While on vacation earlier this month, I read Nate Silver’s book, The Signal and the Noise: Why So Many Predictions Fail – But Some Don’t. Silver is the statistician who developed PECOTA, a system for forecasting baseball performance that was bought by Baseball Prospectus. He is also the writer and founder of The New York Times political blog FiveThirtyEight, who accurately predicted the outcomes of the U.S. elections in 2008 and 2012.
Although Silver’s book doesn’t make any predictions about where online video is headed, he does say, “Companies that really get Big Data, like Google, aren’t spending a lot of time in model land. They’re running thousands of experiments every year and testing their ideas on real customers."
That’s an insight we can start using next week.
Silver is also a proponent of Bayesian statistics. He observes that the most accurate forecasters tend to have a superior command of probability. They can distinguish the predictable from the unpredictable, and they notice a thousand little details that lead them closer to the truth. That’s why they can also distinguish the signal from the noise.
We can start using this insight today.
For example, Chris Atkinson noted late last week that “some of the overall numbers are down” in comScore’s February 2013 U.S. Online Video Rankings. In February 2013, 178 million Americans watched 33 billion online content videos for 17.4 hours per viewer. In February 2012, 179 million U.S. Internet users watched nearly 38 billion videos of online video content for 21.8 hours per viewer. And in February 2011, 170 million U.S. Internet users watched online video content in February for an average of 13.6 hours per viewer.
So, is the typical American viewer now spending 4.4 fewer hours a month watching 5 billion fewer videos than he or she was a year ago? Was 2012 the high water mark for online video?
Before we panic and jump to conclusions, let’s double-check the data and our assumptions to try to distinguish the signal from the noise.
It turns out that comScore made three important methodological changes to its estimates with the release of the August 2012 U.S. Video Metrix data. So, comparing this year’s data to last year’s data is like comparing apples to oranges.
Fortunately, I was able to reach one of comScore’s PR people over the weekend, who put me in touch with Andrew Lipsman, comScore’s VP of Industry Analysis. He was more than happy to share the latest apples-to-apples trends for online video across a variety of metrics.
As for the dip in February, he pointed out that if you multiply January video totals by 28/31 to seasonally adjust the data, you actually get a number slightly lower than February’s total. As for the higher figures in the last five months of 2012, that could also be impacted by seasonality. Viewership of online video tends to go up with the start of a new TV season.
|Total Unique Viewers (000)||188,016||181,411||182,574||182,050||181,717||179,515||177,955|
|Videos per Viewer||200.5||216.9||204.0||219.5||212.8||201.6||185.6|
|Minutes per Viewer||1,335.4||1,399.0||1,254.4||1,182.9||1,150.2||1,140.2||1,045.7|
Nevertheless, my interest in seeing what has changed over the past two years – particularly at YouTube.com, the top U.S. online video content property ranked by unique video viewers and the top U.S. online video ad property ranked by video ads viewed – prompted me to seek out another source of information to see where online video might be headed.
For example, check out the first chart below from Compete PRO. It looks at the unique visitors to YouTube.com over the past two years. It shows a slow but steady increase from 140.5 million unique visitors to YouTube.com in February 2011 to 147.1 million in February 2012 to 163.2 million in February 2013. There’s no drama in this data because the slight dip in February 2013 from January 2013 is similar to the slight dip in February 2012 from January 2012. Hey, February is the shortest month of the year.
Next, check out the second chart from Compete PRO. It looks at visits to YouTube.com over the past two years. It shows a bumpier road than the off-road jeep tour I took two weeks ago while on vacation in Aruba. There were 1.5 billion visits in February 2011, 1.3 billion in February 2012, and 1.4 billion in February 2013. Fasten your seatbelts; it’s going to be a bumpy night!
Then, check out the third chart from Compete PRO. It looks at the page views on YouTube.com over the past two years. It shows a roller coaster ride that drops from 20.4 billion page views in February 2011 to 14.5 billion in February 2012 to 12.7 billion in February 2013. Holy moly, what happened?
