accurate and effective.
12 / 29
Figure 2.Total dollars raised plotted against the number of contributors. Line represents best fit from model described in the text. Shaded grey area
represents the 95% confidence interval around the fit relationship.
doi:10.1371/journal.pone.0110329.g002
Df
Pr(.x2)
Twitter Followers
0.041
0.84
Facebook Friends
5.397
0.02
12.849
.0.001
44.601
.0.001
doi:10.1371/journal.pone.0110329.t003
13 / 29
Table 4. Coefficient estimates, standard errors, and t-tests of predictors in analyses of number of contributors in round 1.
Estimate
Std. Error
t value
Pr(.|t|)
(Intercept)
4.497
3.925
1.146
0.263
Twitter Followers
20.001
0.006
20.224
0.825
Facebook Friends
0.019
0.008
2.301
0.03
0.009
0.003
3.544
0.002
0.048
0.009
5.139
.0.001
doi:10.1371/journal.pone.0110329.t004
Figure 3. Factors affecting number of contributors to a project. Plot shows the number of contributors plotted against the number of Facebook friends.
Size of points shows the number of page views before achieving success. Color shows the number of project page views after goals were reached with blue
representing no views to red representing many views. Line represents best fit from generalized linear model between x and y. Shaded grey area represents
the 95% confidence interval around the fit relationship.
doi:10.1371/journal.pone.0110329.g003
14 / 29
Table 5. Likelihood ratio tests evaluating predictors of pre- (a) and post-goal page views (b) in round 1.
(a)
(b)
LR x2
Df
Pr(.x2)
Twitter Followers
11.621
0.001
Facebook Friends
0.97
0.325
Facebook Likes
58.85
.0.001
Twitter Followers
0.307
0.579
Facebook Friends
1.463
0.226
Facebook Likes
8.466
0.004
doi:10.1371/journal.pone.0110329.t005
scientists Twitter network has viewed their project. Overall, our effort model
provided modest explanatory power for pre-goal page views (R250.67). Post-goal
page views seemed to be relatively uninfluenced by all factors (Table 9b). Instead,
a simple model where post-goal page views was explained by pre-goal page views
(i.e., a popular project continues to be popular) appears to provide some
explanation for post-goal page views (LR x257.09, DF51, p50.008,
slope50.1130.047 SE, intercept5118.28388.942 SE, R250.20).
(a)
(b)
Estimate
Std. Error
t value
Pr(.|t|)
(Intercept)
528.414
165.058
3.201
0.004
Twitter Followers
0.782
0.284
2.752
0.011
Facebook Friends
20.345
0.355
20.971
0.34
Facebook Likes
10.04
1.769
5.675
.0.001
(Intercept)
5.674
1.147
4.949
0.016
Twitter Followers
20.001
0.001
20.503
0.649
Facebook Friends
20.003
0.002
21.114
0.346
Facebook Likes
0.018
0.009
1.925
0.15
doi:10.1371/journal.pone.0110329.t006
15 / 29
Figure 4. Relationship between Facebook Likes, number of Twitter followers, and project page views before a project hit its goal. Line represents
best fit from model described in the text. Shaded grey area represents the 95% confidence interval around the fit relationship. Point size is proportional to the
number of Twitter followers.
doi:10.1371/journal.pone.0110329.g004
Figure 5. Relationship between Facebook Likes and the number of page views after a project has achieved its funding goal. Line represents best
fit from model described in the text. Shaded grey area represents one standard error around the fit relationship.
doi:10.1371/journal.pone.0110329.g005
16 / 29
Figure 6. Relationship between monthly blog posts and number of Twitter followers. Line represents best fit from model described in the text. Shaded
grey area represents on standard error around the fit relationship.
doi:10.1371/journal.pone.0110329.g006
analysis in that Facebook networks and sending out e-mails to social networks
were among the most important drivers of a successful crowdfunding campaign.
The other component of a successful campaign, according to the statistical
analysis, is press contacts. However, this was not considered a key reason for
success by the majority of participants. Less than 5% of the sample across the
three rounds identified #SciFund publicity (4%), national media (2%), and local
media (1%) as being important to their success.
