David Boyle for Johgn Deere - BIA 2023 Keynote Seeding the Future

It's my pleasure to introduce our first keynote today, David Boyle. He is the founder of Audience Strategies. David has spent his life at the intersection of data and creativity. He's led the transformation of analytics at companies like EMI Music, the BBC and HarperCollins Publishers, and advises many of the world's largest entertainment and media companies. He's a mathematician by training and passionate about the underlying rigorous of what we do. And an expert at understanding how businesses work by trade, which allows him to craft data products that really make a difference. Ladies and gentlemen, I'd like to have welcome David Boyle to the stage as he tells us about seeding the five bold lessons for transforming insights and analytics capabilities from the creative industries. David.

Welcome.

Thank you so much for such a warm welcome and for having me here today. It's such an honor to come work with such an awesome brand, an iconic brand since my childhood. You make such a difference in the world. Thank you for having me. And thank you also for having me here. For what is my honeymoon this week? I got married on Saturday in the uk. I know honeymoons are traditionally relaxing on the beach, but I'm a big data nerd and so this is one of my big passions. Yes, that is a problem in life.

Okay.

We're here to talk about how analytics transformations can be so much more powerful when creativity is appropriately involved throughout. And I come from a background in creative industries, as you just heard. I've worked with some of the most passionate and creative people in the world. Worked with Snoop Dogg and Katy Perry. I've worked with Hillary Clinton and Tony Blair back in his day. Actually, last time I was in Iowa was for the caucuses in 2008. Side note, it's a good lesson where awesome analytics isn't enough. You have to have a great product. Obama won that election at caucus. A product wins over analytics often. I've been lucky to work with some of the world's best and most loved TV shows, video games, books and luxury brands. I'm going to bring lessons from all of that here today. But I didn't start in the creative industries. I started in the dry and mechanical world of analytical strategy consulting. Using data to value companies and do mergers and acquisitions in all sorts of relatively uninteresting industries. I say mechanical as part of that because I used to think that decision making in companies was a relatively mechanical thing. I used to think that if you got the right data, analyzed it in the right way, and drew the right conclusions, then the right business decisions would follow from that. Makes sense, right? Turns out when you spend lots of time in the creative industries, you realize that's very much not the case. And it turns out that you then look back at what you thought was mechanical decision making and you realize it never was in the first place. There's always creativity involved. So we're going to learn that lesson today as well. Data and creativity go together like yin and yang, and the interaction is what's most important. That theme will run throughout this morning's session. I'm also drawing on experience being part of and at the center of a number of big transformations. I worked on the dot com boom back in 2000 when E Commerce started. I worked on 3G network rollouts when they promised to change the way we communicate and live our lives. I worked on the launch of streaming services in the music industry and in the television industry, which totally disrupted both of those business models. I worked on social media as it became the place where companies need to communicate with their audiences and therefore the place where you need to go to get data on who your audiences are and what it is that they want. Most recently, I'm at the heart of AI helping lots of large companies to understand, particularly text based AI and how transformative it is for their business. So I've seen a lot of disruption in my time as well. I'm going to draw on those lessons today as we as you heard, five big lessons and my goal is about 45 minutes, leaving 15 minutes for Q& A. So get your questions ready and then after this build up for an hour and then we're going to do a joint Q and A in one of the breakout sessions afterwards. So loads of time for question and discussion on this as well. Really looking forward to it. So let's keep this central theme in mind and walk through these five lessons that you can practically apply in each of your different areas to help analytics to be as transformational as possible. The first lesson is all about audiences or customers or consumers and how you can deeply understand their needs in order to build the right products and services for them. And I use the word audiences instead of customers or consumers in my language because I think there's so many valuable lessons when you think about these data points, these customers, these consumers, when you think about them as audiences. In the music industry, we know this very, very well. We don't think of people as buying stuff or consuming stuff. They're individuals with wants and needs and passions and hopes. People fall in love with music not for its functional benefits, but because it helps them to feel or to love or to party or to cry. Music is about serving those deep underlying needs. And this is a lesson we're going to explore. I found myself learning this lesson again just a few years ago when I was responsible for insight and data transformation at this place. This particular photo is a room called the Salon de Parfum, which is a perfume room or something like that in Harrods, a luxury department store in London. It's got a million square foot of retail space. In this single store, the most wealthy and famous people in the world shop there. Also, hundreds of thousands of tourists buying things like teddy bears. So you can simultaneously buy a bottle of champagne for a quarter of a million dollars, or one of my favorites, small girl's dress for $100,000. Or you can buy a souvenir teddy bear and some chocolates for your friends. Best of both worlds. And I went to Harrods under the promise we understand our customers brilliantly. We have fantastic data on all of our customers because almost every pound that's spent goes through the loyalty scheme. And therefore we know exactly who's been buying what. And therefore we understand our customers well. And if only we could unlock the power of that data, then surely we could drive growth. Sounds good. I'm in. Let's do this. But there's a problem. Transaction data often doesn't reveal what you really want to know about your customers. Think about a customer whose transaction records say they've only bought a single teddy bear. What would you think about that person? How would you characterize or classify them? What might you recommend to them as their next purchase? A box of chocolates, maybe. How would you market to them? Maybe you tell them about the upcoming sale. Well, I realize the error of this thinking. Transaction data, prediction, thinking. When I talk to a very wealthy Chinese individual over a very lavish dinner in Shanghai one day, I said to him, have you heard of Harrods? Oh, yes, of course I've heard of Harrods. Have you been to Harrods? Oh, yes, of course I've been to Harrods. Not very communicative, this person. Did you buy anything at Harrods? Of course I did. What did you buy? A teddy bear.

