Future of TV Briefing: A Q&A with NBCUniversal’s chief data officer John Lee

Big Data

NBCUniversal’s first-party data efforts are moving into a new phase, with the number of individual-level NBCU IDs having surpassed its end-of-year goal of 200 million individual-level IDs and the number of attributes associated with each ID surpassing the 1,200 mark.

After unveiling its first-party data platform NBCUnified in January, the Comcast-owned media conglomerate launched the platform into production in May and announced its first deal with a brand to use NBCUnified in September. NBCUniversal has also made a series of hires under chief data officer John Lee: svp of brand engagement Dan Bernard, svp of data product and partnerships Kaitie Coghlan, svp of product strategy and business development enterprise data Nick Illobre, and svp of data science Leah van Zelm.

These appointments were made with an eye toward deepening the platform’s development and broadening its adoption after what Lee described in an interview as “a rapid build phase.”

The interview has been edited for length and clarity.

Do these new hires signify any new phase or development in NBCU’s data efforts?

Some of these new hires are representative of coming out of a rapid build phase. This NBCUnified platform is still relatively new. We probably embarked on this whole enterprise data initiative about maybe less than 18 months ago. We launched the platform, it was live in production early summer, somewhere in that May timeframe. And some of these hires — a Kaitie Coghlan, a Nick Illobre — are representative of our readiness to take that data, commercialize it and bring it to market in support of our clients and actually our internal marketers as well.

In a nutshell, we’re going from this big back-office building of the data pipes to now a fully commercialized, market-facing, customer-supported product suite.

On that point, last month NBCU announced General Motors as the first marketer to sign on to use NBCUnified. What has been the process for getting GM set up? Does the brand need to be sending its data to NBCU to match with NBCU IDs?

It’s really facilitated through an integration between Dentsu’s M1 platform [Editor’s note: GM is a client of Dentsu’s media agency Carat], which is their agency data platform supported by [Dentsu’s data arm] Merkle. What we’ve done is integrate the M1 ID directly through a crosswalk to the NBCU ID to NBCUnified.

The setup is Dentsu works with GM to define that Cadillac audience using whatever sources they have available to them, including Cadillac first-party data. They match that to the NBCU ID through a function that we call consumer match. It’s as described. It’s the ability to match consumer IDs directly to the NBCU ID. And then from there, it just follows essentially the path to the clean room where they have the option of adding the NBCU first-party behavioral data to those audiences for additive insights, for additive segmentation. And then they can activate. In the case of GM, they’re activating in a cross-platform manner, which means they’re optimizing consumer reach across both our data-driven linear and our addressable streaming products.

I just did a video breaking down clean room models between the ID-based approach and the newer model-based approach, in which an advertiser can share the recipe for the audience it wants to reach without matching back to its audience or being limited to matching only against its audience. It sounds like NBCU’s clean room currently is taking the ID-based matching approach. Is there any plan to add on a model-based option?

In the case of some of our first advertisers, the way that the market has matured, the pathway has been through that ID-based approach where these IDs are being anonymized to an anonymous key but are being matched one to one. The model-based approach, we have the capability to do it. 

I don’t want to go off on a tangent, but our clean room approach is we haven’t developed a proprietary clean room technology. We leverage [clean room providers such as Habu, InfoSum and Snowflake] to facilitate making our data accessible. We can take either [an] ID-based approach there or a modeled approach within any of these clean room constructs. We haven’t had somebody ask us on the model side yet, but from a technology and a data science perspective, there’s no reason we couldn’t do it if somebody asked.

Given the history of platform clean rooms like Google’s and Facebook’s, there’s the question from advertisers and agencies when it comes to all clean rooms at this point: Does a clean room connect to a marketer’s own clean room, or does it require companies to provide their data? In the case of NBCU’s clean room, are you connecting to marketers’ clean rooms, or do you require companies to provide their data to NBCU’s clean room?

The way we think about the paradigm is we can set up a joint clean room. If somebody’s a Snowflake user or Habu user — I’m just throwing around names — we can set up a node or a share to that clean room technology. But it’s going to be partitioned. We would never just quote-unquote share our data unfettered into somebody’s clean room. We have to have our own requirements met around how that data is handled or what functions — a segmentation, modeling, attribution — we’ll permit against that data. 

