The TV industry's lost generation.

Watching television is still the number one leisure activity for the majority of people in the developed world and, as I’ve explained before, measuring who, what, when and how individuals view TV is extremely important.  This data is used to shape the kind of programmes that are created in order to appeal to advertisers - the all-important advertisers whose business depends on knowing that the right kind of people are viewing their TV advertisements.  Put simply, if you’re an advertiser and you pay to show car ads to people who don’t buy cars, or who have no influence on people who do buy cars, then you have wasted your money. Nielsen, the dominant company which measures these essential TV audiences, faces a major problem - the Internet has radically changed television viewing habits and this change has completely disabled Nielsen’s measurement system.   Nowadays people under 30 years old are far more likely to watch TV on a laptop computer, a tablet, or a mobile phone rather than a TV screen, and Nielsen’s measurement system isn’t flexible enough to work at all well with Internet devices.  The two pie charts above illustrate the size of the problem.  Nielsen’s measurement system was created for a time when the only way to watch TV was on a big screen in the living room, or perhaps on a second TV in the bedroom.  This was the era of appointment-only TV, when programmes were transmitted at a particular time over the air, or via a cable connection, to the dominant TV screen.  Like so many of the market research business reports on Internet behaviour that I see, Nielsen’s methodology has two basic flaws – “the law of small numbers” and “time.”  For the pie chart on the left I’ve used U.S. data from Nielsen’s latest State of the Media report that apparently shows the relative percentage market share of monthly time spent watching streaming video from YouTube, Netflix and Hulu.  For the pie chart on the right I’ve used North American data from the recent Global Internet Phenomena Report from Sandvine to calculate the relative percentage market share for the same three companies.  From the outset I should point out that there is a difference in the two datasets between a North America that includes Canada and one that doesn’t.  In an ideal comparison both datasets would be solely U.S. based, or both would contain North America, although the inclusion of Canada in one dataset is more likely to dampen down the difference in results rather than create an emphasis.  If you work in the U.S. television industry and believe Nielsen’s data, then YouTube is foremost and Netflix and Hulu are not that significant.  But look at the Sandvine data and Netflix is now the clear leader with nearly twice the usage compared to YouTube.  Why such a large difference in two results? For many years now, in order to understand and measure how people watch TV in the U.S., Nielsen has used a sample population of a putative 10,000 - 25,000 homes (the actual number is opaque, try finding it on Nielsen’s website) equipped with electronic devices that monitor which channel the TV set is receiving, and at what time it is being used.  This is a comparatively tiny sample that has been calculated as just 0.02% of the total U.S. TV audience.  So in order to get richer audience data a larger sample, using paper diaries, is also captured.  This is where people self-report what and when they have watched TV.  You can predict the problem: along came the Internet like Topsy: growing and growing until everything changed.  Eventually in 2007 Nielsen responded by purchasing NetRatings, a web-tracking company.  That may have been fine for a time but some of the key websites on which people watch TV programs online do not implement NetRatings SiteCensus tracking code.  Of the three major video websites shown in my charts, NetRatings only claim Hulu as a client, and I haven’t detected a SiteCensus tracker on the other two websites.  This means that Nielsen doesn’t have any really large, and therefore more accurate, sample data about YouTube or Netflix viewing.  They are using a small sample of people and extrapolating this data to represent the entire U.S. online TV audience.  The end result of this modelling is that the information is totally misleading due to the fact that small samples will produce apparently significant effects which, when using bigger samples, will be averaged out and not appear as important.  This is known as the law of small numbers – for a good explanation of this see Chapter 10 in Daniel Kahneman’s best seller Thinking, Fast and Slow.  Worse still for the TV industry, relying on such small samples means that significant effects may go undetected. The Sandvine data that I’ve used for the pie chart on the right is derived in a totally different way.  This is anonymous behavioural data captured from U.S. Internet Service Providers (ISPs) which use Sandvine’s network monitoring software.  This is definitely not a small sample as Sandvine’s software is utilised by a significant proportion of U.S. network providers, including big players like Comcast and Clearwire.  These companies use Sandvine’s software to understand, shape, control and, of course, charge subscribers for the data flowing through their networks.  Obviously this much larger data sample provides a far more accurate understanding of American online TV behaviour, especially as the information isn’t self-reported – self-reporting is notoriously unreliable.  On the contrary, Sandvine’s data is collected and measured in real time as people actually access the three websites in the chart.  It is obvious that measuring large samples of people’s actual behaviour will always be more accurate than smaller samples that have been self-reported.  