As it does with its AdWords search system, Google plans to sell spots based not only on price but on how well an ad performs. Better-quality advertising -- advertising the audience deems most relevant -- will be rewarded, as marketers will pay less for the same spot. Lower-quality advertising -- the ads deemed less relevant by consumers -- will acquire higher media bills. Simply put, Google's TV Ads could go a long way toward answering who is responsible for bringing viewers in and who is responsible for retaining viewers, the network or the advertiser?
Such a system is still a few years away, but Google is serious about it. The company has a deal to sell ads on 120-plus cable TV networks through EchoStar's Dish Network and collect second-by-second audience ratings for the buys made through the system. Google has long asserted that some of the tools it has brought to the online world, particularly the search-advertising marketplace, could be important to the offline world. That's why it's worth paying attention when Google's top ad-sales executive, Tim Armstrong, said that the biggest change in Google's business model had nothing to do with technology but "had to do with implementing ad quality into our business model."
Google is still a bit TV player -- its ads reach less than 15% of the total TV universe -- and to assume that it would be powerful enough to influence the entire TV marketplace would be misguided. But the premise is an intriguing one and echoes sentiments TV titans from ABC's Mike Shaw to CBS's David Poltrack, have made over the past few years.
Consider how it works in search, where an ad is essentially a product of two variables: the price of the bid and the historical click-through rate. Paying the highest price for an ad doesn't automatically mean you'll win the top search-ad slot. In search, Google uses click-through rate as a proxy for ad quality; in TV, it's looking at other potential proxies -- for example, the percentage of an ad's audience that is tuned in at the beginning of the ad and stays tuned in for the duration.
A few weeks ago, on a sunny brisk April day at the Googleplex in Mountain View, Calif., Keval Desai, director-product management for Google TV, enthusiastically showed off how the Google TV-ads system works. It looks remarkably similar to the buying tool Google uses with search advertising, and with good reason -- it's built using that architecture.
The idea of TV Ads is predicated on the notion that TV is beginning to look more like the internet. Fragmentation means that networks with less than 1% of TV share on their own collectively account for 50% of viewers, and networks with less than 0.5% of a share together make up 30% of total viewers. Using the remote control as a proxy for the mouse, Google can study TV "surfing" behavior.
Mr. Desai listed all the information Google can glean from its second-by-second data: the number of impressions an ad got, when the audiences began dropping off, how many stayed tuned in for the entire ad and the average seconds tuned in per impression.
Adding quality scoring is a natural evolution. The proxy for quality, Mr. Desai said, could take the form of "initial audience retained" -- a metric currently offered through the TV-ads system. He said there are three steps to getting its system ready to add an ad-quality metric: 1) create an accurate and timely measurement model; 2) decipher the quality signals; and 3) add those quality signals to the ad-buying auction. Right now, Google appears to be somewhere in between steps one and two and is moving deliberately.
"We've only been up and running for six months, so we need more data to start productizing it," he said. "We don't want a repeat of C3." (C3, of course, being the controversial and long-debated commercial ratings system Nielsen implemented last year that measures DVR playback up to three days.) Google has an advisory committee of ad-buying executives who help it determine what kinds of data are most valuable.
"When you get to an environment where a commercial unit becomes the unit of negotiation ... taking a look at commercial's ability to hold the audience is becoming more important," said Bruce Goerlich, ZenithOptimedia's head of research and a member of that Google advisory team. He said there is potential for ad quality to factor into TV-ad buying, but right now he said his agency is using the data more as a diagnostic tool.
Tim Hanlon, who heads Denuo's new-ventures practice, is blunt about the data's shortcomings: "Second-by-second data is interesting but by no means a sole arbiter of anything." He said that to really create a robust, realistic look at how people view TV ads, the data need to include more robust demographic sampling information and other qualitative factors.
It's true that right now the data is limited: Google runs ads in about 14 million homes that use the Dish Network. But not all of those homes can transmit second-by-second ratings data back to Google, as that requires "return path" technology. Google won't comment on exactly how many homes let it gather data, but a few TV research estimates have pegged the figure as slightly more than half.
Google also doesn't know the demographics of those people who have tuned out during an ad. Through a deal with Nielsen, it can get some demographic data, but for the most part, it can't tell exactly who is tuning out. If an advertiser is running a spot for men's deodorant, for example, and 20% of the people tune out halfway through but they're almost all women, that ad may be no less effective despite the drop off.
Google TV Ads Director Mike Steib said Google is working with most every major agency on the TV Ads product but said right now the data is most valuable directionally and that's how agencies have used itdto gauge the performance of certain Long Tail TV networks that may not be measured by Nielsen or to understand where creative weaknesses may lie. When asked how long it may take to introduce a quality factor into the buying process, Google was noncommittal.
"It took us two years to get here in search," Mr. Desai said. A reporter baited him with the question: So will it take two years for ad quality to factor into the TV-buying system? But he didn't fall for it: "We hope to get there in the near future."