Among the factors that did not work according to the participants, 19% of the
sample thought that engaging their online networks (Facebook, Twitter, blogging,
and Google) was unsuccessful. Related to this, 13% of the participants thought
that they did not promote their project enough (to a variety of potential networks
and press outlets). The third most cited factor considered to be unsuccessful was
having a small or non-existent online network or social media presence. These
impressions are in line with the statistical analysis in that the most frequent
answers to this question were related to engaging social networks.
Table 7. Likelihood ratio tests evaluating predictors of number of contributors in rounds 2 and 3.
LR x2
Df
Pr(.x2)
Facebook Friends
2.981
0.084
58.206
.0.001
17.797
.0.001
doi:10.1371/journal.pone.0110329.t007
17 / 29
Table 8. Coefficient estimates, standard errors, and t-tests of predictors in analyses of number of contributors in rounds 2 and 3.
Estimate
Std. Error
t value
Pr(.|t|)
(Intercept)
6.523
2.504
2.605
0.011
Facebook Friends
0.011
0.006
1.816
0.074
0.018
0.003
6.524
.0.001
0.036
0.01
3.64
0.001
doi:10.1371/journal.pone.0110329.t008
Discussion
Our analysis shows that engagement of broad audiences is the key to successful
science crowdfunding. To engage, a scientist must first build an audience for their
work, hopefully well before their crowdfunding campaign begins, such as through
the Twitter and Facebook networks we quantified here. Once the crowdfunding
begins, a scientist must then put effort into maintaining the connections between
these networks and their science, such as through tweets or direct contact via
email. Some activities, such as reaching out via the press, even accomplish the
goals of both building a wider audience and connecting these audiences to a
scientists crowdfunding proposals all at the same time. Engagement via science
communication then leads to research dollars by bringing people to view project
pages. In turn, those views translate into contributions for new scientific work (
Fig. 8; see Fig. S1 for a full path diagram with coefficients, and Figs. S2 and S3 for
a similar visualization from round 1). In short, audience multiplied by outreach
effort equals successful public engagement, and successful science crowdfunding.
Table 9. Likelihood ratio tests evaluating predictors of pre- (a,) and post-goal page views (b, c) in rounds 2 and 3.
(a)
(b)
(c)
Google+ Followers
LR x2
Df
Pr(.x2)
0.118
0.731
# of Google+ Posts
3.198
0.074
# of Twitter Followers
2.432
0.119
# of Tweets
0.189
0.663
21.47
.0.001
# of Press Contacted
33.88
.0.001
0.12
0.729
5.394
0.02
# of Twitter Followers
0.839
0.36
# of Tweets
0.348
0.555
0.072
0.788
# of Press Contacted
0.342
0.558
0.249
0.618
7.096
0.008
doi:10.1371/journal.pone.0110329.t009
18 / 29
Table 10. Coefficient estimates, standard errors, and t-tests of predictors in analyses of pre- (a, d) and post-goal page views (b, c) in rounds 2 and 3.
(a)
(b)
(c)
Estimate
Std. Error
t value
Pr(.|t|)
(Intercept)
572.711
93.726
6.11
.0.001
Google+ Followers
0.003
0.097
0.028
0.978
# of Google+ Posts
214.324
11.371
21.26
0.214
# of Twitter Followers
20.269
0.199
21.354
0.182
# of Tweets
25.025
4.06
21.238
0.221
1.72
0.371
4.634
.0.001
# of Press Contacted
92.645
15.917
5.821
.0.001
20.001
0.002
20.347
0.73
0.024
0.014
0.006
2.323
(Intercept)
156.213
57.159
2.733
0.012
# of Twitter Followers
0.005
0.21
0.023
0.982
# of Tweets
2.04
2.732
0.747
0.463
20.05
0.188
20.268
0.791
# of Press Contacted
6.263
10.703
0.585
0.564
20.002
0.003
20.499
0.623
(Intercept)
118.283
88.943
1.33
0.194
0.114
0.043
2.664
0.012
doi:10.1371/journal.pone.0110329.t010
Figure 7. Relationship between pre-goal page views, press contacts, number of people emailed, and effort times engagement on Twitter. Line
represents best fit from model between press and pre-goal page views. Shaded grey area represents the 95% confidence interval around the fit relationship.
doi:10.1371/journal.pone.0110329.g007
19 / 29
Table 11. Factors mentioned by SciFund project creators that helped with project fundraising.
