What to Read Next
The Holy Grail of modern online marketing is video content that “goes viral,” meaning that it captures an enormous number of views and leads audiences to share, comment or click that they “like” a video. Advertisers, marketing consultants and filmmakers have all ventured theories about what kind of content makes for a hit. The trouble is that the advice varies widely and is even contradictory. Depending on the expert, success is thought more likely if a video is humorous, shocking, dramatic, topical, warm, arousing, angry, scary, socially beneficial, cute, violent, sexy, uplifting, intriguing, quirky, interesting, authoritative, tear-jerking, educational, controversial or baby- and animal-filled.
One of the reasons for the wide range of recommendations is that researchers have often looked at only popular videos. For example, one study that tracked the distribution of videos on Facebook focused only on those that were shared most often, which meant that the researchers did not compare the most popular clips with the content almost no one saw. Nor have marketing scholars reached conclusions about the characteristics of other kinds of popular shared content, despite the vast amount of data now available. One of the more successful papers on viral messages,1 which looked at forwarding behavior in viral email marketing campaigns, suggested that many emotions can play a role, including surprise, joy, sadness and fear. Another study looked at the sharing of New York Times articles and found that still other emotional responses, such as awe and anxiety, also predicted sharing.2
To see if we could clarify some of the contradictions, we decided to take a different approach. Rather than catalog a hodgepodge of content elements found in popular videos, we examined a mix of popular and unpopular videos, then systematically coded and empirically tested the effect of each element on some relatively objective and observational measures of viewer engagement.
First, we gathered a data set of 750 YouTube videos that varied across a wide range of topic categories (including automotive, comedy, gaming and politics) and a wide range of success in gaining viewership. (See “About the Research.”) We excluded music videos because they were associated with an exceptional number of views (in fact, 29 of the top 30 YouTube videos of all time are music videos) and would therefore skew our analysis. We also excluded videos longer than 10 minutes because these were a specialized type of content likely to be governed by different rules than the majority of YouTube content. (Think a 15-minute-long instructional video on using a pressure cooker, compared to a video of a digitized cat with wings on a rainbow.)
We assigned a team of research assistants as blind coders to watch the videos and to independently score each on a range of attributes. Did the video feature babies, attempt to be funny or use sexually suggestive content? How would watching the video make the typical viewer feel (for example, sad or surprised)? We collected information on dozens of different video elements and correlated these with three measures of engagement: the number of times people left comments on the video, the overall “liking” index for each video (calculated by subtracting the number of “dislikes” from “likes”) and the number of views.
We also coded for three content themes that we hypothesized might be particularly powerful in creating engagement because they are likely to induce a strong emotional reaction: novelty, incongruity and hyperbole. The first, novelty, was chosen because we know that people enjoy experiences that are new and original. For example, research suggests consumers pay more attention to novel advertisements and public service announcements.3 New or “fresh” content creates feelings of surprise, pleasure, entertainment and interest.4 We thought this might also be true online, so we coded for video novelty.
Second, we assessed incongruity, which is the presentation together of two contradictory or unrelated things.5 Incongruity forms the basis of much humor, but serious videos with incongruous content may also be more engaging than those without it. For example, the famous “Kony 2012” video6 discussed crimes against humanity, and that topic was juxtaposed against the youth and hopefulness of the protagonist. It was not remotely funny, yet received almost 100 million views. We thought the concept of incongruity might help to explain this type of engagement.
Third, we looked at hyperbole, which essentially boils down to the use of excess or exaggeration.7 This included content that was very dramatic, extravagant or even ridiculous. Video creators hope that, by making an over-the-top claim or elaborating effusively, they can more easily get and hold the viewer’s attention. We believed hyperbolic content might be an effective way to create a strong emotional reaction and engage audiences.
What Is Engagement?
One reason so little is understood about video engagement is that the phenomenon is less than 20 years old. The first real example of the power of engaging online content was the “Dancing Baby” video, created as a product demonstration by a 3-D character animation software development team. Released in 1996, the Dancing Baby “ooga-chuckah’d” its way across Internet forums, websites and email inboxes.8 Though a relatively low-budget animation, the Dancing Baby video captivated viewers’ attention and became a cultural phenomenon noted around the water cooler and even on the evening news. In 1998, the Dancing Baby landed on the hit TV comedy “Ally McBeal,” cementing the video’s status as a cultural icon.
