Merchants spend tens of billions of dollars on advertising in the U.S. Cable operators are of the opinion that they deserve a bigger piece of that pie, but until they can more accurately target potential customers, advertisers aren't going to be very accommodating.

Picking Your Spots

Most of the major cable operators last year implemented a way to target ads by regions sometimes as small as metro zip code areas, but that measure was largely preparatory for what comes next.

Comcast Cable, Charter Communications, Time Warner Cable and other operators are now testing various technologies that will enable them to target audiences with a finer granularity, from zip codes to neighborhoods of a few thousand subscribers and beyond to individual set-top boxes—the ultimate target in targeted advertising.

Telcos, with their fiber-based switched video services, are in a position to ship specific ads to specific set-tops the moment they hook up a subscriber. For now, the issue is that they haven't hooked up nearly enough customers to attract an appreciable amount of business from major advertisers. That won't be the case for long.

For competitive reasons, operators want to be first in the traditional TV world to offer the ability to target customers and simultaneously have enough customers to attract advertisers. The expectation is that within six months, cable operators will start rolling out the capability to aim ads at neighborhoods and smaller groups of potential customers, and that within two years, most will be zeroing in on individual set-tops, one way or another.

In addition to subscriber numbers and the technological capability, the other elements necessary for attracting national advertisers are verification and the availability of demographic data. When advertisers spend money to reach consumers, they like to have assurances those consumers actually saw their ads and have a high likelihood of becoming customers.

Viewership numbers (verification) and demographics have long come from Nielsen Media Research. Advertisers have been grumbling for years that Nielsen's TV viewership numbers are unreliable, and if they weren't before, the proliferation of devices that enable ad skipping and ad zapping now justifies the complaints.

Figure 1
Figure 1: Cable operators install Digital Program Insertion (DPI)
systems in hubs to enable the capability of splicing local ads in digital program streams.
Scientific Atlanta uses its DCM (Digital Content Manager) to splice ads into program streams.

The availability (or lack thereof) of verification and demographic data goes a long way in explaining why advertisers have been shifting their ad spending from TV to the Web. Spending on all advertising was up 4.2 percent in 2005, but spending on Web advertising was up 23 percent, by far the fastest growing category, according to Nielsen.

With ads on the Web, advertisers know not only that people saw their ads, but they can also identify the specific PCs the ad was delivered to, determine if the person behind that PC clicked on their ad, and also get some details about that person's online behavior. That data can then be cross-referenced with other information (e.g., zip code) to render a demographic profile.

When national advertisers buy TV ads, they overwhelmingly buy ad time directly from networks—CNN, Disney, etc. Although verification is weak, the networks can still offer millions of viewers with definable group demographic characteristics.

National advertisers are key, because they're the ones with the biggest ad budgets. National advertisers rarely buy ad time with individual local cable operators (ignoring for the moment that some MSOs also own national programming networks), mainly because they're not big enough (with a very small number of obvious exceptions), but also because demographic and verification is rudimentary, especially in comparison with the Web.

What cable operators get to sell are the ad slots the networks reserve for them in their program feeds. Operators sell them primarily to regional and local advertisers.

Local/regional advertising is nothing to sneeze at. Those advertisers spent over $5 billion with cable operators in 2005, and even with rudimentary regional targeting, most expect that can double or even triple by the end of 2007.

But once cable operators can target individual set-tops, they believe they'll be significantly more attractive to national advertisers. While operators might siphon some national ad dollars from the networks, they are more likely to attract the "spend" that advertisers are shifting away from network TV and print (now being diverted to the Web), and to also grab a good share of any increase in spending.

In the last year, most of the major cable operators have upgraded to digital distribution. By distributing digital program insertion (DPI) capability throughout a system, typically at neighborhood nodes, operators have enabled themselves to aim ads at viewers on at least a regional basis at a fairly low cost.

In this configuration, an operator can have as many ad zones as it has nodes equipped with DPI. From the cable operator's perspective, that might be overkill, however. There's no economic justification for selling a single ad slot in 20 different zones to 20 different advertisers.