To compound its errors Nielsen appears to be interested only in online TV viewing that happens within 7 days of a show being aired on a network channel, despite the fact that in a digital environment viewing happens over much longer timescales.  Sandvine’s data more practically covers a 30 day period as well as including viewing on laptop computers, tablets, or mobile phones. Television watching behaviour slowly changed from the very first public use of the Internet in 1994, through to escalating differences as adequate broadband speeds have made viewing online TV more pleasurable.  Yet, extraordinarily, Nielsen’s audience data has been consistently under- reporting any form of online “binge” TV viewing.  This is where people use services like Amazon and Netflix to view all episodes of a TV series over the course of one or more evenings.  I first came across this behaviour six years ago when interviewing a busy senior executive about TV viewing behaviour.  Both people in this household led highly pressured lives and when they watched television they wanted “good TV”, with no commercial breaks, and all the episodes of a program available at once at a time to suit them.  This is exactly what Amazon and Netflix supply with their online streaming services, but this important aspect of modern TV viewing is simply not covered in Nielsen audience data.  Amazon and Netflix shrewdly keep such valuable information to themselves but Amazon knows exactly what and when programmes are being viewed using Netflix, as Amazon’s cloud services provide Netflix’s video streaming.  Unsurprisingly Amazon and Netflix are now using this data to create TV productions based on the genuine popularity of specific types of programs. For me, the interesting aspect of all this is how two very large industries - TV and Advertising - have continued to mislead themselves for so long.  Both industries are still persisting in making fundamental flaws about defining who watches which programs and when, never mind which devices are being used for viewing, and this despite the truly enormous sums of money which are involved.  Heads in the sand comes to mind.  The era of “big data” may now be revealing how totally inaccurate Nielsen’s numbers actually are, and have been for some time.  The question to ask is just who or what benefits from trying to maintain the status quo?  It’s certainly time now for the incumbent TV industry to get TV viewing in proper focus before inadequate measurement, and subsequent poor feedback, eventually leads to the loss of a whole generation of younger viewers. June 2013
Click here to download the PowerPoint chart: Click here to download the PowerPoint chart:
...with analysis & insight...
Archive: Free PowerPoint download Free PowerPoint download
Click image to enlarge
Click here to download the PowerPoint chart: Click here to download the PowerPoint chart: Click here to download the PowerPoint chart: Click here to download the PowerPoint chart: Click here to download the PowerPoint chart: Click here to download the PowerPoint chart:
How Netflix and Amazon are changing TV viewing
Click to return to page
View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles View All 2016 Articles
2013
Click to return to page

The TV industry's lost

generation.

How Netflix and Amazon are changing TV viewing
Watching television is still the number one leisure activity for the majority of people in the developed world and, as I’ve explained before, measuring who, what, when and how individuals view TV is extremely important.  This data is used to shape the kind of programmes that are created in order to appeal to advertisers - the all-important advertisers whose business depends on knowing that the right kind of people are viewing their TV advertisements.  Put simply, if you’re an advertiser and you pay to show car ads to people who don’t buy cars, or who have no influence on people who do buy cars, then you have wasted your money. Nielsen, the dominant company which measures these essential TV audiences, faces a major problem - the Internet has radically changed television viewing habits and this change has completely disabled Nielsen’s measurement system.   Nowadays people under 30 years old are far more likely to watch TV on a laptop computer, a tablet, or a mobile phone rather than a TV screen, and Nielsen’s measurement system isn’t flexible enough to work at all well with Internet devices.  The two pie charts above illustrate the size of the problem.  Nielsen’s measurement system was created for a time when the only way to watch TV was on a big screen in the living room, or perhaps on a second TV in the bedroom.  This was the era of appointment-only TV, when programmes were transmitted at a particular time over the air, or via a cable connection, to the dominant TV screen.  Like so many of the market research business reports on Internet behaviour that I see, Nielsen’s methodology has two basic flaws – “the law of small numbers” and “time.”  For the pie chart on the left I’ve used U.S. data from Nielsen’s latest State of the Media report that apparently shows the relative percentage market share of monthly time spent watching streaming video from YouTube, Netflix and Hulu.  For the pie chart on the right I’ve used North American data from the recent Global Internet Phenomena Report from Sandvine to calculate the relative percentage market share for the same three companies.  From the outset I should point out that there is a difference in the two datasets between a North America that includes Canada and one that doesn’t.  In an ideal comparison both datasets would be solely U.S. based, or both would contain North America, although the inclusion of Canada in one dataset is more likely to dampen down the difference in results rather than create an emphasis.  If you work in the U.S. television industry and believe Nielsen’s data, then YouTube is foremost and Netflix and Hulu are not that significant.  