No one watches TV, Nielsen, and you know it
Scraping social media doesn't work either
Even in the modern world where there is more pay TV, there are few, if any, sources of professional video where consumers can know that they will encounter little or no advertising.
Pay TV networks such as Comcast, DirecTV and Time Warner Cable in the US – and Sky and Liberty Global in Europe – all carry the advertising which is sold into the channels they carry, and they add a sheen of their own advertising in the VoD, EPG or other portions of their content delivery.
Adverts are not only here to stay, but they are proliferating, and form a bedrock of revenue that prevents consumers from bearing the full financial cost of content creation.
And for as long as TV is watched every day within around 1.5 billion households globally, there will need to be ways of measuring the audience for programmes that are watched – and those which are barely tuned into – in order to put a value on that advertising.
This week, Time Warner Cable (TWC) put out a paper that attempts to make sense of new methods coming to market which try to put a different spin on the way TV audiences are measured. Nielsen is the big bad boss of the audience measurement industry and at Faultline we have had no problems criticising some of its frequent statements, mostly because the company continues to purvey the myth that TV viewing is growing, and while we would not argue that video watching is on the rise, we can see sometimes that straightforward linear TV viewing is dying on its feet.
We know of nobody who watches linear TV if it can be avoided, and yet Nielsen continues to put forward the idea that this is on the rise. This is, in part, because there is no room for a dissenting voice and Nielsen, almost a monopoly, can say what it likes.
Each year, sometimes each quarter, Nielsen changes its methodology slightly in order to make its numbers add up. TWC, along with the authors of the paper – titled "Programme Value in the Evolving Television Audience Marketplace" – have explored the idea that simply calculating how many hours a TV is switched on in a household is no longer enough, and that asking people what they watched does little more than that. Instead some measurement of "engagement" needs to emerge.
But while this paper is shrewd and intelligent at some levels, it entirely misses the point and analyses purely what‘s going on in audience measurement, rather than putting forward any theories of what else might be tried. You can download the paper here (PDF).
Why scraping comment from social media is just not accurate
The issue is that a number of measurement companies, perhaps as many as a dozen, are using screen-scraping and text-analysis tools to take comments off the public segments of social network and movie criticism sites, and attempts to parse them to establish how many positive and negative comments are made about each TV programme. They then use this as a measure of audience engagement, to go with Nielsen measures of People Meter samples which say which programmes were watched.
The paper cites operations such as General Sentiment, Radian6, Crimson Hexagon, Bluefin and Trendrr.tv, as well as opt-in versions.
GetGlue, Miso, IntoNow, Wikia, and SocialGuide all want to be the next Nielsen.
Right now US networks are all subscribing to multiple audience measurements, and with good reason. When a programme has a low level of absolute views, the only way to argue up the price of advertising on those networks is to claim – and perhaps prove – that the viewers are more "engaged," which might mean they are more loyal or that they actually pay attention when they watch this programme, instead of also doing their homework.
In fact, perversely, talking to friends on Facebook about a TV programme may indicate greater involvement, or not. It may be that a) splitting your attention between Twitter or Facebook and the TV, means you take your eyes off the screen more often, and b) that the people you are talking to may bring up other subjects and take your mind OFF what you are watching, leading to a lower level of engagement.
Another issue is that screen scraping of social media does not distinguish between comments made while watching the TV programme and those made in response to a question brought up after the programme ran. For instance a dialog on Facebook happening the morning after a popular show might go, "Did you see Twilight last night?" and the answer may be: "No, don‘t you think Taylor Lautner is mental?" "Yeah he‘s sick."
Now to a piece of software that is expected to analyse these words, it might conclude that Lautner is not popular. But of course that language is high praise when it comes from a female 14-year-old‘s vernacular. But beyond that, even if a comment on an actor is a poor one, it might be because he is a very well-acted villain, and surely that means a high level of engagement with the programme, not a low one? The whole idea of using social media has come up with these companies not because just because it‘s "possible" but because it is relatively easy to scrape comments from the public walls of Facebook and from Twitter feeds.
No one has said that this is the BEST way to measure audience engagement, but it‘s one that is available to lots of players. The paper makes this and other points and talks about this being problematic, and so is coming up with a semantic scoring system for comments. Surely measuring the volume of comments relative to comments about other shows tells us more than trying to work out who likes the show versus who doesn‘t.
While Facebook does not play in this market, it actually has the edge, in that it could peep inside of every Facebook message even the private ones, and let‘s face it, most social media comments made during TV shows are made in the private messages of Facebook, and these are the least available to mass distortion by the programme makers. They might otherwise create tons of fake Facebook addresses and fill them with commentary about a TV programme (or if you created a million fake Facebook accounts could they be made to talk non-stop about TV programmes, automatically?)
The light's on, but is anyone home?
But still these services offer something better than Nielsen, simply because of the 20 rule. The 20 rule is that every person under the age of 20 in any particular US house leaves a TV on, when they leave a room. It is also the N+1 principle. In any given US home, however many people there are in the home at a given point, this many TVs are left turned on, plus one. Nielsen has our sympathy in trying to accurately measure this. Its people meters ask people to turn on the meter when they enter a room and switch it off when they leave, but young viewers will wander between programmes and rooms incessantly and when asked say: Yes, they watched both programmes, in both rooms.
Because of young people's ability to multi-task, they may even be able to supply the plot of two TV programmes they viewed at one time, with some accuracy. They will however, have no recall of the adverts, because that‘s when they turned to their tablets and laptops and Facebooked.