Perhaps because the phenomenon is relatively young, scholars have not yet even reached a consensus about what engagement is,9 and we won’t try to settle that debate here. We take a broad view, and define engagement as behavior that includes sharing but also extends to other forms of measurable user involvement. This view is in line with TV advertising norms, which link engagement loosely to attention and viewer interaction.10 Ultimately, it seems likely that different engagement-related behaviors — commenting, sharing, “favoriting,” “liking” and so on — are highly correlated.11 For this study, we measured engagement using YouTube commenting, liking and viewing behaviors. These measures of engagement were readily available for the full range of videos we studied (both successful and unsuccessful), which allowed us to perform a representative content analysis.
Our study allowed us to dispel a number of myths about online videos. For one thing, when it comes to garnering views, professionals seem no better than amateurs at creating compelling content. We found no difference in the aggregate number of comments between homemade and commercially created videos. While branded videos tend to be liked more, they are also disliked more.
The presence of babies also did not have any impact on views or comments. Sure, babies are cute and give people warm feelings, but they don’t seem to increase viewer engagement. Safe creative choices are less risky, but because they don’t generate much emotion, they are also less likely to engage.
Likewise, videos of animals do go viral sometimes, but probably not because they contain animals. Rather, we think engagement with animal videos happens because many of these videos contain animals doing something surprising. In other words, it’s not about the animal per se but about the feeling of surprise inspired by using novel or incongruous content such as Grumpy Cat,12 an apparently frowning cat whose owners signed an endorsement deal with Nestlé Purina PetCare. One might think that anthropomorphized animals such as dancing dogs, trumpet-playing ponies or talking mice would be old hat by now, but it seems not. Among Internet viewers, all those talented animals are apparently still novel and incongruous enough to inspire engagement. Our findings that babies and animals in general are not particularly effective are consistent with other research.13
We also found that attractive people don’t engage viewers particularly well. Videos with attractive actors were linked to lower levels of surprise, which in the context of online videos is not a good thing. Maybe this is because it’s hardly unusual to see beautiful people in front of the camera. In fact, it’s perhaps entirely too expected, which makes it an ineffective way to get people’s attention. Even sexually suggestive content doesn’t seem to especially interest people.
Nor are people especially interested in seeing stories about the triumph of underdogs. We looked at a lot of videos that contained the idea of the underdog and found evidence that they don’t engage viewers. Perhaps they have become a cliché.
Other kinds of content did drive more comments and views but could be problematic for advertisers. Satire was associated with hyperbole and feelings of anger and with high levels of both comments and views. If the goal is to engage viewers by creating anger-inducing content, satire seems to be a good way to stir up a hornet’s nest.
People also liked what we called “stunts and amazing feats.” Watching other people do remarkable things gave viewers feelings of surprise and sometimes fear, which didn’t seem to hurt engagement and led to increased views. This finding probably explains why the video of Felix Baumgartner performing the world’s highest skydive was one of the top 10 videos of 2012.14
Disgusting content also seems to boost views. If it seemed exaggerated, viewers sometimes felt angry, but other times it was seen as novel, incongruous and surprising — and led to more liking and views. It is likely a polarizing creative choice, repelling certain viewers (such as older consumers) while attracting others (such as teenagers).
Finally, anger- and fear-inducing content seem to trigger more views, but make a sad video and you’ll probably cry alone.
The Element of Surprise
What matters most? Surprise!
In our study, emotionally surprising videos generated liking and views more than any kind of specific content element we studied. Surprise had indirect effects, too. In addition to the direct positive effect surprise had on liking and views, it was linked to fear. Frightening videos got significantly more views. This probably does not reflect a “horror movie” effect (where audiences enjoy being frightened) because the overall liking index of such videos also went down. This suggests that fearful content can increase views but at the risk of generating negative attitudes, whereas content that is surprising but not scary doesn’t seem to have a downside, because both liking and views increase.