Factor
Round 1 (n547)
Round 2 (n549)
Round 3 (n522)
43 (36%)
17 (36%)
18 (37%)
8 (36%)
Personal networks
43 (36%)
13 (28%)
23 (47%)
7 (32%)
Online networks
37 (31%)
20 (43%)
7 (14%)
10 (45%)
Effective video
13 (11%)
7 (15%)
4 (8%)
2 (9%)
8 (7%)
6 (13%)
2 (4%)
5 (4%)
4 (9%)
1 (5%)
4 (3%)
4 (9%)
National media
2 (2%)
2 (4%)
Tastemaker involvement
2 (2%)
2 (4%)
Local media
1 (1%)
1 (2%)
Luck
1 (1%)
1 (2%)
Merton effect
1 (1%)
1 (2%)
Rewards
1 (1%)
1 (2%)
1 (1%)
1 (2%)
Round 1 (n547)
Round 2 (n549)
Round 3 (n522)
23 (19%)
10 (21%)
7 (14%)
6 (27%)
15 (13%)
8 (17%)
5 (10%)
2 (9%)
14 (12%)
8 (17%)
4 (8%)
2 (9%)
10 (8%)
3 (6%)
3 (6%)
4 (18%)
10 (8%)
3 (6%)
6 (12%)
1 (5%)
8 (7%)
6 (13%)
2 (4%)
7 (6%)
4 (9%)
3 (6%)
6 (5%)
4 (9%)
2 (4%)
5 (4%)
3 (6%)
1 (2%)
1 (5%)
4 (3%)
2 (4%)
1 (2%)
1 (5%)
4 (3%)
4 (8%)
1 (1%)
1 (2%)
1 (1%)
1 (2%)
1 (1%)
1 (2%)
1 (1%)
1 (5%)
20 / 29
This occurs largely before the crowdfunding campaign begins, and time invested in
engagement yields a larger audience and proportionately greater funds raised. For
example, our analyses suggested that Twitter and Facebook network size influences
project success. While some of this did indeed come from family and friends, a
scientist cannot grow these audiences. Rather, they must build other audiences, as
reflected by contributions of Twitter, emails to supporting organizations, and direct
contacts with the press. Additional forms of outreach to build ones scientific fan
base not measured by our survey (e.g., involvement with museums, public lectures,
TEDx talks, authoring popular science articles for a newspaper, media history, etc.)
quite likely help in crowdfunding a project. These kinds of community engagement
activities may facilitate access to local mailing lists as well as the likelihood of a press
contact translating into an article. All of these forms of audience building
demonstrate the importance of building and maintaining a consistent public
presence for raising money through crowdfunding.
21 / 29
that audience wont come. However, in the survey data, many scientists admitted
to doubts that their efforts were successful. The quantitative data, in contrast,
shows that while promotion of a crowdfunding project may at times feel like
shouting into the void, the effort can and will lead to success. During a
crowdfunding campaign, more effort that is more tweets, more emails sent,
more people in the press contacted - all led to higher funding. Crowdfunding
takes effort. Informally, some successful participants reported spending K
1 hour per day on outreach during their crowdfunding campaign period. Note
that this is after the time-intensive process of producing crowdfunding materials,
such as a short video, necessary to engage with a broad non-expert audience.
These activities are different from the traditional grant-writing models that are
comfortable for most scientists. Rather, these are the activities of a successful
outreach program, but with the added benefit of research funding for the time
invested.
22 / 29
(Figure S5), and after a recent fourth round (run on a different platform,
Experiment.com, and hence not included here for analysis), scientists are now
achieving a 62.5% success rate.
Furthermore, since the inception of #SciFund, several science crowdfunding
projects have raised substantially more money than the most successful #SciFund
projects. Two projects investigating the bacterial communities associated with
humans each raised over US$300,000 [60, 61]. A project to launch a space
telescope raised over US$1,000,000 [62]. The difference between these projects
and #SciFund projects was rewards that directly involve citizens in the scientific
process. Donors funding the two microbial projects at a certain minimum level
had their very own bacterial communities analyzed by those projects. Funding the
space telescope at high levels gave funders direct access to time on the telescope.
Examples of US$100K+ science crowdfunding efforts reinforce the basic lessons
of our analyses. The scientists behind these high-earning crowdfunding campaigns
also went to great lengths to promote their work. But more importantly, they went
to extreme lengths to engage citizens in their scientific process. Audiences were
captivated by taking part themselves in microbial and space research. They will
likely be engaged with those scientific groups for years to come, potentially
crowdfunding future projects.
23 / 29
Supporting Information
Figure S1. The pathway of interactions leading to money raised for projects in
round two and three. Diagram shows the relationships between different variables
in our analyses. Only those relationships that explained significant amounts of
variation are included (LR 42 test p#0.05). Coefficients represent linear
relationships and are in the units of variables described. Sample size varies
24 / 29
Acknowledgments
We thank K. Kilgrove, S. Hampton, J. Stachowicz, N. Baron, M. Miner, J. Fischer,
K. James, K. Weinersmith, and J. Balch for comments on early versions of this
manuscript. JB, BW, and ZF raised money for their individual research projects
through #SciFund. The authors made no money from organizing #SciFund. We
25 / 29
thank Brian Meece, Vlad Vukicevic, Jed Cohen, and Alon Hillel-Tuch (founders
of RocketHub.com) for providing an online platform for #SciFund Rounds 13,
and for contributing server data. We thank the 49 participants of round 1 of
#SciFund for trying something new, and the 110 participants of round 2 and 3 for
furthering the world of science crowdfunding.
Author Contributions
Conceived and designed the experiments: JB JR. Performed the experiments: JB
JR. Analyzed the data: JB. Contributed reagents/materials/analysis tools: JB JR BW
ZF. Wrote the paper: JB JR BW ZF. Designed the survey instrument: JB JR BW.
Collected the data: JB JR BW ZF.
References
1. Ordanini A, Miceli L, Pizzetti M, Parasuraman A (2011) Crowd-funding: transforming customers into
investors through innovative service platforms. Journal of Service Management 22: 443470.
2. Gamerman E (2013) The trouble with Kickstarter. Wall Street Journal D1-D2. Available: http://online.wsj.
com/news/articles/SB10001424127887324021104578551313657138252. Accessed
3. Intersociety Working Group (2012) Research and development FY 2013. Washington, D.C. 290 p.
4. National Science Board (2012) Diminishing Funding and Rising Expectations: Trends and Challenges
for Public Research Universities, A Companion to Science and Engineering Indicators 2012. In: National
Science Foundation, editor. Arlington, Virginia: National Science Foundation.
5. Committee on Prospering in the Global Economy of the 21st Century (2007) Rising above the
gathering storm: Energizing and employing America for a brighter economic future. Washington, D.C.:
National Academies Press.
6. Committee on Research Universities (2012) Research Universities and the Future of America: Ten
Breakthrough Actions Vital to Our Nations Prosperity and Security. Washington, D.C.: National
Academies Press.
7. Rockey S (2012) Our commitment to supporting the next generation. Rock Talk Available: http://nexus.
od.nih.gov/all/2012/02/03/our-commitment-to-supporting-the-next-generation. Accessed 13 May 2014.
8. Gaggioli A, Riva G (2008) Working the crowd. Science 321: 1443.
9. Palmer A (2012) How Amanda Palmer Built An Army Of Supporters: Connecting Each And Every Day,
Person By Person. Case Studies by TechDirt. Available: http://www.techdirt.com/blog/casestudies/
articles/20120502/15324918745/how-amanda-palmer-built-army-supporters-connecting-each-everyday-person-person.shtml. Accessed 14 May 2014.
10. Gobble MAM (2012) Everyone is a venture capitalist: the new age of crowdfunding. Research
Technology Management 55: 45.
11. Wasik J (2012) The Brilliance (and Madness) Of Crowdfunding. Forbes. pp. 144146.
12. Rubin S (2012) The wison of crowdfunding. Forbes. pp. 62.
13. Greenwald T (2012) Crowdfunding. Technology Review 116: 117128.
14. Valanciene L, Jegeleviciute S (2013) Valuation of crowdfunding: benefits and drawbacks. Economics
and Management 18: 3948.
15. Hughes V (2012) Strapped for funding, medical researchers pitch to the crowd. Nature Medicine 18:
1307.
