Now that “Smart” TV’s are prompting us to update their software nearly as often as our computers, it’s becoming easier to imagine a world where real-time programmatic television advertising is a reality.  This shift, in itself, is a monumental change in the way the advertising ecosystem works – but will this disruptive innovation have any impact on the TV programming we have today?

With this problem on my mind, I sat down this weekend to relax and, well, watch TV.  After watching more hours of “Shark Week” than I care to admit I had an epiphany: programmatic TV advertising could mean the end of a large portion of cable TV.

Let me explain:

Thinking back to my media planning days, I remember that the average cable TV show has a rating of 0.5.  For those of you who are unfamiliar with Nielsen’s TV ratings, this means that the average cable television program is being watched by ½ of 1% of US households.  On the surface, this seems fine – but you have to remember that Nielsen ratings are derived from Nielsen’s TV panel. When you take into account that Nielsen’s TV panel includes ~25,000 metered households (about 0.02% of the 114MM total US households) this means that the average cable TV program is viewed by just 125 metered Nielsen households.  By contrast, a program with a rating of .1 (such as an average program on Soap Network, the Science Channel, Speed Network, the Golf Channel and many others) is watched by just 25 metered Nielsen panel households.  An increase of only 10 households watching a 0.1-rated program would increase the programs rating by nearly 50%.  Needless to say, at this level of granularity, there is a huge amount of sampling bias and inaccuracy in cable TV ratings.  Also remember that today there are thousands of cable TV channels – there is no doubt that many of these very small cable TV properties are getting more advertising spend than they deserve just because of the sampling bias introduced by Nielsen’s panel based ratings system.

Now, what happens when we no longer have to count on a relatively small panel of individuals to estimate how many people are watching TV and instead, we have access to advanced metrics, including: how many people actually watched a specific commercial, whether the TV was muted or unmuted, whether the viewer changed the channel part way through the commercial, etc.?

When we have access to all of this rich data, rather than just a rough estimate of one data point (how many people are watching), I think the rich variety of cable TV channels we know today will be no more.


Will Programmatic TV Advertising Kill Cable TV?
  • Well, the idea of a cable “channel” exists only to work around some of the limitations in metrics. If you have more advanced metrics, than you can disaggregate your unit of measure into individual shows, commercials, etc. The more data you have, the smaller your unit of measure (well, to a point – it’s really the more disaggregation you have).

    I mean, I disagree a bit with your criticism of the currently available statistics – I’d prefer to see the margin of error in the actual sample. Also, I’d imagine that the Nielsen devices actually measure a few of the things you bring up (muting, skipping channels midway through commercials, etc). Depending on the degree of annotated data they have on each channel (when commercials run, what was running at the time) they should be able to derive a lot of that information.

    Also, people are working around some of the limitations… Arbitron, for instance, employs people to carry around little devices that record samples of whatever the individual hears, and then they can cross-reference the audio samples to actual shows and commercials. I’m not sure how granular they get, but with increasinly sophisticated measurement devices I don’t think “smart TVs” are necessary.

    (Also, the replacement lifecycle for TVs would necessarily be rather short for your scenario to become effective, i.e. to be able to simulate the level or tracking we have with online advertising and/or TV).

    Finally, one of the core problems remain – it’s still a form of display advertising, i.e. you can’t track the conversion rate (except for those commercials that ask you to call in “in the next 10 minutes”). Knowing exactly how many people see your ad, skip your ad, etc doesn’t make it any easier to determine the real impact of a single piece of advertising in a multi-channel advertising campaign over time (time series analysis).

  • Andrew Eifler

    I like the Arbitron thing. I think someday there will be panels of people who wear special eye glasses that record everything they see and then we cross reference the video data from their life-streams with known advertisements in the market at the time.

    I suppose video-eye glasses would solve the conversion problem too – because you can see when they buy stuff, but that would be harder. Might also be hard to get people to sign up to be part of the panel.

  • Yeah, more accurate data collection is better.

    I think the core point of yours that I disagree with, though, is the idea that you need to measure everything – that samples are invalid. Sure, they can be invalid, but if measure are taken to select appropriately from the population, they should be pretty damn close to accurate.

  • Andrew Eifler

    Good points here as well. I guess I’m just stubbornly skeptical of sampling bias. My rule of thumb is usually: I trust all survey based research studies – except those studies that I did not conduct myself.