Hyperbole, Incongruity and Novelty
We started this study with three ideas about how a strong emotional response like surprise might be created: novelty, incongruity and hyperbole. No matter how we looked at it, there was never a connection between hyperbole and surprise. This suggests that in terms of creating unexpected or fresh content, exaggeration is not the way forward. However, hyperbole had effects that we didn’t anticipate: We found it correlated with increased anger, which in turn boosted comments and views. This suggests that using exaggerated content may earn views, but at the expense of angering customers. Perhaps some “extreme” brands or certain types of organizations (for example, political campaigns or nongovernmental organizations) can use hyperbolic themes without fear of damage to their reputation, but for most business-content producers, this option will be problematic. The creative choices most strongly linked with hyperbole were satire and disgusting content, whereas nonsatiric humor was not.
When we looked at novel and incongruous content, we found that both were associated with feelings of surprise, which increased views and liking. This suggests to us that if a marketer’s goal is to get viewers’ attention by surprising them, they have two good choices: Show them something they have never seen before, or show them two things they are familiar with but in an original, juxtaposed way — to “make it new,” as the poet Ezra Pound once advised.
Why Do People Share Videos?
A final important question, but one that goes beyond the scope of this study, is why people share videos. Recent research has shown that people are more likely to share videos linked to brands they like.15 This makes sense, because a consumer who feels connected to a particular brand will do all sorts of exceptional things for that brand, such as paying a premium for it and recommending it to friends. We also know that people prefer to share videos that they have received from their own friends and family because it’s easy and safe to pass along something that is socially sanctioned.16 These two factors suggest that if businesses create compelling video content, then pushing this content to brand fans or other key influencers might result in a cascade of sharing.
Other researchers have found that videos with superior creativity are better liked and are both forwarded and viewed more.17 This research doesn’t provide clear definitions of what is meant by creativity,18 but we speculate that our study might provide clarification: Creativity often is evident in contexts where something is novel or juxtaposed for a dramatic or comic effect, such as a hamster eating a tiny slice of pizza off of a china plate on a picnic blanket.19 Our results confirm that both approaches are likely to be effective in engaging audiences because of their powerful influence in surprising viewers. What else might make a video seem more creative? Our research suggests some candidates in elements such as humor, stunts and amazing feats, and disgusting content, which all drive higher views.
Much of the academic research on viral videos has tried to contribute a better understanding of the role of emotions in the decision to share a video by focusing on the intensity of the emotion that a person feels watching a video and whether that emotion is positive or negative. For example, the results of several recent papers seem to suggest that more emotionally arousing content is shared more, but being positively arousing (for example, hilarity) generates more sharing than being negatively arousing (for example, shock).20 The main conclusion of these studies, as in ours, is that, while it pays to be intense, to be intense in a positive way usually pays off. These results are largely consistent with ours and seem to point in the same direction: If emotionally intense videos prompt more sharing — whether positive or negative — then, generally speaking, those videos are watched more often.21 For example, we find that anger-inducing videos can lead to high numbers of views, just as joy-inducing ones can. But our research suggests that an element of surprise, a highly arousing emotion that can be either positive or negative, may produce the strongest viewer reaction of all.
1. While the following paper only considers successful (branded) viral marketing campaigns, it is a serious, systematic effort to understand why sharing occurs; see A. Dobele, A. Lindgreen, M. Beverland, J. Vanhamme and R. van Wijk, “Why Pass on Viral Messages? Because They Connect Emotionally,” Business Horizons 50, no. 4 (July-August 2007): 291-304.
2. J. Berger and K.L. Milkman, “What Makes Online Content Viral?,” Journal of Marketing Research 49, no. 2 (April 2012): 192-205.
3. See, for example, G. Fennell, “Attention Engagement,” Current Issues & Research in Advertising 2, no. 1 (1979): 17-33; and L. Shen and E. Bigsby, “Behavioral Activation/Inhibition Systems and Emotions: A Test of Valence vs. Action Tendency Hypotheses,” Communication Monographs 77, no. 1 (March 2010): 1-26.
4. See P.J. Silvia, “Interest: The Curious Emotion,” Current Directions in Psychological Science 17, no. 1 (February 2008): 57-60; and J.L. Sherry, “Flow and Media Enjoyment,” Communication Theory 14, no. 4 (November 2004): 328-347.
