I'm a partner in the advanced analytics group at Bain & Company, the global management consulting firm. My primary focus is on marketing analytics (bio). I've been writing here (views my own) about marketing, technology, e-business, and analytics since 2003 (blog name explained).

Email or follow me:


« Media as Software: A Conversation With Doug Turner | Main | Video Search Part II: ScanScout »

April 29, 2007

Think Viral, Act Tribal Part III: "Dissecting Numa Numa"

Last Thursday morning I attended a MITX Digital Marketing Series presentation titled "Dissecting Numa Numa: A Critical Analysis of Viral Video Content", given by Jeremi Karnell of One to One Interactive, Professor Jeffrey Bardzell of Indiana University's School of Informatics, and Dr. Carl Marci, Chief Science Officer at Innerscope Research.  The questions considered (my version):  why did this amateur work go as "viral" as it did, and how (well) can neuroscience help us predict viral media propagation?

Gary Brolsma first posted his "Numa Numa" video to the Newgrounds Flash animation community in December 2004.  To date, it's the second-most-downloaded video, with 700 million views across all video sharing sites, according to a BBC study Jeremi cited ("Star Wars Kid" is #1 at 900 million). It's been estimated that this equates to roughly $18 million dollars in free media.  15,000 mashups have been made of the original; the top 3 have attracted 13 million views on YouTube alone.  Gary's MySpace page lists 8,600 "friends".

Yet after its release Gary was mocked by mainstream media that only examined his work superficially, and so he dropped off the pop culture radar screen for a while before coming back to try to exploit his fame, with only modest success to date -- http://newnuma.com gets about 9,000 visits per day, according to Compete research cited by Jeremi.

The crucial success factor for Gary's original effort (as I absorbed it) was its significant degree of slyly humorous "intertextuality".  This is to say that it flowed directly from an amalgam of content and themes from communities which Gary participated in actively, if previously obscurely.  If the first 10,000 views of any candidates for viral glory are the hardest, Gary's riffs on those pre-existing memes were sufficiently clever to get the Newgrounds community talking intensively.  Or, put another way, Numa Numa went "macroviral" partly because it went "microviral" first. (See also "Experimental Study of Inequality and Unpredictability in an Artificial Cultural Market", Science, February 10 2006. Summarized in "The Science of Hit Songs" .)

Prof. Bardzell, who has a PhD in comparative literature, suggested "Numa Numa's" humor can only be understood in the context of the shared aesthetics of the Newgrounds community where he launched it.  These  included:

  • the aesthetic of "crap"
  • the aesthetic of "OMG/WTF"
  • the aesthetic of "disorientation"
  • the aesthetic of "virtual tagging/ mastery"

The first refers to an appreciation for content so bad and schlocky that it's campy and funny.  The second refers to valuing work that generates surprise and puzzlement in its audience.  The third refers to presenting content in a non-sensical way: googly eyes, animated figures spinning and bouncing around.  Finally, the fourth is about introducing icons (penguins, bears, raw steaks) that are icons to insiders in the community, and are presented so quickly that for some it takes several viewings, often requiring viewers to pause the video to catch a reference in a frame, to "get it".

Of course it's impossible to talk about the provenance of the video without talking about the roots of  the music.  The Numa Numa song's original name is "Dragostea Din Tei", which has something to do with love and linden trees.  With lyrics in Romanian, it was originally released in March 2004 by a Moldovan boy band called O-Zone, but it didn't go anywhere at first.  Shortly afterward it was covered by a Romanian singer living in Italy, who made it a minor hit.  Both these efforts adhere to heavily stylized pop video aesthetics that feature parody and emphasize exaggerated physical gestures, and which often mix animation with "real" footage as well -- elements which helped it bridge into Japan's anime-inspired Flash media scene.  (I can't leave out what appears to be an evolutionary spur of the Numa Numa phenomenon Prof. Bardzell showed -- a cover by the German goth metal band Audiosmog, which broke up shortly after this effort.)

