BBC Worldwide's David Boyle uses Facial Recognition to Study Audiences

So let's start with a bit about your background. You started out in retail analytics, is that right?

Yeah, and lots of different corporate environments. Banking a little bit, consulting a little bit. Yeah. Pretty far from the entertainment world where I found myself now.

And how did you go from there to where you are now?

Well, I think one big switch was I realized I wanted to make the world a better place. So that led me into philanthropy and then politics. And that was when I think I really started to care very, very deeply about the outcome of my. Instead of just doing good analytics, it used to be that just good data science was fun and that's all that mattered. And you're trying to change the world, obviously you have to achieve something with your work. So that's when I really started to focus on how people receive data science, how people use data science and how you make it most effective in influencing decision making, which is what I'm currently obsessed with.

Right. And you're working on a data insight team at the BBC, is that correct?

Yeah, our job is to look at data across all different aspects of the BBC. So mainly television content, but also a little bit of online, a little bit of direct to consumer digital, mainly broadcast tele though all around the world. And our show is broadcast in a couple of hundred countries. So we have a whole range of different data sets and our job is to help the organization. There's lots of different ways we can do that. Lots of different data sets and trying to find the biggest and the best.

Can you talk a little bit more about like the types of data that you gather and like specific outcomes that you're looking for?

Yeah, well, I think our shows are the heart of what we do, our TV shows, and so they're our kind of unit of analysis. So some questions are how well are they performing out in the wild, which countries are they popular in? But the more interesting questions also which countries are they not very popular in? But you know what, there is some demand. If we could find a different route to market or market the show differently, then we could unlock some of that demand. So where's hidden demand for our shows? So a little bit of measurement, a little bit of opportunity analysis, understanding audiences, that kind of thing, but around specific TV shows usually.

Right, right. And you've also started using a facial recognition software to do some testing. Can you talk a little bit about that program?

Yeah, well, we launched a program where we would put TV shows in front of people in research environments to understand the types of people who are interested in different shows and the Kind of reasons they were interested in different shows, which is a much better way of understanding it than seeing the show's performance in the wild controlled environment. But we integrated this facial recognition technology to test the hypothesis, which is that people aren't as truthful as they might be in research. So the research alone was really helpful. It told us lots of things we didn't know. But when you add facial recognition in as well, you get to understand that there are some shows that people tell you they're not that interested in, but boy, do they really engage with the show. Boy, do they. Emotions tell a different story. So guilty pleasures, or dark horses we call them.

Right. So people will say that they don't like a show, but then your software will say, oh yes, you do like the show.

If you giggle that much or you pull that face and you do that all the way through the clip, we're doing something right. So you might not want to tell us that you like the show, but there's something going on, there's something that excites you about that show.

And so why do you think that is? And from a data analysis standpoint, what does that mean?

Well, I mean, some shows are clearly exciting off the bat. So if you take something like Doctor who, one of our big shows, you know, people tell us that they love it, their face tells the same story, very exciting. But then there are some shows trying to characterize them. I mean, the guilty pleasures is the best way to describe it, you know, where you're not, you're not going to be proud of spending your time watching, you know, this type of content, but boy, is it super engaging for you. I have to say the BBC doesn't produce a lot of that type of content. Right, right, right, right.

Well, there's some guilty pleasure stuff is alrigh. Cop shows and stuff for me would be my guilty pleasure.

I like that. Thank you.

So, shifting gears just a little bit, you've got a talk where you take a look at the use of data in different cultural creative industries like book publishing and music, and you use a battleground analogy. Now, why do you think it has been such a block to, to embrace data in creative industries?

Yeah, well, I think the battleground analogy works if you're in the data science world. I think many data scientists or data people feel like they want the entertainment industries to use data more, feel like they have answers that they could be using and that they're not, and feel like it's a bit of a fight to get attention and to get people to Use data to make decisions. And sure enough, I argue in my talk that actually the decisions at the heart of the creative process, which artists do I sign in a record label, for example, which authors do I publish in a book publisher? Are largely driven by gut instinct or skills and judgment and not very influenced by data. But I also argue in my talk that it's more about how you introduce data to an organization. It's more about how you respect skills and judgment as part of the decision making process. And it's not really a war. It's actually more about finding ways to work together. And there are lots of mutually beneficial ways to work together. Data can be a really important part of even very creative decisions. Like which artists do I sign without stomping on the toes of creativity, without crushing creativity, and channel it back to like an average audience, let's say. So it can be done, but it's often the two sides talk different languages and end up not really communicating.

Right. And so what do these industries, especially like book publishing and music, is on fire right now with the whole data scene. What do they need to do going forward to survive or even to exist in the future?