So this kind of obviates ALL of this social media research, and yet it does clearly have a place, and as we said earlier, measuring the social babble on a programme, is a measure of engagement in its own right – the more people write about a programme, the better for likely future engagement. Kids think: "I‘d better watch and or catch up with this programmeme, or else I won‘t have anything to say to everyone else and they will all tell me the plot."
Another problem the TWC paper points out is that the methodology of each supplier is different and for the most part secret so that others won‘t copy its approach. That makes for a real weakness when advertisers want to compare like with like. So they end up doing something insanely stupid: they compare these new social systems with Nielsen. So that‘s comparing a system which measures just one thing – who watches for how long – with an engagement measure (who or how many people comment about this programme).
If the Nielsen measure supports the social media measure, they figure it‘s safe to use, which is the exact opposite of what they should be doing. Some people have figured it out, says the paper, and they use the social metrics to distinguish between two programmes that have the same Nielsen measurement. The one that people talk about more is the one where the advertising should be more expensive.
But of course there are two huge underlying problems with this. Firstly certain segments, age and sex are invisible in social media entirely, and others are over-represented. So while 60-year-olds still respond to adverts, and are less likely to have DVRs, they rarely have millions of friends on Facebook. So it is true that 14 per cent of the population creates 85 per cent of consumer-generated online content and while males make up just 47 per cent of people online, they make up 57 per cent of the population of sites talking about TV programmes, so they skew the data.
This is perhaps why no one gives their methodology, because it will be inordinately complex if it tries to unravel that skew.
What might be interesting is to look at other ways to measure engagement which may soon become available to us. Broadcasters and pay TV operators have apps, either "viewing apps" or "companion apps". If we assume that the great majority of TV content these days is viewed while doing something else with a companion device (in the US this is definitely true, with the same starting to happen elsewhere), then if both the TV and the companion device are focused on the same TV programme, that might be considered 100 per cent of engagement.
If they watch TV and comment on it simultaneously on a social media site, then that too is 100 per cent of engagement. We would never know which people were commenting on which programmes, but a correlation between the amount of social commentary done in real time plus viewing figures, might get us there.
But apps which listen to your screen, recognise a TV programme and then offer you companion data for that programme, could also be the primary form of audience viewing collection (except for those irresponsible teenagers who leave their tablet in one room listening to the TV, while doing their homework in front of another TV in another).
Of course the reach of tablets needs to become almost universal, something we cannot count on until at least a five-year time frame. Another area that might be fruitful is the search and recommendation elements that are gradually being put into pay TV systems, which will come to fruition over a similar period.
If a system holds recommendations for an individual, then those programmes are more likely to get viewed. Combine this with metrics on how many programmes are watched all the way through, which in many cases came from the recommendation process in the first place or from a recorded search process, and if that data was amalgamated, it would form the backbone of what "will" be watched in the coming days.
There is the problem that the method of finding a programme, using an EPG, or zapping or search, changes your attitude to the programme, and that in turn may make audiences more or less engaged. If it takes you hours to find a programme and you come in halfway through, that may lead to less engagement, for instance.
Combine all of this with underlying demographics on the household and that could form the basis for advertising placements based on programmes that are going to attract big audiences, but which may not all be watched at the same time, but which will have a high level of engagement, due to the high levels of calculated interest. Anyway, that would be in the future, when the EPG is fully replaced by search and recommendation platforms.
Facebook + new partner = next Nielsen?
But for now we feel that the social media side of this is a no-brainer and whatever the system of choice in a few years, it will need to have a social media element. So how about this?
Facebook needs to partner with someone, perhaps Nielsen, perhaps a newcomer, to make available all the TV commentary which is buried in Facebook and perhaps do this in partnership with Twitter, and provide one definitive social media "buzz" factor to augment traditional audience measurement.
Actually if Facebook could simply leverage aggregated social media commentary in all walks of life, it could probably double its revenue without showing a single advert of its own. So that would be comments about movies, but also comments about other brands, music, devices, etc...
Whichever way we end up going, this will be a slow process. Until Nielsen produces its own social measure (something it has had a stab at), it will undermine any other social media offerings out there in order to continue its hegemony and also because there are two issues here.
First coming up with a better audience measure metric which can be safely collected, and Second getting the acceptance of the advertising industry – these two are not the same thing, one lagging the other by a good few years.
Companies like Optimedia have come up with its Content Power Ratings, and these are supposed to be derived from a combination of data sources, including social media. Nielsen worked with the Sundance Channel (says the paper) to develop what it calls its Measurement Innovation Metric, which also captures a television audience‘s loyalty, engagement and ad receptivity.
So already the paper argues there is a two tier system. No-one is content to stay purely with eyeball counts any longer, so there is progress. But similarly a token introduction of social media by the same audience partners everyone is used to, could highjack any genuine change in how audiences are measured.
Finally Addressable advertising, where it is possible, has the power to change all that. Putting adverts into programmes in real time, to the right people who are known to have an interest in a subject, who have the wherewithal to afford the advertised product and who are watching TV now, and perhaps even watching a programme on a similar subject, leverages advertising appeal by multiple dimensions.
So, you have to measure by demographic, region and proven interest, and if anyone sets up such a system, they have already moved beyond engagement metrics, and we note that many pay TV operators have already begun to move down the addressable advertising route, although its acceptance among advertising agencies is almost zero at this point.
Copyright © 2012, Faultline
Faultline is published by Rethink Research, a London-based publishing and consulting firm. This weekly newsletter is an assessment of the impact of the week's events in the world of digital media. Faultline is where media meets technology. Subscription details here.