For the local advertiser, that's also frequently too fine a dice. Rarely will an advertiser have as many as four or five different ads (or variations of a single ad), and in demographic terms, one only needs a few geographic divisions before getting to a point of diminishing returns.

The more common scenario in regional targeting is for a cable operator in, for example, San Francisco, to sell an ad slot to a local auto dealership in Marin County and show that ad only in the northern half of San Francisco, and then sell that same ad slot to a local dealership in San Mateo and show that dealer's ad only to the southern half of San Francisco.

Boosting accuracy

Eric Conley
There's limited ability for verification in this system, and even when monitoring is in place, it's still inaccurate, according to Eric Conley, CEO of Mixed Signals. "Failure rates are higher than embedded monitoring systems say [they are]," Conley says.

Problems include everything from improper insertions to incorrect metadata to compromised audio, video or both. Many of these problems don't manifest themselves until the signal is sent downstream.

Mixed Signals’ Sentry is a box that sits in headends and monitors video signals after they are transmitted by the video servers but before they hit QAM modulators. The Sentry then provides full audit reports. The system is being used by Adelphia, Cablevision, and Time Warner Cable.

With solutions for verification becoming available, the next targeting step is to aim at individual set-tops, standing in for the consumers behind them.

Terayon Communication Systems also envisions a scheme in which multiple ads would be transmitted, and the set-top would be equipped with the capability to select the single ad most appropriate for that viewer. Terayon says it is working with a number of set-top box companies to create the necessary software. The technique would allow operators to target individuals until the operators begin switched transmission, a la the telcos.

Mixed Signals' Sentry Device
Mixed Signals’ Sentry device monitors video signals before they
hit QAM modulators, and provides full audit reports.

BigBand Networks advocates jumping right to the switched broadcast model, and several operators are considering that advice, according to the company. It says several partners are testing its switched broadcast technology. The vendor would not disclose specific partners, but Time Warner and BigBand did co-author a paper on the subject at last month's National Show in Atlanta.

Cable operators could start, if they chose, with a multicast model of switched transmission. If one person in a system is watching Lifetime, for example, the operator would still send it as a multicast signal. If a second person tuned in, a server at the edge would join that viewer in to the multicast, thus conserving bandwidth, explains Paul Delzio, director of broadband systems at BigBand.

The next step is unicast, which not only allows operators and advertisers access to individual set-tops, but also to conserve bandwidth. If they know that only 80 channels of the 150 they have available are being watched by at least one person, that's 70 channels they know they don't have to transmit. Bandwidth reclamation in a unicast model might be less efficient than it is in a multicast approach, however.

"Cable operators are up against the telcos," Delzio says. "Fiber gives the telcos unlimited bandwidth. In a switched environment, operators can compete by better managing their bandwidth."

The switched model is compatible with the approach described by Terayon earlier in which set-tops are enabled to select an ad from among several streamed simultaneously to the box.

Delzio predicts cable operators will support up to 10 million cable homes with switched broadcast by the end of this year, and predicts that Time Warner will have migrated entirely to switched broadcast by the end of next year, at a fraction of the cost of adding bandwidth.

Putting ads on-demand

A parallel development in the TV ad market is the insertion of advertising in video-on-demand.

Today, ads are stored directly with specific VOD content. The problem is that VOD content can sit on a server for weeks (sometimes months) before getting refreshed or rotated, which means any accompanying ads have to be valid for at least that long. That rules against ads for imminent events, sales, or anything else that could go stale.

Advertisers want to decouple ads from VOD content, so that they can refresh ads more often, and do so in a more targeted manner.

One approach to ad VOD is to overlay interactive capability. Viewers might see a banner ad or be presented with a short version of a video ad. If viewers want more information, they can click on the ad, which triggers a VOD session, in which the server retrieves the requested content. The practice is commonly called telescoping.