Well, I mean, the first thing for the music industry, let's say, is I think understanding the different types of consumer that are out there. You know, this streaming revolution is wonderful if you're super digitally engaged and, you know, passionate about music. If you're a very mainstream consumer, then that's probably not very interesting to use. Spotify or even Apple music are not mainstream consumer propositions. They're not totally niche either. They're not tiny, but they're not mainstream consumer propositions. They haven't reached and won't for the foreseeable future reach kind of mainstream adoption levels. So for those people, the radio and frankly, YouTube is like a pretty mainstream way of consuming music. And then today we heard from speakers about like, super engaging methods of listening to music where you can play individual parts of songs and mix and match, you know, elements of songs together. Well, that's wonderful also for a very specific audience. So I think, you know, not getting caught up in, let's say, the streaming revolution and thinking about the whole of the ecosystem, the different products and services required to fill that ecosystem. Putting all your eggs in a streaming basket will be good for a certain type of audience, but you'll miss the vast majority. So different types of consumers, different needs is really important in that industry.

And it's interesting too, from an artist perspective. Like, the artists today have a lot more access to data than they've ever had before. And the different interaction sets and the, what am I trying to say? The battleground would be a good analogy there too, between the artists and the music labels.

Yeah, absolutely. I think the key with the data from an artist's point of view as well as for the label's point of view, is to ask what decisions that they need to make that the data can then help with. So for an artist, you've got questions like where should I go on tour maybe? And then design data systems or processes or visualizations that help with that specific question. So the problem of the industry in general is there's so much data swimming around these days, by the way. Wasn't a problem when I was in the music industry just a few years ago, but nowadays there's so much data flowing around, you need to channel it into very specific use cases. Where should I go on tour? Is a great business question. There's two or three different data sets you probably need to bring together to answer that question. You need to visualize it in a very specific way. For example, you need to show like share of listeners in each zip code that are a fan of your music, not absolute numbers. Or you end up just going to the big cities again to maybe less big places with a higher penetration of your music. So specific visualizations and analysis, specific kind of ways of looking at that data. But in the music industry as well as books as well as tv, for me, one of the key lessons is to focus on specific business questions and build a data tool that helps to answer that question in a very easy way so that relatively unsophisticated data users can look at it and pretty quickly draw the right conclusions. And it would be difficult for them to draw the wrong conclusions. I think most data tools actually fail that simple test, so.

Right, right. Well, to close our conversation today, I want to shift to a broad, very personal question. What people or projects are you following? What are you finding personally fascinating right now?

Okay, interesting. Well, there's one thing that comes to mind quickly. There's a data science organization called Para Analytics. And by way of background in the TV industry, measurement of TV consumption is a multibillion dollar industry. Nielsen ratings in the us, the equivalent in countries all around the world. But it's the key data source on which advertising and performance is measured. But as people switch to different methods of consumption, the digital age, Netflix on demand pvrs. After seven days piracy, then increasingly less and less TV consumption is actually being captured by this gold standard of measurement. It's becoming less and less relevant to us as an industry. I think there's a parallel here with other industries as well, by the way. And what Parrot have done, instead of saying we're going to try and fix it like so many other companies have tried and failed, they said we're going to work around it, we're going to gather other data. There's a universally available for all shows globally and we're going to use that to infer demand for shows instead of trying to deal with this mishmash of data that's on different bases and different availability in different countries. So they take piracy data by city, by country, around the world, by day, they take social media engagement and they cleverly swirl that together in a data science soup to come back to you with underlying demand for a piece of content in a certain country. This incredibly simple output from very complex data and it bypasses this horrible multi billion dollar mess of a measurement industry and gives us actionable data we can use every single day. So it's both great at solving a business problem, great data science and it's delivered and packaged and visualized in a really neat way. There's lots of lessons there for lots of other data science startups, I think.

Well, that's going to be a lot more important with everybody. The core. All the cord cutting happening as well. Is that becoming a more of a mainstream sort of thing?

Yeah, it's going to become more and more important as well. It's pretty far from mainstream, but it's getting there. Well, but that's the problem is the niche activities like that. Cord cutting is a pretty niche activity, but it gets so much press attention that the industry ends up focusing on it. And it's a. You have to constantly remind people. 98% of people are not doing that.

Right.

Yeah. And even video on demand services like Netflix and Amazon, they're at something like 30, 40% penetration. Even in the US that's 60% of people have no real interest in watching Netflix. They just want to watch linear program television. So again, to the point about music industry, you have to understand who all the different segments of the audience are and what each of them want. Some people want Netflix and which is wonderful, let's super serve them. Some people just want to watch schedule telly. We've got to make sure that we sustain and deliver great services for those people as well. It's a complex and fragmented world, unfortunately.

Right, right. Well, that's been fantastic. Thank you so much for talking with me today, David.

Good to.