Then, check out the fourth chart from Compete PRO. It looks at the average stay on YouTube.com over the past two years. It shows a bumpy downhill ride that drops from an average stay of 22.1 minutes in February 2011 to 19.2 minutes in February 2012 to 18.5 minutes in February 2013. Seasonality, my foot!
Finally, check out the fifth chart from Compete PRO. It looks at the visits per person and the pages per visit on YouTube.com over the past two years. It shows two terrible, horrible, no good, very bad trends. First, visits per person drop from 10.6 in February 2011 to 8.9 in February 2012 to 8.4 in February 2013. Second, pages per visit drop from 13.7 in February 2011 to 11.1 in February 2012 to 9.3 in February 2013. San Bruno, we have a problem!
Looking at these last three charts can help us to distinguish the warning signal from the seasonal noise. But they still don’t explain what is driving these downward trends. Here are some probable causes:
- Psychographics: Four months ago, we took a look at Generation V (Gen V), a psychographic profile that cuts across demographic groups. Men 18-34, particularly young men 18-24, and Women 25-49, particularly women 25-34, have strong Gen V behaviors. For example, 40 percent of Men 18-34 and 25 percent of Women 25-34 actively seek out videos related to their particular passions or hobbies. And 64 percent of Men 18-24, 57 percent of Men 18-34, 46 percent of Women 25-24, and 38 percent of Women 25-49 have shared online video content in the past week. However, as a larger percentage of Americans visit YouTube each month, has there been a regression toward the mean? In other words, does a smaller percentage of the site’s audience now have strong Gen V behaviors?
- Content: In February 2011, over 35 hours of video were being uploaded to YouTube every minute. Today, 72 hours of video are being uploaded to YouTube every minute. However, as more and more content is being created every day, has there been a reversion to mediocrity? In other words, is the typical video now less unique, compelling, entertaining or informative?
- Mobile: Today, 25 percent of global YouTube views come from mobile devices. Are people watching video on phones and tablets differently than they watch on desktops and laptops? In other words, has the growing percentage of views come from mobile devices caused the drop in page views, average stay, visits per person, and pages per visit on YouTube.com?
- Seasonality: There appears to have been a short upward bounce at the end of 2012 in visits, pages views, average stay, visits per person, and pages per visit on YouTube.com. So, is YouTube impacted by seasonality just as the broader online video market is? In other words, did viewership of TV Shows
- Redesign: During the past two years, quite a bit has changed on YouTube. Did these changes cause the drop in page views, average stay, visits per person, and pages per visit – or did these downward trends cause YouTube to increase its focus on watch-time and audience engagement as well as begin the “channel-ization” of the platform? In other words, which is the cause and which is the effect?
I don’t have the answers to these questions … yet. But I now understand why it’s so important to be able to distinguish the signal from the noise in order to see whether there will be any significant shifts in where online video is headed.
Over the past eight years, YouTube has vanquished Singingfish, Yahoo! Video, Google Video, and a host of other online video sites. But, past performance is not necessarily indicative of future results. And going forward, YouTube will face new competition (predictable) from unexpected challengers (unpredictable).
For example, on Jan. 24, 2013, Twitter launched Vine, a mobile app that enables its users to create short video clips with a maximum length of six seconds that can be shared on a variety of social networking services, such as Twitter or Facebook. As of yesterday, Vine ranked #10 in the iTunes charts for free apps, behind YouTube at #8, but ahead of Instagram at #14.
Last year, 500 years of YouTube video was being watched every day on Facebook and over 700 YouTube videos were being shared on Twitter each minute. Following the launch of Vine, was more – or less – YouTube video being watched on Facebook.com and shared on Twitter in February?
So, keep your eyes open and check your peripheral vision from time to time. Hey, if things never changed, then we’d all still be optimizing the metadata of our videos for Singingfish.