Tuesday, October 16, 2012

Neilsen Introduces Eavesdropping Meter

Ratings firm Nielsen is introducing a new form of metering device as it expands local samples.  The new device listens for audio cues embedded in television programs (as well as other content like ads and PSAs) to record viewing behavior - and pushes that data to Nielsen using wireless technologies.  Nielsen hopes that this less intrusive technology (compared to the hard-wired Local People Meters) to expand its sample size in some local markets.
  Nielsen will initially install the devices in three local markets - in homes with LPMs as well as in newly recruited households.  Eventually Nielsen hopes to expand the system to include "audio code" samples in all US television markets, to supplement LPMs, traditional diaries, and data gathered from set-top boxes.
   Audio code readers have been used to record air play of songs and commercial in radio for several years, and to track PSA and promo package use on television channels.  Those systems are typically employed on computers, however, as the archive of audio cues needs to be constantly updated to include audio cues for new program segments.  And the TV ecosphere of programming content and alternatives is much more complex.

  For Nielsen, the viability of the data will depend heavily on whether or not it can reliable record TV use.  There are several specific challenges Nielsen faces in that regard.  First, they have to be able to distinguish and record the full range of TV viewing options, and reliably distinguish among the panoply of TV content. Specifically, Nielsen will need to develop and continuously update an archive of audio cues to use to determine what program is being heard, and have the processing power to recognize and distinguish audio cues - otherwise, content is recorded as being unknown.  Having too many unknowns raises questions about validity and reliability.  Essentially that means Nielsen will be to either work with program producers to embed predetermined audio cues throughout the TV/video industry, or develop a way to continuously update archives of audio cues used to distinguish content.
  If Nielsen only codes for program (and not for individual episodes), then the need for updates is reduced - but it also would mean that the system would not distinguish among episodes - particularly among episodes that are time-shifted.  Is the episode of NCIS heard the most recent, or the cliffhanger from last season?  Is it an episode timeshifted from CBS, or watched on cable channel USA, through a MSO's On-Demand offerings, streamed from CBS.com or other video streaming services, or viewed from a DVD?  If Nielsen wants to be able to differentiate among program source, it needs to use a audio code system that assigns unique codes to each source.
  So a comprehensive, continuously updated set of audio cues is needed - and is needed at the location where the determination of viewing is done.  If that's at the audio code reader - then that device needs the processing power to determine viewing, the storage capacity to hold the archive of audio cues, and the means to continually update that archive.  Or the household audio code reader merely listens, and sends what it hears to a location with the continuously updated archive and processing power to determine what it is hearing.  (And if its this latter case, then there are some really serious privacy issues at play).
  Wherever the coding is done, there are other potential problems with trying to tie viewing to audio cues.
First is the obvious problem that Nielsen's been battling since it started using meters: that hearing an audio cue may mean that the TV's on, but not that anyone is watching (or how many are watching).  If the environment is too noisy, will the sensor be able to differentiate audio cues from background noise?  Then there is the issue of what if the audio cue isn't audible - say if someone is watching and using headphones or for some reason has the sound off,  or bypasses the portion of the program containing the audio cue for one reason or another (switching among programs, starts viewing late, skips or fast forwards timeshifted or streamed content, etc.).  Will the audio coder be able to differentiate between viewing on a TV set and second-screen viewing?  Will Nielsen be willing to place an audio coder in each room containing a TV set in multiset households.
  The audio coder systems used to track song play on radio labels designates programming it doesn't recognize as "unknown."  The questions I raise about Nielsen's implementation of audio coding of audio cues comes down to an issue of how much of viewing behavior is missed or coded unknown by whatever system Nielsen develops.  I can see a good system being able to recognize and record when a program with a recognized cue is on, in normal (traditional) viewing contexts.  What's unclear is this particular system's ability to reliably recognize and record the full panoply of program and viewing options, in less-than-perfect contexts.  I suspect that these are the questions that the Media Rating Council (MRC) will ask before it certifies this new approach to measuring viewing.

Source -  Nielsen's New Local TV System Moves Into Living RoomsTVBlog

No comments:

Post a Comment