• by Joe Mandese  @mp_joemandese, May 1, 2023

MediaPost: What has been the impact of the deprecation of Big Data, and how are you adjusting for it?

Karthik Rao: As it relates to accreditation, I just want to reinforce that we have been steadfast in our commitment that this is the right process. It’s an industry body established over decades. They have very high standards.

The audit process is brutal: 25,000 hours. This is not easy stuff. It’s not a joke. It’s a fact -- 25,000 hours of auditing and it ultimately goes to a vote.

You have to ask the question, which I guess most of our competitors do, which is why bother? The reason we bother is because there has got to be trust and transparency and high standards for audience measurement. We take that very seriously.

We have been committed that the MRC process, however rigorous and painful and complex it is. We will do it, because it is the only way to reinforce to buyers and sellers that we stand for something.

We’re not just another random data company that gets a whole bunch of return-path data and turns it into measurement.

MediaPost: Kudos to you. You’ve done that and you deserve the props, but we’re talking about the next stage.

Rao: The next stage is really about the commitment we made to the market -- that we will keep our products up to date with evolving consumer behavior. Appointment-based viewing is going down and people watch anything they want, anytime they want.  And business models are evolving, and we can’t continue to keep the world in silos like digital, CTV or linear.

So our journey -- which we are calling Nielsen One -- is really about bringing these things together. And how do you do that in a world, by the way, where privacy concerns and data-sharing have gotten even more complicated?

It required us to actually create a system that works -- either directly or with an integration, so that you have all this technology like cleanrooms and differential privacy.

These are not buzzwords. These are the only ways in which you can actually operate. We’re working directly beyond the firewall of each of these players.

It doesn’t matter whether it's Meta or Google or even a broadcaster with a streaming service. This is the pattern for how we believe measurement will unfold. We went down this path because we anticipated we can’t be on one player anymore for all forms of identity and demos. That is not a good way for the world to function. This has been a two-and-a-half-year journey for us already.

MediaPost: Your strategy and execution seem absolutely right. You are mirroring the MRC’s standard exactly. My question for you is really about the ecosystem.

Because of things unrelated to what Nielsen is doing (mainly consumer-privacy compliance) the Big Data platforms that have been your partners are -- maybe deprecation isn’t the right word, but come up with a better one for me -- they have reduced the amount of visibility in the data they are passing through to you.

What I’m asking you is: a) how much of an impact has it been?; and b) how are you adjusting for it?; and c) what do you see going forward? Those are the three questions I have for you.

Rao: It has been an impact. The biggest bellwether of this was when Google announced the deprecation of cookies, right? That was probably three-and-a-half years ago. We didn’t wait. We had to adapt. And part of what we had to do was come up with a framework for how we work directly with each player.

The simple problem was, you know, you could work with one player that could give you whatever you want. That day is over.

Now you’ve got to work with each player, and you have to adapt your privacy rulebook and their privacy rulebook and use technology for data-sharing in a completely privacy-compliant way so that data doesn’t leave each other’s building. That’s the concept.

MediaPost: Data cleanrooms -- I get it. Can I ask you a follow-up question to that? For the decades I’ve covered Nielsen, you’ve used these things called “in-tabs” to measure the quality of your sample, even if only wonks and geeks understood them.

Is this the same idea -- that you have a composite of all these Big Data sources and there's some kind of quality score for measuring that?

Rao: It’s less about the idea of an in-tab. There are two components to doing measurement. One is: you need data about exposure -- what was exposed to a human. The second piece is: how do you infer who that person is? Those are the two basic frameworks.

We take in all kinds of exposure data, because there is no privacy issue associated with it. It’s just a log file, if you will. Like: “Hey, at 8:01, this particular thing was delivered to somebody, right?” Then from there, there are a lot of identity systems across all the suppliers or an identity system that exists outside -- like Experian, LiveRamp.

So the magic of doing this ultimately is to connect all of those dots and to prove -- are you resilient enough so that today when you measure something and infer it’s female 25-54 and tomorrow you don’t come back and that it’s the opposite?

It’s your in-tab concept, but it’s more about reliability and consistency of the models. And the only way to do that is that we use our panel to validate. That’s why our panel plays a huge role, because one of the key elements of the panel is that we collect -- observe -- attributes about humans.

That informs this model-based world, because it’s true that people cannot share information about humans as easily as we could have even three years ago.

MediaPost: Is there an analog for an in-tab in this?

Rao: I think it is about good old accuracy. Eventually, you might get into something called a “match accuracy,” because you’ve got multiple identities from different sources, so you might talk about a match accuracy. This is where the industry needs to evolve. But that’s the same concept.

MediaPost: So let me ask you a theoretical question, and give me a back-of-the-envelope answer. Has that theoretical index moved from whatever it was before -- an 80% match or an 83% match to 79%, or something like that?

Rao: I have to assume it has gone down. But it is very recent. And it’s not a coincidence that every company talks about having an identity platform. Everyone has built a stack with identity and their first-party data.

I have to assume because there are so many hops taking place, that it creates a little bit of erosion. And you need some ground truth -- but you can at least get a sense of, “Okay, directionally it’s at least holding to the same level or not." That’s why we’re hellbent on panels not going away.

In a world where everything is virtual with virtual handshakes that we match rate or what not, how do you actually know? That’s why our strategy has always been three-prong.

The panel will continue to play a big role. You do need identity -- because yes, identifiers are going away. And you need all forms of Big Data.

And that’s what we’re continuing to push forward.

MediaPost: So the last two parts of my question were -- now that you’re aware of it, is there anything you’re doing right now to adjust for it? And what do you think will happen going forward?

Rao: I think this is a very important topic. And the only way in the short term to continue to improve it is to get more data partners into the system, because it turns out that when you’re working with this volume of data, more is better.

Eventually, where we all have to get to is accuracy validation -- and that being another standard that needs to be adhered to. It’s not enough to say, “Hey, we measure all this stuff.”

There has to be a standard that says, “What is the accuracy?”

Everyone just wants to use Big Data and let it flow through the system, but it has flaws in the core of the exposure data and it doesn’t have human identifiers.