By the fall of 2004 it appears Numa Numa was making the leap from Japan to the US.  Several clues  from Brolsma's first Numa Numa video make it clear that he was mashing up / riffing on stuff he had found before, rather than creating a totally random work.  He uses the exact clip of the original "DDT" song previously used in a popular Flash animation posted on Newgrounds.  He dances and pumps his arms in a non-narrative, "happy-happy" manner common to a part of the Newgrounds world that focuses on crude animations using static, bitmap images.  He pays homage to a subgenre of animation called animutation, invented originally by a 13-year-old named Neil Cicierega (http://albinoblacksheep.com), which seems to take its inspiration from Monty Python animations.   He weaves in icons (images of penguins, arched eyebrows on pictures of people) that are "tags" for insiders.  And he uses another community convention: mondegreens.

(Literate friends note: I admit I first learned this term Thursday.)   Apparently it's a derivative from an old poem with the phrase "and they laid him on the green."  Apparently in modern use (friends who know more please help here) it refers to words that sound like others, and the comparison of which is humorous in a silly way.  Hence when the raw steak flashes on the screen in Numa Numa, it's a reference to the "beef sounds like a cellphone's 'beep'" mondegreen, as in -- as Gary says at another point in the video -- "I 'beep' you!"  This, and the otherwise apparently non-sensical photo-based reference to feta cheese, are exactly the kind of deeply inside jokes that drove so much interest among fellow Newgrounds participants.

I've left lots out.  Prof. Bardzell presented, among other things, a "network diagram of intertextual relationships" that showed relationships between sites and content across time and continents, that looked like angel hair spaghetti draped across a Gantt chart.

Was this all a happy accident, or was Gary really this calculating?  The general conclusion was that Gary was both really hip to the communities he participated in (he had posted other material on Newgrounds 12 times before), and really clever with his riff on what he found there.  There's no evidence that he had a bigger framework or agenda in mind.  He was also original in at least one important respect.  Apparently it's extremely unusual to post videos on Newgrounds, and parodying what came before using this format was an especially happy "strangeness" in the Harold Bloom sense.

So to sum up, going viral here depended on:

  • one-upsmanship
  • doing stuff that seemed initially crazy
  • but with more views came understanding
  • with that understanding -- getting the jokes -- came a sense of being "in on it" -- which then triggers viral propagation to others who might also be in a position to "get it" -- almost as a challenge
  • the more communities you're in (among the ones that are relevant to the item being sent around) the greater the likelihood you'll get it -- and share it across communities
  • the pre-existing susceptibility of the "host" to the "virus"

Next, Dr. Marci talked about "predicting viral video engagement with neuroscience", presenting research done by his firm Innerscope that One to One Interactive had sponsored.  His first cautionary words: "if you're not thinking in an interdisciplinary way, you're not thinking!"  Hence, the emergence of neuromarketing, a field now made possible by better / cheaper signal processing, faster/ cheaper computing, sorage, and bandwidth, and the accumulation of two decades of brain research to inform what we see and crunch.

Innerscope has developed a metric it calls a "biological index of engagement", or more specifically a measure that represents the "aggregated response of defined [reactions] at a pre-conscious level, timelocked to video.  These reactions, measured with a "smart vest" subjects wear, include:

  • arousal (brain waves?  skin's electrical conductivity or temp? can't remember...)
  • respiratory state
  • heart rate
  • motion (measured via accelerometers in the vest)
  • eye-tracking
  • (click-tracking -- not relevant here)

So, they show  focus group subjects a bunch of these videos and then correlate their "mean audience engagement" results with independent metrics of video popularity, like views, and reviews, and ratings on the video sites where the videos are found.  R-squared = 0.77, to a statistically significant level!