But look at the Sandvine data and Netflix is now the clear leader with nearly twice the usage compared to YouTube.  Why such a large difference in two results? For many years now, in order to understand and measure how people watch TV in the U.S., Nielsen has used a sample population of a putative 10,000 - 25,000 homes (the actual number is opaque, try finding it on Nielsen’s website) equipped with electronic devices that monitor which channel the TV set is receiving, and at what time it is being used.  This is a comparatively tiny sample that has been calculated as just 0.02% of the total U.S. TV audience.  So in order to get richer audience data a larger sample, using paper diaries, is also captured.  This is where people self-report what and when they have watched TV.  You can predict the problem: along came the Internet like Topsy: growing and growing until everything changed.  Eventually in 2007 Nielsen responded by purchasing NetRatings, a web-tracking company.  That may have been fine for a time but some of the key websites on which people watch TV programs online do not implement NetRatings SiteCensus tracking code.  Of the three major video websites shown in my charts, NetRatings only claim Hulu as a client, and I haven’t detected a SiteCensus tracker on the other two websites.  This means that Nielsen doesn’t have any really large, and therefore more accurate, sample data about YouTube or Netflix viewing.  They are using a small sample of people and extrapolating this data to represent the entire U.S. online TV audience.  The end result of this modelling is that the information is totally misleading due to the fact that small samples will produce apparently significant effects which, when using bigger samples, will be averaged out and not appear as important.  This is known as the law of small numbers – for a good explanation of this see Chapter 10 in Daniel Kahneman’s best seller Thinking, Fast and Slow.  Worse still for the TV industry, relying on such small samples means that significant effects may go undetected. The Sandvine data that I’ve used for the pie chart on the right is derived in a totally different way.  This is anonymous behavioural data captured from U.S. Internet Service Providers (ISPs) which use Sandvine’s network monitoring software.  This is definitely not a small sample as Sandvine’s software is utilised by a significant proportion of U.S. network providers, including big players like Comcast and Clearwire.  These companies use Sandvine’s software to understand, shape, control and, of course, charge subscribers for the data flowing through their networks.  Obviously this much larger data sample provides a far more accurate understanding of American online TV behaviour, especially as the information isn’t self-reported – self-reporting is notoriously unreliable.  On the contrary, Sandvine’s data is collected and measured in real time as people actually access the three websites in the chart.  It is obvious that measuring large samples of people’s actual behaviour will always be more accurate than smaller samples that have been self- reported.  To compound its errors Nielsen appears to be interested only in online TV viewing that happens within 7 days of a show being aired on a network channel, despite the fact that in a digital environment viewing happens over much longer timescales.  Sandvine’s data more practically covers a 30 day period as well as including viewing on laptop computers, tablets, or mobile phones. Television watching behaviour slowly changed from the very first public use of the Internet in 1994, through to escalating differences as adequate broadband speeds have made viewing online TV more pleasurable.  Yet, extraordinarily, Nielsen’s audience data has been consistently under-reporting any form of online “binge” TV viewing.  This is where people use services like Amazon and Netflix to view all episodes of a TV series over the course of one or more evenings.  I first came across this behaviour six years ago when interviewing a busy senior executive about TV viewing behaviour.  Both people in this household led highly pressured lives and when they watched television they wanted “good TV”, with no commercial breaks, and all the episodes of a program available at once at a time to suit them.  This is exactly what Amazon and Netflix supply with their online streaming services, but this important aspect of modern TV viewing is simply not covered in Nielsen audience data.  Amazon and Netflix shrewdly keep such valuable information to themselves but Amazon knows exactly what and when programmes are being viewed using Netflix, as Amazon’s cloud services provide Netflix’s video streaming.  Unsurprisingly Amazon and Netflix are now using this data to create TV productions based on the genuine popularity of specific types of programs. For me, the interesting aspect of all this is how two very large industries - TV and Advertising - have continued to mislead themselves for so long.  Both industries are still persisting in making fundamental flaws about defining who watches which programs and when, never mind which devices are being used for viewing, and this despite the truly enormous sums of money which are involved.  Heads in the sand comes to mind.  The era of “big data” may now be revealing how totally inaccurate Nielsen’s numbers actually are, and have been for some time.  The question to ask is just who or what benefits from trying to maintain the status quo?  It’s certainly time now for the incumbent TV industry to get TV viewing in proper focus before inadequate measurement, and subsequent poor feedback, eventually leads to the loss of a whole generation of younger viewers. June 2013
Click here to download the PowerPoint chart: Click here to download the PowerPoint chart: