Beware of Your Own Biases When Analyzing Social Media
1. Beware of Your Own Biases When Analyzing Social Media
Abstract – Social media analyses are vulnerable to biases of those performing the analysis – a critical
point to consider when attempting to analyze your competitor’s use of social media. iMedSocial
uncovered one potential source of bias when performing an analysis of a recent news phenomenon on
social media, which is the difference between how individuals involved in an activity discuss that
activity, as opposed to how it is discussed by traditional news media.
Introduction -
Social media is an excellent communication tool for businesses to reach out to their customers for two-
way, interactive communication. As a result, social media metrics have become an important analytical
tool for businesses to understand whether their social media campaigns are successful, how to define
success in social media and also how to improve their customer and partner outreach efforts. Social
media analysis in general can assist businesses to learn how their competitors use these important tools
and whether they are successful at using these tools, as well as providing significant information on
social trends. However, it is important to be aware of your own biases when analyzing social media.
As a case in point, we recently analyzed a hot news story on children engaging in risky behaviors in the
form of stunts while recording a video of the stunt for posting on Youtube. Although our research
showed that this phenomenon had been occurring for some time, and had also been discussed on social
media for some time, it only made headlines with the tragic death of a teenager, David Nuno, who died
on Sept 25 due to such a dangerous stunt (http://abcnews.go.com/Health/dangerous-stunts-youtube-
hurting-killing-teens/story?id=17342485&partner=skygrid ). This specific story, as well as a more general
discussion of the dangers of these stunts, then rapidly spread on-line. However, we found that the
words used to discuss the stunts, and the types of social media channels on which the stunts were
discussed, differed greatly for the individuals actually performing the stunts as opposed to on-line
discussions by outsiders. By “outsiders” we mean those individuals who do not perform these stunts,
but who instead relate to the stunts as a news item only. The consumers of such news items may be
peers of the subject, parents or totally objective readers.
Our analysis – news media vs. “chubby bunny”
We first looked at terms that describe the stunts from such an outsider, news-only perspective, by
searching through social media from January 30 to September 30 2012, using terms that related to
children and teenagers, stunts, and terms signifying danger or injury, or even death. We found a total of
6,067 hits for all types of social media, using the above terms.
We then drilled down within the results, to look for hits that also mentioned the term “Youtube”, and
found 6027 hits – nearly all of the hits for the above terms.
info@iMedSocial.com
www.iMedSocial.com
2. 2
Figure 1 shows the daily volume for all types of social media with these terms; the highest peak is seen
on August 6 2012, followed by another but smaller peak on September 14 and a still smaller peak on
September 25 2012. This last peak is when the story of these stunts made widespread headlines, but in
social media channels, the story was already becoming less popular.
Figure 1 – stunts as a news story, all hits including Youtube as a term
Figure 2 shows that these three peaks most clearly match to discussion of these stunts on traditional on-
line media sources (three yellow peaks shown below in Figure 2). Microblog hits, such as for Twitter for
example, only start to form a peak as the story broke into the wider news media in the last peak on
September 25 2012. For the entire time period overall, traditional on-line media sources accounted for
more than 60% of all hits, followed by blogs (about 18%). Microblogs only accounted for 1% of the
overall hits, while photo and video sharing sites only had two hits over the entire time period.
3. 3
Figure 2 – most hits relating to the stunts only as news were found on traditional on-line news media
Indeed, Youtube failed to emerge as a top domain for the period of January 30 to September 30 2012
with these keywords; even when the search was restricted to the month of September 2012, Youtube
failed to emerge as a top domain. Instead, the top domains for both periods were all related to
traditional on-line media and news aggregators.
This latter result, in which terms such as “dangerous” and “stunts”, failed to yield hits on photo/video
sharing sites, and in fact that Youtube failed to emerge as a top domain for these terms, is not surprising
when these terms are considered in context of the individuals performing the stunts. Clearly these
individuals don’t expect these stunts to be dangerous, but rather “cool” or at least admirable, and in
fact, as described in greater detail below, they don’t even categorize these acts as “stunts”.
We next looked at terms that describe the stunts from the perspective of those performing them, by
searching through social media from January 30 to September 30 2012, using terms that related to video
and one stunt in particular, known as “chubby bunny” (Wikipedia provides a description of this stunt -
http://en.wikipedia.org/wiki/Chubby_bunny), which has led to deaths in the past. We had to remove
the word “stunt” from our search in order to generate any hits. We found a total of 1,794 hits for all
types of social media, using the above terms.
Figure 3 shows that most of the hits were actually found on microblogs, with peaks at the end of July
and again on September 1 2012. Indeed, nearly 70% of the hits using these “insider” terms were found
on microblogs, followed by video/photo sharing sites with about 17% of the hits. Social networks, such
as Facebook, and blogs together accounted for about 10% of the hits. The lower red line at the bottom
4. 4
of the graph indicates hits on video/photo sharing sites; interest in this stunt on these stunts had
generally ended by the time the first major peak on microblogs is seen at the end of July.
Figure 3 – hits on the term “chubby bunny” were mainly found in microblog entries
However, perhaps more interestingly, the discussion of the stunt “chubby bunny” was dying down in
social media as discussion of such stunts generally was rising in the mainstream news media. In other
words, it appears that interest was gradually waning, at least for this stunt, in social media, while
increasing in traditional on-line news media (represented by the far right blue peak on the graph, at the
end of September).
Beware of biases
Apart from the clear disconnect between traditional news media, including traditional on-line news
media, and other types of social media such as microblogs and video/photo sharing sites, it is clear that
discussions of such phenomena can start in one type of social media, such as video/photo sharing sites,
lose traction in that channel and then move to another channel, such as microblogs – and finally cease
to be of interest to social media channels, even while their popularity is rising in more traditional news
media.
Furthermore, the types of analyses performed on social media dictate the results obtained. From the
first set of search terms we considered, it would appear that such dangerous stunts had suddenly spiked
5. 5
in popularity at the end of September 2012; but the second set showed that at least for one type of
stunt, chubby bunny, interest was already waning at that time. Thus, it is important to be careful to
consider potential biases when analyzing social media, in order to fully understand how different social
media channels are used for communication of different topics.
If you are interested in our social media analyses, please email us at info@imedsocial.com. We would be
happy to discuss such analyses and our tips for successful social media campaigns, particularly in the
areas of hospitals and medical care providers, and also life science companies, including pharmaceutical,
biotech and medical device companies, as well as suppliers to hospital and medical care providers.