Data: Why YouTube Will Never Look Quite Like TV

Even as YouTube pursues television-quality content, data is helping mold decisions about everything from hip-hop dance moves to throat-slitting horror gore.

Data: Why YouTube Will Never Look Quite Like TV

“This is the Steve Martin,” says Charlene “Chi-Chi” Smith, swinging her arms as she demonstrates the dance move: two exaggerated hip-hop steps to the left and another two to the right.


Smith, dressed in jeggings with black converse sneakers laced up to her knees, is Shakira’s former dance captain and has performed with the likes of Snoop Dogg and Diddy. Online instructional video creator Howcast recently recruited her to teach the rest of us how to dance.

When it’s time to demonstrate the Steve Martin to music, Howcast’s senior director of production, Heather Menicucci, hits play on her open laptop. Fifteen seconds later, that’s a wrap. Smith has flawlessly improvised a professional-quality dance lesson without so much as stumbling on a word. Thirty minutes later, she’s also given one-take demonstrations of the Cat Daddy, the Robo Cop, and the Dougie.

Smith is an awesome dance teacher, but the reason she’s here today has nothing to do with hip-hop. It’s about data analytics.

Hip-hop is a topic plucked from Howcast’s topic-selecting program, which selects winning video search terms based on their search volume, search competition, advertising yield, public competitor data, and the past performance of similar themes. The company is filming its fifth series on hip-hop and teaching us how to Dougie for the third time because both topics had a high “z score,” or likeliness of profitability.

In the realm of Internet video, Howcast is as good at data as Smith is at the Dougie. It makes almost every programming decision–from how much content it releases to whether a video should include an intro–based on the numbers.

Not all channels take their data analysis to the same extreme, but almost all of them base at least some creative decisions on their data dashboards. Even as YouTube pushes for high-quality, TV-like content, data is molding Internet video into a distinct medium.


“The way I think about [web video] is how I think about web 2.0,” says Tim Shey, director of the YouTube Next Lab. “The web 2.0 companies don’t spend years building their platform and then launching it and crossing their fingers…you launch with the minimum feature set you need to keep your audience happy. And then you iterate and improve.”

Shey is charged with coaching YouTube’s more than 1 million ad-serving channels to maximize their performance. As of late, that job has involved more data tools. YouTube continually refines its content creator tool set. In November, it launched a revamped data dashboard, including a new option to see how long viewers watch a particular video. In March, it announced new tools to analyzing total watch time on videos.

This sort of minute-by-minute feedback provides cues that Nielsen television ratings never could. And paired with online video’s relative low barrier to entry, it enables experimentation that would be inefficient or impossible to interpret in a traditional broadcast.

Before YouTube acquired his company, Next New Networks, in 2011, Shey led a handful of YouTube channels. He credits attentiveness to data with his success on the platform. For instance, one of his most successful channels, called Barelypolitical, originally targeted an older audience that actually followed politics. But when it introduced Obama Girl, with her tight shirt and gushing songs about the 2008 election’s democratic candidate, it realized the channel had untapped potential.

Not only were Obama Girl’s music videos a hit, but other musical comedy about pop culture on the channel was gaining steam. Data showed the channel’s audience was also getting younger. Reacting to these factors, Shey and his team decided to create a weekly pop culture musical comedy segment called “The Key of Awesome” and aimed it at a teenage audience. It ended up blowing past their previous successes to the tune of more than 1 billion views.

Barelypolitical accidentally discovered a hit format it may have not noticed without analytics. Howcast’s data-driven experiments are more deliberate.


Smith has planned to create 40 videos on the day that I visit the Howcast studio. In a handful of them, she’ll teach dancers in the room with her instead of demonstrating solo. Again, this decision has nothing to do with hip-hop. Howcast wants to know if its audience responds better to a solo teacher talking to them directly, or to a teacher and student combo. So it’s filming samples of each method to use as a test.

The company has run about 35 similar tests since it started in 2007, according to Rick Bashkoff, Howcast’s VP of Business Development & Marketing. At one point, for instance, it added short expert introductions to each of its videos, but a look at the drop-off data showed the 10-second clips were sending users clicking elsewhere. Now Howcast videos jump right into the action.

CafeMom Studios, a channel that hosts mostly talk shows, noticed a similar slump when it verbally transitioned from story to story during its daily news show. It, too, decided to ditch its intros.

“As a result, we have seen audience retention smooth out–we no longer lose viewers between stories–and more are now sticking around for the whole video,” CafeMom EVP of Content Tracy Odell tells Fast Company.

Sticking around for the whole video is something YouTube is pushing for. In addition to investing heavily in high-quality content, it has also begun rewarding videos that retain viewers, not just attract them, by favoring them in search and suggested videos. As a result, Shey says that time spent on YouTube has been rising. Viewers watch 4 billion hours of video on the site every month.

Shey says data can help creators make content that keeps people watching their videos.


“Especially if making long-form content, you can really understand at what point people rewind and watch parts of the video again and again, they fast forward, they drop off. You can really look at videos and see how far they’re watching and then try to understand why,” he says.

Bashkoff agrees. Howcast sticks to a short how-to format, but he’s thought about how data might also shape episodic, long-form content. Do people navigate elsewhere when that guy’s throat is slit? Now you know where your optimal gore line sits. Is a character more engaging when he’s fat and bald or handsome and fit?

Howcast’s co-chairman, Kevin Law, has a unique vantage point from which to approach the question. He’s also the CEO of Uncommon Content Partners, which makes shows of television length and quality for YouTube. According to him, the data-based decision making of Howcast doesn’t translate well to the entertainment-focused content of Uncommon’s Reserve Channel.

It would be too expensive, he says, to do the sort of experiments Howcast uses to program its instructional videos on the long-form shows that populate Uncommon’s YouTube channel. Data isn’t nearly as useful when making programming decisions, either.

“You make a video about how to fix a flat tire because you are addressing an individual need of theirs,” he says. “It’s not that the person who fixes a flat tire wants to see a show about cars.”

“When you’re entertaining somebody, it’s more subjective. You’re trying to anticipate what is going to make somebody laugh or cry or be inspired.”


Uncommon’s creative process may be all art, but it can’t escape science. Just like Howcast–which gets between 30% and 40% of its traffic from search and most of the rest from suggested videos–the majority of the Reserve Channel’s viewers find its shows through some type of query. Most people do not yet commit to watching any YouTube series every week the same way they do television shows.

Data drives how each of Uncommon’s shows is paced. It changes how they’re promoted, and it will ultimately decide which shows get dropped and which shows get their own channels.

“It’s an interesting combination of art and science,” Law says. “I don’t think it’s ever really existed to the extent it does now, for video programming, because of the web.”

[Image: Flickr user Michael Jones]


About the author

Sarah Kessler is a senior writer at Fast Company, where she writes about the on-demand/gig/sharing "economies" and the future of work.