He bought.

He bought a teddy bear. Very exciting. I said, that's it? You only bought a teddy bear? Uber wealthy. Live a life of luxury. There's nothing in the store that excited you? No. So over the course of a dinner, I got to know this person a bit better, and I got to know the data on this person that didn't exist in my transaction database, which Tells me what it is I need to know about this person if I'm going to build the right products and services to really engage them. It turns out this person core need was his family. He'd grown up poor, didn't have a life of luxury, found that in his life, and he wanted his children to grow up that way as well. He wanted them to be welcomed into, to learn about and to be part of this life that he built for himself, but that he wanted his children to have as default. He wanted to give opportunities to his children that he'd never had as an individual. He saw Harrods as a store that sold loads of stuff, didn't achieve the goal he wanted. What we did, after talking to a number of other people who had similar needs, it turns out, is to set up a big program of events, private exclusive events for the uber wealthy where we'd welcome them in and it's almost educational in nature, helping them to understand the background to the products and services and the community and the lifestyle that is the world of luxury. So we moved from, on the surface, selling loads of fancy products to being an organization that builds community and training and welcoming and networking inside a luxury lifestyle. And by the way, we sold a lot of cool products to those people as well. I remember showing my mum, or trying to show my mum this $100,000 children's dress, only to find it wasn't there. I think I hadn't thought it was real, so I asked the sales assistant, where did that dress go? Oh, somebody bought it. I guess it was real then. The lesson being that if you just look at the transaction data, you see a teddy bear purchaser. But if you get to know the real person, their deep underlying needs, you build radically different products and services as a result. I think of this as being more focused on the jobs to be done for people than the products and services that they buy. If we think about hobby of mine, skateboarding, an analogy might be don't focus on things like the carbon skateboard or the Swiss bearings or the hollow trucks. Don't focus on the product and the product features. Focus on the job that person's trying to be done, what they really want, what they want your help achieving, which is feeling really good and having fun and doing awesome things that we call the job to be done. And that job to be done doesn't live in transaction record databases. So we have to find it in other ways. We have to dig deeper into customers and customer data to enhance it, to understand what people really want. If I Might tell you a story from the small farming community where I live part time in Connecticut. There's a man there called Alden. He has, I think probably 10 or 20 acres, you know, hobby farm. He describes it as in your transaction record database. I'm sure after a few generations there's a number of tractors and various other bits and bobs. I'm sure there's truckloads of spare parts because they're all pretty old in your transaction database. But what you don't have is the information on his wants and dreams. And he's actually remarkably similar to our Chinese billionaire. Not in wealth, he's a hobby farmer in Connecticut, but in aspiration. His core goal, his want is for his children to be welcomed into and to feel part of and to be able to succeed in a life of farming. He wants his children to be part of that community and to stay in that community, be welcomed into that community. Remarkable similarities to the needs of our Chinese billionaire, albeit in a very different world. So digging deeper than transaction records and prediction algorithms to really understand our audience, our customers needs, we're going to now apply exactly the same lesson, but internally now to the analytics tools that you're building to help business leaders make decisions. Exactly the same lesson, which is focusing on the jobs they are trying to do. And I'm going to illustrate this lesson with a story from the music industry. David Guetta. Who knows David Guetta? Raise your hand if you do a good portion of the room. I wouldn't say all of the room sounds about right, but a pop star, you would call him these days. When I first met him many moons ago, he would describe him as a French dj. Now he's a global pop star. At the time, he was very popular in France. Household