Dr. Marci noted that most brain processing -- 75-95% -- is unconscious, and that there are no direct connections between the parts of the brain that control emotion and those that control language.  So to ask someone "how are you feeling?" inevitably provides a highly filtered, and thus only partially helpful response.  In research circles this is called the "cognitive bias in emotional self-report."  So, measuring biological response is really much better, as the stats bear out.

In Innerscope's M.A.E. measure, 50% indicates neutral engagement with a video -- "ok, I'm watching, not engaged, not distracted."  70% is high.   Numa Numa peaks early at 80%, then holds mostly at 65%, dropping below 60% only briefly.  There's a typical pattern Dr. Marci reports in which an early peak is followed by what he called "an attention release pattern" as audiences get past the initial effort to figure out what the video's all about and settle into the story.  Highly rated/ high overall M.A.E. videos dip down for range as part of taking the audience on an emotional journey (Numa Numa flashes still images of people frowning about 2/3 to 3/4 of the way in), but they end high to leave their audiences with closure and feeling happy, as Numa Numa does. 

He notes that while they observe this strong correlation between videos'  M.A.E. and popularity right now, it's early for causality.  Dr. Marci noted he has not been able to do controlled experiments here that isolate the impact of the images from the music yet.  It's also unclear to what extent pre-selection for length (we're more likely to view -- and rate, and therefore rate highly -- shorter clips) plays a role, since he pre-selected the videos in this research sample.  One observation nonetheless: pictures of food correlate with higher M.A.E.

Other comments from the speakers:

  • "The agency of success here was the predisposition of the pre-existing social network, not the individual genius of the work"
  • "Viral content has a certain Shakespearean 'hi-lo' appeal:  references to Greek myths in one breath, crude jokes in the next"
  • "Success factors for amateur videos with no prior public awareness clearly differ from those for professionally produced videos where a brand may have significant pre-existing recall among audiences -- contrast Numa Numa with  Kylie Minogue's more highly-produced but ultimately less-sophisticated 'Agent Provocateur' , which is the fourth most-downloaded video ever."

My take:

The idea that viral success depends on triggering the pre-dispositions of the pre-existing network resonated most strongly for me, given past experiences and ideas.  The idea of working humor into content to help its viral strength is one that I hadn't focused on as much, but as I think about it (Diet Coke Mentos Experiment, Smirnoff Rap) seems to make lots of sense -- maybe something I can work into my posts ;-)  Here, totally gratuitously to see if I can boost the viral fortunes of this post, is one of my favorite cartoons about blogs, Pat Oliphant's "At The Gates":


Related: Reflections from Bob Metcalfe on adjustments to Metcalfe's Law in the context of social networks, here on his partner Mike Hirshland's blog.

Postscript: "There Will Never Be Another Star Wars Kid"


TrackBack URL for this entry:

Listed below are links to weblogs that reference Think Viral, Act Tribal Part III: "Dissecting Numa Numa":

» Dissecting Numa Numa | A Critical Analysis of Viral Video Content from OTOinsights
Last week I had the pleasure of presenting along side with Dr. Jeffery Bardzel (PH.D. Assistant Professor of HCI/Design and new media at the School of Informatics in Indiana University) and Dr. Carl Marci (Chief Science officer Innerscope Research) One [Read More]


You say:

The idea that viral success depends on triggering the pre-dispositions of the pre-existing network resonated most strongly for me...

so, you mean that if a video clicks with a certain demographic it will be viral?

Robin, Thanks for your question. I inferred from Prof Bardzell's comments that Numa Numa exploits the pre-existing habit in the Newgrounds community to try to one-up the next guy with parodies of pre-existing content that use lots of inside jokes, and that people there share stuff that does this well. Best, Cesar

I like your analysis of how this phenomenon occurred. I hope that you will enjoy my post celebrating the continued success of this viral phenomenon: http://whoaitsaimz.blogspot.com/2008/12/happy-4th-birthday-numa-numa-dance.html

The comments to this entry are closed.