Algorithm And Blues: How Google Cracked The TV Ad Marketplace

When Google TV Ads chief Mike Steib and his team first approached Bloomberg Television about placing inventory up for bid on the fledgling Google system, executives at the cable network were intrigued. The opportunity to attract a new crop of advertisers -- and more so, tap into a wealth of data -- held promise.


Still, as with many networks approached by the Google sales force, deciding to upend the traditional sales process and sell spots via an online auction was cause for debate. As negotiations began a year ago, executives at the financial news network weighed the pluses and minuses internally for some time.

Bloomberg eventually made the decision to pull the trigger -- reaching a deal with Steib, the director of Google TV Ads, and his team to sell some of its national inventory (which reaches some 50 million homes) on the platform.

But offering advertisers the chance to bid for spots on its "Starting Bell" or "Taking Stock" shows wasn't going to happen overnight -- multiple logistical and technical hurdles remained. And engineers on both sides worked for months to make sure Google and Bloomberg's inventory management systems were synchronized.



"The world's worst nightmare would be for an advertiser to buy an ad on Google TV Ads and for it not to appear properly," said Trevor Fellows, Bloomberg's head of advertising sales.

Now, some time this month, advertisers will be able to bid for Bloomberg spots and watch them run on the network nationwide.

NBC Universal also has a deal with Google that enables it to place inventory on six of its networks (Sci Fi, Oxygen, MSNBC, CNBC, Sleuth and Chiller) up for bid on the TV Ads system. The first NBCU network will be online by the end of the year--it's unclear which one--and the remainder early next year.

The two deals are the product of immense effort on the part of Steib and his sales force, who doggedly pounded the pavement looking to sign up networks willing to place some national inventory in the system. Google has declined to discuss the size of its sales team.

Many cable networks approached by Google declined to participate for a variety of reasons, including a reticence about ceding inventory to an electronic buying system that could lead to lower rates. A sales executive who held unsuccessful talks with Google suggested a scenario where an advertiser already on a particular network would switch to TV Ads.

"Maybe I already have them on the air, and at the last minute through Google they can buy it cheaper than I sold it to them the week before--why would they ever buy it from me again?" the executive said.

Despite that executive's concerns, Google continued to attempt to get a deal done, offering various proposals--some it clearly thought were favorable enough to tip the scales--but the executive took a pass.

In order for its system to be a game-changer, Google will have to persuade more networks to place national inventory up for sale. Even NBCU didn't dive in headfirst. Its offerings don't include the NBC network or the popular USA and Bravo cable channels yet.

But Steib has indicated that others are coming. He said the platform is tailor-made for networks similar to Bloomberg and NBCU's horror-themed Chiller and mystery-oriented Sleuth.

All three are not rated by Nielsen, which can be deterrent for advertisers. But the Google platform allows advertisers to bid for time, and to receive second-by-second data on how their spots performed.

"Advertisers are increasingly demanding accountability for the money they spend and as an unmeasured network, it can be harder," Steib said. "We're going to measure them as well as anyone's measured in television."

Before its recent deals--since April 2007--Google had been selling spots on Dish Network, reaching the satellite provider's 13.8 million homes. Because Dish homes have set-top boxes, Google can obtain "census-level" data on viewing patterns down to the second. (Google also places some spots reaching some 25,000 cable subscribers in Northern California.)

But not all Dish homes are equipped for Google to obtain that type of viewing data. It has a subset that it believes is large enough (it won't say how large) that allows it to extrapolate and provide ratings data reflecting what happens in all 13.8 million Dish homes.

(In that sense, it acts like Nielsen--albeit with a much larger sample size and one that has "hard" set-top box data to work with.) Going forward, the Dish subset will provide the basis for metrics Google produces for the national spots on TV Ads. Google's complex algorithms will look at the performance of ads in the subset of Dish homes--perhaps between 2 million and 5 million of them--and project what takes place with viewership nationally.

Take NBCU's Sci Fi Channel, which is in 93 million homes: Google will look at what the network's viewers are watching in several million Dish homes, and then extrapolate it to provide the second-by-second data for what happens across the country. Google has a deal with Nielsen, where it can use Nielsen data to get demographic breakdowns for a channel's audience--such as what percentage are men ages 18 to 49.