Interactivity typically requires software on the set-top. The interactive approach is being championed by companies such as GoldPocket (now part of Tandberg TV) and Navic Networks.

Charter is using Navic's Addressable Advertising Service in Riverside, Calif. Charter is testing the technology not only for advertising, but also for polling, and other interactive functions.

The other approach is to use a decision manager. VOD programs are encoded with "bumpers"—spaces for ads at the beginning and end. When a subscriber orders a program, the decision manager tells the server which bumper ads to play. The server then retrieves the ads and the program and streams them to the subscriber. A program stream built this way is referred to as a playlist.

An advertiser can associate specific ads with specific content—always send the latest ad for beer with this film, for example. But the decision rules could be set to depend on the demographic character of the subscriber—send the traditional lager ad if the viewer of this film fits one demographic profile; send the ad for the microbrew if the viewer fits the other demographic profile.

It appears that Comcast will be the first MSO to roll out ad VOD across its entire digital footprint (about 9 million homes) starting in June. The capability will initially be available in only three less prominent VOD networks (SuccessTV, Music Spy Videos and DriverTV), but the experience will be instructive for all involved. Comcast is using Tandberg Television's AdPoint system.

Additional storage might be required for this approach, and bandwidth usage has to be managed carefully, but can be implemented in software with existing VOD servers.

Atlas on Demand—which honed its ad-delivering skills on the Web—is one company providing software that can provide rapid decision-making for ad VOD. Atlas SVP and GM Scott Ferris says the company's software works on VOD servers from any vendor, including those from C-COR Inc., Concurrent Computer Corp., and SeaChange International.

With Atlas, C-COR notes, it can provide verification in spades. "I can tell if someone watched an ad, fast-forwarded through, skipped it, or paused it," explains John Boland, C-COR's VP of market development.

Figure 2
Figure 2: Cable operators can target ads by installing more DPI systems farther downstream.
Every DPI system in every node is capable of splicing in a different ad from every other,
effectively creating ad zones. The decision of which ad goes where can be
based on demographic information about the area served.

Some advertisers are said to be considering an ad VOD model in which they give viewers the option to not watch ads, or to get significantly briefer ads. They would charge, for example, $3.00 for a program with ads and perhaps $3.50 for the same program without ads or with shorter ads.

Whether an operator opts for the interactive approach or the decision manager approach, if an operator has already migrated to digital with Gigabit Ethernet transport, the cost of implementing advertising VOD is "at the noise level," according to C-COR's Boland.

A potential hitch is a lack of standards for formats and metrics. On the latter, "a view from SeaChange/Comcast might be defined differently than a view from Concurrent/Cox," says Terri Swartz, director of advanced advertising for SeaChange.

There is no set protocol for what a playlist looks like, just as there's also no agreement on data interchange formats, a potential headache if advertisers want to aggregate information from multiple operators using different equipment vendors.

"Another key concern is ad skipping," says Tom Kennedy, senior director of marketing for Broadbus Technologies, which has a DRAM-based VOD server that can also act as a splicer. "Will operators allow ad skipping or not?"

Whether any of that hinders the proliferation of ad VOD will depend on how eager advertisers are to take advantage of the increased effectiveness of ad VOD.

One other issue: In the broadcast model, it's contractually clear who gets allotted which ad slots, under what circumstances. No such agreements exist for ad VOD.

"The technology is there," says SeaChange's Swartz. "We can do targeting and we can measure the audience. Now we just need the MSOs, agencies, and content providers to agree on how to split up the proceeds." Getting those agreements is just a matter of time, Swartz adds, and won't hold back the introduction of ad VOD.

When talking about targeted advertising, consumer privacy is always an issue, but the industry believes it has very clear rules about assuring viewer privacy, and any number of ways to conform to them.

Of course, there might be a dark side to targeted advertising. What if everyone watching the Super Bowl gets different commercials—what will everyone talk about the next day? Worse, what if you get so demographically typecast that no matter what channel you flip to, you always get the same commercial?