• by Frank S. Foster , November 10, 2008

"You're saying that as if the television industry differentiates between the two," I answered.

"There is a difference. You should know that."

"I agree there is a difference. But 'viewing' as defined in context of the people meter is not the same as 'viewing' with respect to a tuner meter/diary environment."

"They are both viewing -- both have people data. You are making my point for me," he retorted.

"Really? The tuner meter has always produced people data?"

"No. The VIP contains more than tuner meter data."

"Combining four sets of virtually worthless diary data with tuner meter data from a completely separate panel and, VOILÁ, what was crummy on its own has been transformed into wonderful 'viewing' data?"

"Yes. Besides, modeling demographic data is problematic."

"And what would you call Nielsen's approach to demographics? Are you suggesting they are NOT modeling? Just because they use a really crappy model doesn't negate the fact that it is a model."

"True. But their model is familiar."

"So because it is YOUR crappy model, it's okay? You plan to make that argument with me, of all people?"

"What you don't understand is that no one cares about boxes, we buy and sell access to 'viewers'."

"No, you don't. You just think you do."

There was a pause in the conversation, and then he said, "Excellent argument. How 'bout another round?"

The lesson is clear: good tuning data be damned, it is really all about crummy viewership data. While I will be the first to cede that data obtained from set-top boxes is tuning data and not viewing data, there is the small issue of relevance. When the subset of a small, opt-in panel watching a particular network is in the low double digits -- as it is for nearly all locally inserted, advertising supported television stations and networks -- the demographic data is inherently error-prone and the resulting demographic ratings rife with error. It would not matter if the process used to select panelists were flawless -- nor if the technology used to measure viewers were perfect -- when the numbers are so small.

Add to these inherent problems the fact that traditional "viewing" data requires ancillary action on the part of the sample -- write down your viewing patterns, push a button, push it again -- and the resulting ratings should be viewed with more than a fair degree of apprehension. But of course they are not, they are deemed "perfect" and without error. It is not uncommon to see Nielsen ratings in local newspapers with three significant figures. Someone from Nielsen should go back to college and take notes during statistics class.

Compare this situation with new forms of television distribution. When analyzing broadband, researchers need not rely on data obtained from sample populations monitored by a research firm. Broadband video, even when it is broken into five- or 10-minute streams, can measure "tuners," which -- last time I checked -- were counted, not estimated. When video-on-demand numbers from Rentrak are published, those figures again are tallied, not projected. Addressable advertising powered by Acxiom or Experian can direct advertisements to specific households -- no estimates, no projections.

While STB data currently has no inherent demographic attribution, the tuning information is based on very specific and auditable transactions. Those transactions can be grouped at the household level and compared to actual subscriber numbers. The data can be used to project analog-only households and broadcast-only as well. The resulting analysis has the potential to be much more accurate and reliable.

Are we really that uncertain about the demographic makeup of the audience that regularly tunes into Fox NFL football? How about loyal viewers of the Golf Channel? Does anyone have a question about who watches Oxygen? At the end of the day, the historical tuning behavior of a set-top box is arguably a more accurate tool to project audience demographics than labels developed from a biased, error-prone small panel. While such analysis would require a fair understanding of mathematics, that should be viewed as an asset, not a liability.

At the end of the day, media-buying decisions need to be made with eyes wide open. The future of television audience measurement is set-top-box data. Anyone suggesting otherwise is either lying, or trying to sell you something rotten.