Comcast Revamps Set-Top Data As NBCU Audience Platform

by Wayne Friedman , May 15, 2015

Comcast Corp. long-protected set-top cable subscriber box data -- highly desired by many TV research services -- is getting another release, this time as NBCUniversal’s new audience/data platform.

On Thursday night, Linda Yaccarino, president of advertising sales of NBCUniversal, in speaking at the NBCU Cable Entertainment upfront presentation, said NBC’s Audience Targeting Platform will now get anonymous set-top box data from Comcast’s 20+ million cable subscribers.

NBCU launched Audience Targeting Platform in January, initially to be driven by third-party data sources -- from Acxiom and Experian -- with set-top box viewing data.

But not from Comcast’s set-top box subscribers. Many TV providers want to keep set-top data proprietary; almost all TV providers also say they worry about releasing data in an effort to protect viewers’ privacy.

This past January, NBC said its ATP would include some of NBC premium TV inventory -- about 30% would through the ATP system and offered to clients. ATP recommends TV commercial inventory allocation to TV marketers.

In 2014, Comcast did a somewhat similar move with NBCUniversal for another product. Then ‘NBCU+ Powered by Comcast’ was launched -- a platform that allows marketers to match their own consumer data, and third-party consumer data, with anonymized Comcast 20 million plus set-top-box subscriber data.

All this is to help marketers optimize national TV campaigns across the NBCUniversal networks on linear television and target them at the household level on video on demand inventory.

Comcast recently abandoned its deal to acquire the second-largest cable company, Time Warner, due to increased Federal agency scrutiny.

audience targeting, big data, data management, television, tv

8 comments about "Comcast Revamps Set-Top Data As NBCU Audience Platform".

Check to receive email when comments are posted.

Ed Papazian from Media Dynamics Inc, May 15, 2015 at 4:07 p.m.
Let's hope that users of this data realize that all they are getting is tune in ratings melded with other information to determine how well various TV shows "target" product buyers or users. Just because a household tune-in tally seems to note a high "index" of consumption or purchase for Show A's "audience" versus Show B that does not mean that the human consumer or decision maker in the household was the one who was watching. Often, this is not the case, which is why advertisers stopped using TV home ratings half a century ago.
Reply

Bill Harvey from Bill Harvey Consulting, May 15, 2015 at 5:05 p.m.

Ed, the screening out of (or downweighting of) choices that don't have the right sex/age group in the audience happens when the agency activator combines the Nielsen and TiVo Research (for example) insights. Vehicle selection therefore reflects, cost, currency rating against sex/age buying target, and the skew to or away from the chosen purchaser target, at the point of execution. We agree that substitution of the people data currency with exclusively set top box data is not the right idea at all.

Ed Papazian from Media Dynamics Inc, May 15, 2015 at 6 p.m.

Bill, I'm a bit puzzled by your comment regarding TiVo and Nielsen data and viewer demos. If Nielsen is measuring viewing by age, sex, etc. why would this be combined with TiVo panel data? Are you saying that TiVo's panel is supplying an adjustment factor---say viewers per home ratios---which is used to combine the set-top-box tune-in ratings with Nielsen's peoplemeter findings?In other words, are you using TiVo data to estimate viewer ratings from the cable homes, then marrying this, via indexing, to Nielsen's ratings. I wasn't aware that TiVo, necessarily knows who is watching when it captures TV activity?Can you clarify, please? Thanks.

George Shababb from SQAD LLC, May 16, 2015 at 5:32 a.m.

Two comments that I would offer:

1) As Erwin Ephron so often reminded us ratings whether calculated on a persons basis or household basis are nothing more than a measure of an "opportunity to see." So even with a persons rating we really don't know whether "the human consumer or decision maker in the household was the one who was watching" a paricular ad.

2) So if one accepts that the choice of media placement always comes down to "opportunity to see", it seems to me the question is, is it better for a TV advertising campaign to be based on broad based age/sex demos (which may or may not accurately reflect consumer interest in a particular product) or product purchase behavior? Mirroring the demonstrated succes of targeting online ads, it would appear that the TV industry is increasingly adopting a data driven behavioral approach over traditional age/sex demos.

Ed Papazian from Media Dynamics Inc, May 16, 2015 at 6:13 a.m.