There is the potential hitch that the Dish subscriber may, in general, have a different profile than those who subscribe to cable or even satellite competitor DirecTV. So, if the sample is generated from Dish set-top boxes, it may not be representative of national viewing behavior. But Steib said Google has accounted for that, with sampling "done at such a scale that there is a high degree of statistical relevance."

In addition to the opportunity to gather performance data, Steib believes networks will increasingly sign on to offer national inventory on Google TV Ads for a more immediate reason: revenue. The system offers the potential to drive sales on inventory that's difficult to sell, he said, by offering the opportunity to tap into a new base of possible clients.

Google says it has done business with a slew of advertisers that would be interested in advertising on TV and can afford it. But they don't have the creative, and may perceive TV as something out of their price range. Google has a system that can create an ad for a marketer.

"You bring TV advertising within the reach of companies that may never have thought about it before," Bloomberg's Fellows said. "And that's really interesting to us."

Much of TV Ads' promise is indeed the opportunity for a mid-size business without an ad agency to begin to use television. And Google cites examples of several. But the system is also employed by some of Madison Avenue's largest agencies, which would appear to welcome the chance to expand its use with national spots.

As the cable industry has explored ways to more efficiently sell its loads of lower-tier ad time, the term "remnant inventory" has spread further into the lexicon. But to networks, it's a bête noire, as it implies some ad space holds little appeal. Regardless of any preferred term, Steib believes TV Ads offers a prime opportunity to drive value for it.

"We think overnight spots, under-monetized daytime spots, spots that may not be sold on Chiller, or at 2 in the morning on Bloomberg--we love that inventory," he said. "People use the word 'remnant' as if it's a bad word.

"We're very much inspired by the way advertising has been sold on the Web, where there isn't remnant inventory. You are targeting audiences and you're targeting behaviors and whenever there is something that's unsold, a marketer swoops in and buys it so they can reach their audience. We think TV should work much the same way."

Steib said Google isn't looking to persuade broadcast networks to place high-demand prime-time inventory up for bid. "That inventory doesn't need our help," he said. "Where we're helpful is with the next few tiers."

While NBCU won't be offering time on "The Office" or even "Project Runway," he said "there is a good-faith commitment to make inventory available" that is more valuable than, say, "Simon & Simon" at 7 a.m. on Sleuth. "Both parties are committed to seeing this useful for advertisers," said Steib, a former NBCU executive.

MediaVest executive vice president Pam Zucker said the addition of national inventory "begins to fulfill the promise of what Google TV can be." MediaVest has used the system since its inception, and she suggested agencies might begin to experiment with using it as part of a media plan, rather than more of an experiment.

"As Dish Network-only inventory, it had learning opportunities, but I think it was never going to be a long-term value proposition," she said. "As it begins to get national supply, it begins to potentially fulfill strategies as opposed to fulfilling test-and-learn scenarios."

As many of the cable networks that Google looked to cut deals with balked, others looked under the hood and saw a risk-free opportunity. "There's no downside," a sales executive said.

Many of the intricacies about the opaque system are either kept under wraps by Google or are just simply hard to comprehend. Fellows, the Bloomberg sales head, said there's a "secret sauce."

Nonetheless--to be certain--what follows is a highly oversimplified flow chart showing how a network could find the system risk-free:

Take NBCU. The company's deal allows it to set a "floor" for any spots it places up for bid--meaning there's a minimum price it will accept. If the auction yields a winning bid below that "floor," the spot effectively drops off the board and isn't sold.

(NBCU also has control over its inventory. Fellows said Bloomberg "can put inventory on the system for sale and can move it in and out as you wish.")

Using Sci Fi Channel as an example, no advertiser can buy a spot on "Battlestar Galactica" for $20,000--if NBCU has determined it won't take less than $25,000.

(Advertisers actually don't offer up a per-unit price in the auction, but a CPM that leads to the network's take. So NBCU could set a $10 CPM "floor" and if the top bid is $6, it can nix the auction.)

Some industry executives have raised the issue of what Google takes home on a winning bid (effectively its commission). A network could worry that Google is simply making money off its inventory.

