Philipe Casarotte

Philipe Casarotte is a Director of Product Development for NBCUniversal within the content distribution / digital products division.


When it comes to research, what we're here [at the IIEX conference] for, how does that relate to your role? 

In my role, I don't do research myself, but we consume a lot of research and try to understand consumer trends, especially around media. 

You'll see the trend of cable subscribers declining, but then you'll see people moving more to streaming, but you also see what we call “digital cable providers” getting a lot of subscribers, but it's not necessarily offsetting the change of losing cable subscribers. There's still demand. So a lot of what we try to understand is what is driving that. People cancel cable, and then after a while they miss cable, and then they find a new way to consume what they were looking for. Are people signing up for more technology-driven cable providers just because they have more sports content?

So we're always trying to figure out: is it only about the content offering or is it something beyond the content offering like the user experience on those platforms, and things like that. That's more or less the overarching theme of what we look at. 

And then we try to see what we can do to help enhance content discovery from NBCUniversal in all these different distribution channels. 


What types of sources of information are you tapping in order to get some of the answers, the types of information, you just described?

It will vary. 

There is qualitative research done. 

There's also access to first party data we have access to from our own platforms that we're looking to. 

But when it comes down to people switching services, then we have to tap into third parties where basically you pay for a data set. 

Or, you can tap into credit card data that you can buy to understand what people are canceling. It's a wide variety. 

Some of it is in-house captured (like for more D2C platforms).

Some of it is going to be just purchasing external data sources. 


Something we've heard from a number of different people here [at IIEX] is there are these different cadences and types of research you might do – three buckets. Generally, it sounds like:

• Big projects, something that was planned on for a while. It's a quarterly goal, whatever it might be. You have a clear start, you know what the outcome looks like, you know how you're going to use it. 

• Small stuff, the ad hoc. “Someone on my team asked me a question.” “I have to get smart before I go into a meeting.” 

• An in between: an always on, stay smart on my competitors or be monitoring trends around platform consumption or otherwise.

I'm curious if those three buckets even resonate, and then if you tend to gravitate towards one or the other?


They make sense. For NBC, it will be more of the ongoing things. Especially in trying to understand churn, how people are switching between services, and also content trends. Those are coming out all the time across different media companies. So from what I see, we would be more under that specific bucket. 

Of course, if you drill down to Philipe – my team’s needs – it might touch all three buckets. But overall as a company, they're always constantly trying to profile viewers. “What are the type of viewers in this platform? What type of viewers are consuming ‘fast’ right now?” (a new model of TV). 

So out of the three you mentioned, I see NBC being more under “ongoing” – always trying to update and understand what's happening now. Because it changes so much, right? With many outlets, it's so fragmented. And then content types, and then people are very complex. It's media. 

Consumption of video is not like buying tennis shoes. With tennis shoes, you can kind of reduce why people are buying to maybe two or three questions: Are they comfortable? Do they look good? Can I afford them? 

But why people consume content or a show is very different, because you might like a very different comedy style than I do. And it changes so much – even from time of day, the mood that you are in. So it's much more complex and demands something ongoing to really try to refine the profiles of people you're trying to tailor your experience to. 


Makes a lot of sense. And when you think about desk research (mining what might be publicly available on the internet or otherwise), does that play a role in your “research stack?” 

A lot of data that is publicly available is just not tapped to, oftentimes. If you had a way to harvest that data and put it together and have something that can help us summarize the body of content (like with AI now – it's really good at doing that)...


There's a big barrier you just described: it can be a lot of work. Is that a barrier for you? Or do you guys still mine it (whatever you can)? 

I don't think it's a barrier, to be able to tap into that.

Now, sometimes, we do need to get a more granular type of data to help prove our point, and that will come from data you have to acquire – it's not out in the public. 

But personally speaking, as a personal point of view, I do think there's a lot of public data, public knowledge, that is just not tapped to because maybe we're not thinking about it or it's just the lack of tools. So can we have the tools to mine that data?! 


When it comes to mining anything publicly available, if there was a wand you could wave to fix one part of it, what might you fix about finding information that you need from anything that's publicly available?  

I come from a software engineering background. Can I have tools that can easily scrape multiple sources and bring those together in one place? So then I can run an LLM model, or whatever – I can feed that into another system that then can summarize that body of data.

So for me, it's more: instead of manually having to go to multiple places, how can we actually more easily scrape multiple sources where we know they have this public data available, and then bring them all together under one umbrella. 


Very helpful. Last question for you. 

When you think about AI in research, whether related to what you just described or not, one word that describes how you feel or your general sentiment towards the future of research with AI?

If it's one word, it's very hard for me.


I'll give you a few more!

The word is good. I don't think it means much, but I can expand. It's good in the sense that I think it can help with the things we've been talking about. It is here already and we are seeing it at the conference – everybody's talking about it, right?

But at the end of the day, the way that I think about research is you're trying to get to the causation as to why people watch our content or why people buy your shoes or why people want to use your health insurance or health service. It's the causation. What caused people to make that decision?

With AI, I'm not saying that you will get to that level, but it can help assist you with getting to a level where you're trying to go deeper and help you be more creative in generating better questions and all of that. 

But ultimately, I don't think AI will replace researchers or people that are good at connecting with other humans, because the translation of what I'm saying, no pun intended, is that you require a human to create empathy with the other human, right? Or have empathy to the level that they're understanding the underlying sentiments and emotions of what they're communicating.

So I know you just asked for one word, but I think it's a “good” thing. BUT I also think research will get better, not because of AI, but because of how people continue to use AI to try to be more human, more empathetic with whoever they're dealing with (whether it's a panelist or whoever you're working with).

That's why I think we have meaningful connections. When you see “oh, this person is actually listening or caring about or trying to understand what I really feel, what I really want to communicate here.”  

Super helpful. That was a lot of great words – better than just a single one.