It’s a long way of saying: We will work in an advertiser’s or an agency’s clean room technology. But each instance is going to be its own, and we are going to partner with that advertiser on what are the joint pre-approved use cases and functions that we allow in that.

It feels like the clean room landscape is going to be a very big hub-and-spoke model where a hub is also part of a spoke.

Yeah. We kind of think about [clean room technology] like it’s a database technology, but it’s not. It’s a collaboration technology that allows you to move data without moving data or share without sharing — everyone has their own expression for it, but essentially to be able to bring data together in a very safe and compliant manner, quickly but that also then has a rules layer — like a set of controls — that sits on top of it and a set of functions that you turn on. That’s what these platforms are good at: making that very standardized and making it fast. So that once you’re an expert user of, say, Snowflake, then you could set up dozens of these things very quickly, very easily.

A backbone of NBCUniversal’s first-party data platform is the NBCU ID. When NBCUnified was announced in January, it was announced as having 150 million person-level IDs and 80 million household IDs. What are the latest numbers?

We’ve increased the individual-level match to over 200 [million]. I think the count right now is 204 million individuals. And the household count is somewhere in the 80 [millions]. So we’ve been able to fill in more individuals into the residences and households that we had records for previously. And then there is the number of attributes, and that’s in the 1,200s.

And that’s the depth of those IDs. 

Exactly. Back to the old database analogy, think of it as columns and rows. The rows are 203 million individuals, 80 million households; the [rows] are 1,200 columns wide.

You all had set the goal of 200 million individual-level IDs by 2023. Obviously you’ve reached that. Have you updated that goal?

Our goal is in the 220 to 225 [million range]. But at this point, if you say there’s somewhere between 240 and 260 million adults 18-plus [in the U.S.] — which is all we would ever aggregate in the U.S. — when you start getting into the low 200s, what is that 80 percent of the U.S. Census already keyed in our data spine? 

So it’s a little bit less about adding more and more of those IDs. If you look at an Amazon or a Facebook or a Google customer match, the effective match numbers [for advertisers are] going to be quite similar at this point to NBCU. The real question gets back to how deep are the attributes and how interoperable. 

Adding more consumer IDs, of course we always want more, but it’s now fallen much lower on our priority list. We’re much more focused on the number of clean room integrations that we manage, managing volume through OpenAP and our programmatic partners like The Trade Desk and Yahoo DSP. If I’m going to count stuff, I’m counting the number of advertiser custom audiences and models. That should probably go on the dashboard because that means people are using it, they’re getting value from it and we’re making money.

In an interview earlier this year with Beet.TV, you said, “The future is… First party rich data owners working in a peer to peer networking manner, making their data interoperable through privacy safe means like the clean room.” How would that work? What would that look lie?

I think the predominant model on a go-forward basis is going to be the clean room. It’s not the only way to do it right now. But I do think that’s the future. This reliance on third-party data, third-party IDs that are being traded out in an open auction, open web manner — we clearly don’t think it’s sustainable. It’s our view that efforts to try to prop up that model are doomed to fail.

What we’ve heard

“We’ve all just spent two and a half years communicating over video. It’s become the way we do business now. And so LinkedIn has to be able to have a first-class platform for video.”

— LinkedIn’s Dan Roth on the Digiday Podcast

Netflix’s Q3 2022 earnings report

Netflix is growing again. After two consecutive quarters of sequential subscriber losses, the streaming service added 2.4 million subscribers in the third quarter of 2022, which is more than double the additions Netflix executives had anticipated. Even the company’s subscriber base in the U.S. and Canada grew for the first time this year.

“After a challenging first half, we believe we’re on a path to reaccelerate growth,” Netflix said in a letter to shareholders published on Oct. 18.

The key details:

• 223.1 million subscribers, up 5% year over year
• $7.9 billion in revenue, up 6% year over year
• Gained 2.4 million subscribers in Q3, compared to the company’s projection of gaining 1 million new subscribers in the period
• Added 100,000 subscribers in the U.S. and Canada

Profitability push

Apparently feeling itself coming off the subscriber gains, Netflix took the opportunity to flex on its rivals a bit by touting the company’s profitability in its shareholder letter at a time when rivals like Disney, Warner Bros. Discovery, NBCUniversal and Paramount are losing money in streaming.

“Our competitors are investing heavily to drive subscribers and engagement, but building a large, successful streaming business is hard — we estimate they are all losing money, with combined 2022 operating losses well over $10 billion, vs. Netflix’s $5 to $6 billion annual operating profit,” Netflix wrote.

To be fair, though, it took Netflix’s streaming business a while to consistently generate a positive cash flow, with 2019 marking a low point.

Fourth quarter forecast

Netflix is projecting its subscriber growth to continue in the fourth quarter, which is typically a strong subscriber acquisition period for streaming services. In Q4, the company expects to add 4.5 million subscribers.

However, even that Q4 subscriber projection is somewhat muted. In Q4 2021, Netflix added 8.3 million subscribers. Moreover, this Q4 will see the company rolling out its ad-supported tier, which is supposed to help Netflix to add subscribers. But Netflix doesn’t expect the ad-supported tier’s launch to have much of an immediate impact on subscriber growth.

“While we’re very optimistic about our new advertising business, we don’t expect a material contribution in Q4’22 as we’re launching our Basic with Ads plan intra-quarter and anticipate growing our membership in that plan gradually over time,” the company said.

Numbers to know

$199: Annual price for the Los Angeles Clippers’ new direct-to-consumer streaming service.

42%: Percentage share of survey respondents who use streaming services and don’t have a pay-TV subscription.

-937,000: Decline between January and September this year in the number of U.K. households that pay for at least one streaming subscription.

1,024: Number of original TV show episodes that Netflix released in the third quarter of 2022.

44%: Percentage share of survey respondents who said they have both a smart TV and use a connected TV device, such a streaming dongle.

What we’ve covered

Why LinkedIn is stepping up its original video and audio content ambitions:

• LinkedIn has hired former CNN executive Courtney Coupe as its first head of original programming.
• The hire is meant to help further professionalize the platform’s original video programming, LinkedIn’s editor-in-chief Dan Roth said on the latest Digiday Podcast episode.

Listen to the latest Digiday Podcast episode here.

Creators react to Twitch’s updated revenue share model:

• Twitch is tweaking the sweetheart deals it had cut with its top streamers.
• Twitch streamers are considering asking for viewer donations instead of paid subscriptions and diversifying to other platforms.

Read more about Twitch here.

Netflix unveils its ad-supported product:

• The streaming service’s ad-supported tier will launch in the U.S. on Nov. 3 and cost $6.99 per month.
• Netflix will air pre-roll and mid-roll ads and offer limited targeting and measurement options at the outset.

Read more about Netflix here.

TikTok’s latest ad product rollout aims to make creators’ and advertisers’ lives easier — for the most part:

• TikTok is releasing updates to its branded content collaborations platform.
• The updates include a creator recommendation tool and the ability to work with creators who are not officially part of TikTok’s program.

Read more about TikTok here.

Apple is quietly pushing a TV ad product with media agencies:

• Apple has had exploratory conversations with agencies related to streaming advertising.
• It’s unclear to what extent the conversations are focused on Apple’s streaming service Apple TV+ or its connected TV platform Apple TV.

Read more about Apple here.

What we’re reading

Warner Bros. Discovery’s diversity dilemma:
By laying off non-white executives and sidelining some Latinx programming, the media company appears to be the latest example of how mergers can compromise media companie’’ diversity pushes, according to Daily Beast.

Apple’s NFL demands:
The iPhone maker wants wider-ranging rights to the NFL’s Sunday Ticket package than the league may be willing or able to give, according to CNBC.

TV’s political ad dollar disparity:
Republicans are paying more for TV ads than Democrats, in large part, because Republican candidates have had a hard time raising money so Political Action Committees are picking up the tab without being granted the same pricing candidates would receive, according to The New York Times.

Netflix’s new era:
The dominant streaming service’s struggles this year have taken a toll on its employees’ morale, and the company is trying to address its subscriber losses with a crackdown on password sharing and an upcoming ad-supported tier, according to Bloomberg.