It is. Still, a network arguably still comes out ahead.

Back to Sci Fi: Say it is having trouble selling afternoon inventory on Saturdays. The network then puts some of that inventory up for bid. It sets a "floor" where it won't accept less than what amounts to $5,000 a spot.

Next, an advertiser uses the system and successfully buys a spot for $10,000.

Sci Fi doesn't collect the full $10,000. Instead, it takes home about $7,500. Google gets about $2,500 (25%).

Even so, the $7,500 Sci Fi collects presumably is more than it could have brought in on the open market.

NBCU's deal calls for Google to collect about 25% to 30% of a winning bid's total amount, a person familiar with the pricing structure said.

Google does not comment on commission structures. Bloomberg's Fellows declined to comment on his network's deal.

Is the system transparent? For the buyer, no. A "winning" advertiser knows what it paid--say $15,000 for a spot.

That's what it's worth to them--and presumably they walk away happy. They do not know what the network ultimately collected or Google's commission or what other advertisers are paying for the same inventory. That would, of course, allow them to alter their bidding strategy the next time. Steib said Google encourages advertisers to "put in what it's worth to you."

On the other side of the fence--the network-selling side--there seemingly is some transparency A network would appear to know what the winning bid is for particular inventory--if for no other reason than it's aware of the commission agreement it has with Google. Knowing how much its inventory goes for could prove to be valuable information for adjusting rates on the open market and for other purposes.

The system lacks transparency, of course, in that networks don't know how others fared. So, competitors Bloomberg and CNBC don't know how much advertisers paid for their respective spots via the system.

For the buyer, the system has one other intriguing wrinkle--put in place to avoid "buyer's remorse." It's complicated, but essentially the highest bidder only has to pay what the second-highest bidder offered.

So, if Microsoft successfully bid $2,000 for a spot on MSNBC and Intel then came in second with a $1,000 bid--Microsoft would only pay the $1,000. MSNBC would lose out on half.

The bid would actually be done using CPMs. So, Microsoft may bid a $20 CPM for a certain number of impressions and Intel $10 for the same reach. Since it's the high bidder, Microsoft's cost would drop to the $10.

Within 24 hours of a spot running, an advertiser receives a detailed report on how it performed, including how many impressions an ad generated, and the percentage of viewers who tuned in to the full spot.

At Bloomberg, Fellows said the opportunity to collect data on viewer tendencies was the driving force behind the network's signing on to the Google system--even more so than the potential for increased revenue. "This is really just part of our interest in understanding and working with robust metrics for media," Fellows said. "I don't think it will have a significant impact on revenue in the short term."

Fellows suggested the second-by-second tracking could serve to boost revenue in what might be described as "through the back door." He said mining the data could help improve the on-air editorial product. Viewership patterns could be evaluated based on what happens when there's a switch to an anchor in London, or a breaking news alert.

"At the moment, whether an anchor is popular with the viewers is really the opinion of the producer and the TV executives," he said.

More directly in the ad realm, there's the potential to gauge how well a particular tease works before a commercial break in getting an audience to return.

Over time, Fellows warmed to the possibility that Google data will offer what Nielsen local-market data does--performance metrics for DMAs such as New York and San Francisco.

Sales and buying executives also cited Google TV Ads' potential to help them streamline back-office and human-resources operations. While an automated system will never allow for complex multi-platform deals or network upfront agreements, it could create some efficiencies for some aspects of day-to-day operations.

"If you're in the business right now, you've got to be looking at ways of changing the model of doing business to ease it and take some of the labor costs out of the process," said a cable sales executive who has met with Google.

A new report from Citi Investment Research wrote that for agencies, "easier ad buying could ease cost pressures, so long as they are not dis-intermediated by Google."

So, will Google TV Ads ultimately prove to be a notable player for a company that had second-quarter revenues of $5.37 billion?

Steib thinks the platform offers a compelling proposition for enough networks to join that will propel growth.

"Here's what I think is a really favorable deal: a technology system that we essentially make available for free that gives a network access to every advertiser in the world on a self-service basis--and one that optimizes the yield for the network and gives them control over their inventory," he said. "We think that's about as good an offer as you can possibly give somebody."

Next story loading loading..