George, just to clarify my own remarks about set usage versus people viewing data, the two kinds of data often produce quite different estimates of demographic audience profiles. This is largely because of the different number of residents per home. For example, overall, you have about 2.6 persons aged 2+ per household, but in younger homes with kids this figure rises to 4.0 or higher, plus there are many more visitors---any and all of whom can be using one of the home's TV sets. In an older home, without kids, your typical number of residents is 1-2, with many being single person households. Obviously, when such homes use their set you can be pretty sure who is watching. The same thing applies to income. Because so many older homes are also low income, they tend to have fewer persons per household, In contrast, upscale homes tend to have more residents than the norm.

The distinctions I have described, above, cause a seeming contradiction in the findings of set usage versus "people" studies. Typically, set usage studies tell you that younger, larger and upscale households use their sets to a far greater extent than older and lower income homes. However, people studies tell us that older persons and those with low incomes outview their younger, upscale counterparts by a big margin. So if an advertiser were using set-top-box set usage ratings as an indicator of a TV show's viewer profile, it is quite possible that programs that are thought to target young adults are actually sub par in this regard---someone else is often watching----while other shows, which index low on set usage, actually perform better---because in such homes it actually is the younger adult who was attending the set.

Ed Papazian from Media Dynamics Inc, May 16, 2015 at 6:29 a.m.

George, one final comment about targeting. I believe that all of us agree that using age/sex demos in TV "targeting" is not the way to go. What we need to understand is that in reality, "demos" like 18-49 or 25-54 are used primarily as a basis of negotiation between time buyer and seller, providing both with a single "currency" for audience guarantees. This is especially true for upfront buys which are mainly "corporate" in nature, not brand by brand operations. In effect, the buyers are buying gross rating point tonnage, as if the corporation's 10-15-20 brands were one big brand. To do this, they need a single number---a "currency" ----so they can negotiate the audience guarantees that are demanded by the client. If you had a different, and more marketing-relevant targeting definition for each brand you would have no way to make a corporate buy. It would be chaos and the networks and cable channels wouldn't guarantee audience delivery on a myriad of varying product usage indicators anyway.

Until a major ovehaul of planning, buying and selling methods takes place, more refined ways of targeting, which everyone wants, are best suited for allocating the fruits of an upfront buy to individual brands--after the fact. Otherwise, you go to a brand by brand approach to buying upfront, which changes the whole system and puts the bargaining power of the buyers relative to the sellers regarding CPM pricing and the ability to get acess to "premium content at grave risk.

Stanley Federman from Telmar, May 18, 2015 at 3:01 p.m.

Hello Bill, Ed and George, as your comments have expanded the scope of the discussion to the limitations of CPMs in general, it is worth noting that Telmar’s MultiBasing technique allows TV planners and buyers to index product performance of each TV show through on the fly integration of TV ratings and Product User data from a user’s own selection of respondent data sources. The technique is a fast and simple way to augment the currency rating alone that is so necessary for that “single number currency” for negotiating audience guarantees.

Even Erwin thought it worth looking into this technique when he said in his 2005 article, The CPM Below that MultiBasing “may provide tangible proof of the unique value of programs hidden right in the data we use.” See Erwin’s more detailed description at http://bit.ly/1AgEz2Z .

Of course this begs the question of the value of set top box data but maybe Comcast and other cable providers could add facial recognition technology to the boxes so we can use these data at the demo level and MultiBase from there.

Ed Papazian from Media Dynamics Inc, May 18, 2015 at 4:48 p.m.

@Stanley, I'vr been saying for some time now that a good deal of the required information is already available if anyone would bother to look at the data. For example, I'm sure that your programs allow a client to select MRI's doublebase sample, if desired, to index product buying or usage across numerous TV shows and/or genres, based on single source data---no attribution needed. Yes, such data is based on peoples' responses not electronic monitoring, and some care is required to meld the findings with Nielsen's ratings, but the scope of the MRI survey is extremely broad, covering all sorts of ad categories, not just cars, credit cards, and packaged goods. Even if we distrust human responses, do we really believe that people will lie about what brand ( s ) they use/buy or how frequently the use a product?

I'm not taking sides about the accuracy of people surveys as opposed to "big data" sources which use actual purchase information or electronically scanned records to define product usage or purchase. But wouldn't it be interesting if someone, who had access to both types of information for a number of products/services, did a comparison of their findings. Do you usually get the same answer or are the "big data" sources coming up with something quite different----and, presumably, a more accurate result? Inquiring minds want to know.