5. A.P. McGraw and C. Warren, “Benign Violations: Making Immoral Behavior Funny,” Psychological Science 21, no. 8 (2010): 1141-1149.
6. Invisible Children, “Kony 2012,” March 5, 2012, www.youtube.com.
7. L. Cano Mora, “All or Nothing: A Semantic Analysis of Hyperbole,” Revista de Lingüística y Lenguas Aplicadas 4 (October 2010): 25-35.
8. Awesome, “Dancing Baby,” 2010, http://knowyourmeme.com.
9. For a good discussion, see B.J. Calder, E.C. Malthouse and U. Schaedel, “An Experimental Study of the Relationship Between Online Engagement and Advertising Effectiveness,” Journal of Interactive Marketing 23, no. 4 (November 2009): 321-331.
10. R.G. Heath, “Emotional Engagement: How Television Builds Big Brands at Low Attention,” Journal of Advertising Research 49, no. 1 (2009): 62-73.
11. Business scholar A.J. Mills argues that the willingness to consume is highly related to the willingness to distribute — meaning sharing and views are likely to be positively correlated; see A.J. Mills, “Virality in Social Media: The SPIN Framework,” Journal of Public Affairs 12, no. 2 (May 2012): 162-169.
12. CNN, “Grumpy Cat Goes From Meme to the Big Screen,” May 30, 2013, www.youtube.com.
13. K. Nelson-Field, “Viral Marketing: The Science of Sharing” (Victoria, Australia: Oxford University Press Australia & New Zealand, 2013).
14. D. Hurwitz, “Watch: YouTube’s Most Popular Videos of 2012,” Dec. 21, 2012, www.usatoday.com.
15. J.L. Hayes and K.W. King, “The Social Exchange of Viral Ads: Referral and Coreferral of Ads Among College Students,” Journal of Interactive Advertising 14, no. 2 (2014): 98-109; and J. Huang, S. Su, L. Zhou and X. Liu, “Attitude Toward the Viral Ad: Expanding Traditional Advertising Models to Interactive Advertising,” Journal of Interactive Marketing 27, no. 1 (February 2013): 36-46.
16. R.E. Guadagno, D.M. Rempala, S. Murphy and B.M. Okdie, “What Makes a Video Go Viral? An Analysis of Emotional Contagion and Internet Memes,” Computers in Human Behavior 29, no. 6 (November 2013): 2312-2319.
17. E. Botha, “A Means to an End: Using Political Satire to Go Viral,” Public Relations Review 40, no. 2 (June 2014): 363-374; and D. Southgate, N. Westoby and G. Page, “Creative Determinants of Viral Video Viewing,” International Journal of Advertising 29, no. 3 (2010): 349-368.
18. Botha provides no definition of creativity while Southgate, Westoby and Page define creativity in terms of involving and enjoyable (branded) content without explaining what these terms mean and how they differ from, for example, outcome measures of engagement. For example, one might argue watching a video is a measure of involvement, not an indicator of creativity.
19. AprilsAnimals, “Tiny Hamster Eating a Tiny Pizza,” May 7, 2014, www.youtube.com.
20. P. Eckler and P. Bolls, “Spreading the Virus: Emotional Tone of Viral Advertising and Its Effect on Forwarding Intentions and Attitudes,” Journal of Interactive Advertising 11, no. 2 (spring 2011): 72-83; Guadagno et al., “What Makes a Video Go Viral?”; and K. Nelson-Field, E. Riebe and K. Newstead, “The Emotions that Drive Viral Videos,” Australasian Marketing Journal 21, no 4 (November 2013): 205-211.
21. The results do, however, suggest some interesting research questions. For example, Nelson-Field, Riebe and Newstead suggest that sad videos may be shared more, whereas our data suggests sad videos are watched less. Between the two, there is a logical disconnect that bears examination. For example, could it be that if people knowingly receive a link to a sad video, most prefer not to watch it in order to maintain their mood?
i. The number of views does not reflect differences in viewership composition or social transmission (in other words, sharing). For example, it does not distinguish whether one person has viewed a video many times or whether many people have viewed a video only once.