16. Wheat RE, Wang Y, Byrnes JE, Ranganathan J (2013) Raising money for scientific research through
crowdfunding. Trends in Ecology & Evolution 28: 7172.
26 / 29
17. Johnson DR, Ecklund EH, Lincoln AE (2014) Narratives of science outreach in elite contexts of
academic science. Science Communication 36: 81105.
18. Burns TW, OConnor DJ, Stocklmayer SM (2003) Science communication: a contemporary definition.
Public Understanding of Science 12: 183202.
19. Pinholster G, Ham B (2013) Science communication requires time, trust, and Twitter. Science 342:
14641464.
20. Kouper I (2010) Science blogs and public engagement with science: Practices, challenges, and
opportunities. Jcom 9: A02.
21. Stilgoe J, Lock SJ, Wilsdon J (2014) Why should we promote public engagement with science? Public
Understanding of Science 23: 415.
22. Mitchell K (2008) Practising public scholarship: experiences and possibilities beyond the academy.
Oxford: Wiley-Blackwell.
23. Sismondo S (2008) Science and technology studies and an engaged program. In: Hackett EJ,
Amsterdamska O, Lynch M, Wajcman J, editors. The Handbook of Science and Technology Studies,
Third Edition. Cambridge: The MIT Press. pp. 1331.
24. Russo G (2010) Outreach: Meet the press. Nature 468: 465467.
25. Silka L (2010) Community research in other contexts: learning from sustainability science. Journal of
Empirical Research on Human Research Ethics 5: 311.
26. Ellison J, Eatman TK (2008) Scholarship in public: Knowledge creation and tenure policy in the
engaged university. Syracuse, New York.
27. Green R (2008) Tenure and promotion decisions: the relative importance of teaching, scholarship, and
service. Journal of Social Work Education 44: 117128.
28. Poliakoff E, Webb TL (2007) What factors predict scientists intentions to participate in public
engagement of science activities? Science Communication 29: 242263.
29. The Royal Society (2006) Survey of factors affecting science communication by scientists and
engineers. London: The Royal Society.
30. Nyden P (2003) Academic incentives for faculty participation in community-based participatory research.
Journal of General Internal Medicine 18: 576585.
31. Weigold MF (2001) Communicating science: A review of the literature. Science Communication 23:
164193.
32. Gascoigne T, Metcalfe J (1997) Incentives and impediments to scientists communicating through the
media. Science Communication 18: 265282.
33. Holbrook JB (2005) Assessing the sciencesociety relation: The case of the US National Science
Foundations second merit review criterion. Technology in Society 27: 437451.
34. Holbrook JB (2009) Editors Introduction. Social Epistemology 23: 177181.
35. Ahmed SM, Palermo AGS (2010) Community engagement in research: frameworks for education and
peer review. Journal Information 100: 13801387.
36. Hackett E, Rhoten D (2011) Engaged, embedded, enjoined: science and technology studies in the
National Science Foundation. Science and Engineering Ethics 17: 823838.
37. Kennedy C, Vogel A, Goldberg-Freeman C, Kass N, Farfel M (2009) Faculty perspectives on
community-based research: I see this still as a journey. Journal of Empirical Research on Human
Research Ethics 4: 316.
38. Williams K (2012) The devil is in the details: community based participatory research. Journal of Cancer
Education 27: 34.
39. Delgado A, Lein Kjlberg K, Wickson F (2011) Public engagement coming of age: From theory to
practice in STS encounters with nanotechnology. Public Understanding of Science 20: 826845.
40. Resnik DB, Kennedy CE (2010) Balancing scientific and community interests in community-based
participatory research. Accountability in Research 17: 198210.
27 / 29
41. Goldberg-Freeman C, Kass N, Gielen A, Tracey P, Bates-Hopkins B, et al. (2010) Faculty beliefs,
perceptions, and level of community involvement in their research: a survey at one urban academic
institution. Journal of Empirical Research on Human Research Ethics: An International Journal 5: 6576.
42. Meyer JL, Frumhoff PC, Hamburg SP, de la Rosa C (2010) Above the din but in the fray:
environmental scientists as effective advocates. Frontiers in Ecology and the Environment 8: 299305.
43. Scott JM, Rachlow JL, Lackey RT, Pidgorna AB, Aycrigg JL, et al. (2007) Policy advocacy in
science: prevalence, perspectives, and implications for conservation biologists. Conservation Biology
21: 2935.
44. Waananen L (2012) Three years of Kickstarter projects. Available: http://www.nytimes.com/interactive/
2012/04/30/technology/three-years-of-kickstarter-projects.html. Accessed 9 May 2012.
45. Pebble Technology (2012) Pebble: E-paper watch for iPhone and Android. Available: http://www.
kickstarter.com/projects/597507018/pebble-e-paper-watch-for-iphone-and-android. Accessed 9 May
2012.
46. Portalarium Inc (2013) Shroud of the Avatar: Forsaken Virtues. Available: https://www.kickstarter.com/
projects/portalarium/shroud-of-the-avatar-forsaken-virtues-0. Accessed 14 May 2013.
47. Ecklund EH, James SA, Lincoln AE (2012) How academic biologists and physicists view science
outreach. PLoS ONE 7: e36240.
48. Wilcox C (2012) Its time to e-volve: taking responsibility for science communication in a digital age. The
Biological Bulletin 222: 8587.
49. Shema H, Bar-Ilan J, Thelwall M (2012) Research blogs and the discussion of scholarly information.
PLoS ONE 7: e35869.
50. Procter R, Williams R, Stewart J, Poschen M, Snee H, et al. (2010) Adoption and use of Web 2.0 in
scholarly communications. Philosophical Transactions of the Royal Society A: Mathematical, Physical
and Engineering Sciences 368: 40394056.
51. Priem J, Piwowar HA, Hemminger BM (2012) Altmetrics in the wild: Using social media to explore
scholarly impact. arXiv:1203.4745 [cs.DL].
52. Open Science Federation, Byrnes J, Faulkes Z, Ranganathan J, Open Science(2012) SciFund
Success and Social Media. Available: http://figshare.com/articles/SciFund_Success_and_Social_Media/
90818. Accessed 13 May 2014.
53. Faulkes Z (2012) #SciFund Round 2 success and social media. Available: http://dx.doi.org/10.6084/m9.
figshare.92541. Accessed 13 May 2014.
54. Faulkes Z (2012) #SciFund Round 3 success and social media. Available: http://dx.doi.org/10.6084/m9.
figshare.104620. Accessed 13 May 2014.
55. McCullagh P, Nelder JA (1989) Generalized Linear Models, Second Edition. Boca Raton: Chapman &
Hall/CRC.
56. Ver Hoef JM, Boveng PL (2007) Quasi-Poisson vs. negative binomial regression: how should we model
overdispersed count data? Ecology 88: 27662772.
57. R Development Core Team (2012) R: A language and environment for statistical computing. Vienna
Austria.
58. Cameron AC, Windmeijer F (1996) R-squared measures for count data regression models with
applications to health-care utilization. Journal of Business & Economic Statistics 14: 209220.
59. Perlstein E (2013) Microryza campaign trajectories. Perlstein Lab Available: http://www.perlsteinlab.
com/slideshow/successful-science-crowdfunding-trajectories. Accessed 14 May 2014.
60. Human Food Project (2013) American Gut - whats in your gut? Available: https://www.indiegogo.com/
projects/american-gut-what-s-in-your-gut-7. Accessed 14 May 2014.
61. The uBiome Team, Richman J, Ludington W, Apte Z (2013) uBiome Sequencing Your Microbiome.
Available: http://www.indiegogo.com/projects/ubiome-sequencing-your-microbiome. Accessed 14 May
2014.
62. Planetary Resources (2013) ARKYD: A Space Telescope for Everyone. Available: http://www.
kickstarter.com/projects/1458134548/arkyd-a-space-telescope-for-everyone-0. Accessed 28 October
2013.
28 / 29
63. Barry D, Oelschlaeger M (1996) A science for survival: values and conservation biology. Conservation
Biology 10: 905911.
64. Laursen S, Liston C, Thiry H, Graf J (2007) What good is a scientist in the classroom? Participant
outcomes and program design features for a short-duration science outreach intervention in K12
classrooms. CBE-Life Sciences Education 6: 4964.
29 / 29