name outside of France. Unless you're really into electronic music, not so much. How can analytics help David Guetta to grow his career? Well, the first thing is understand his real challenge, his job to be done. And it's not the job that a lot of analysts think need to be done. This is a reoccurring problem in business where analysts have loads of data and they think, ha, I could do all this cool stuff with the data. But it turns out not to be a job that decision makers are trying to do in the music industry. Lots of analysts want to do things like predict how well a song will do or challenge the notion of the elements of a song and suggest how that song could be better and perform better in the market. These are things that analysts often and want to tackle in the music industry. It turns out almost nobody in the music industry is trying to do that job. Almost nobody wants to use data in that way. And it's not something that's a good use of analytical time. What is a good use of analytical time is talking to artists about their challenges and then using analytics to help them grow their career. So it turned out that for David, he'd had a tremendous period of growth in his career in France mostly, but that was now plateauing and he wanted to know what was next, like, how can he grow from here? That was the job to be done. So we focused analytics on doing that job. We took his audience, or more precisely, the music listening audience around the world, and we clustered it based on the jobs that they wanted to be done by music to find different interesting audience segments. And some of the audience segments that were most impactful for David had names like Nighthawk. What they wanted from music is that it soundtracked their night out. Names like Pop Idol. What they wanted was escapism from the daily grind. Music took them to a place and helped them to escape from their life and Casuals. What they wanted was music as a background to their life. They weren't that excited about music, but they needed it in their life, albeit in the background. These three segments, once we'd identified them based on the jobs to be done, we could then target to grow his career. A really good example, one of many examples around product and marketing would be this album, where the first album, the first disc, the main album, totally geared around this new audience, these pop idols and these casuals. It's a vocal album, unlike his usual electronic music albums. Loads of guest vocalists, different styles, totally appealing to the mainstream new audience that he wanted to reach. Whereas the second album, also available separately, was the electronic album. No Guests, we don't need those. No vocals, we don't need those. Just great electronic music for his core audience, which was the Nighthawks. And through activities like this, and through analytics like this, we're able to look very differently at David's career and to say, huh, that initial period of growth we talked about, that was almost exclusively this one consumer segment, Nighthawks. Whereas the new growth which he's achieved, and hence so many of you knowing his name already today, Pop Idols and Casuals. So he kept hold of his existing audience while growing new audiences. And by focusing on this one job that he wanted us to do, which is help him grow his career, we were able to get to tremendous success with our analytics tooling, helping all sorts of different artists in all Sorts of different countries around the world. It became the standard way to help artists grow their careers. And that was our first ever analytics product inside of EMI Music. And it taught us some lessons that are useful for you both internally in building analytics tools as well as externally thinking about John Deere products. Three questions you should always ask first, who are we for now? Who we were already working with, already selling to, already have a good relationship with? For David, that was Nighthawks at that time? Number two, who's next? And for David at that time, pop, idle and casual. Keep the Nighthawks, lock them down, but also grow. And then who are you not for? Who are you not going to be distracted by waste your time and energy on? And for David at that time, globally, on average, there were eight other audience segments that we just didn't bother with. Don't worry about them, don't worry about their media channels, don't worry about their product preferences. We'll worry about that later. And again, I think the same lesson applies to end products for consumers as it does for you building analytics tools internally in your organization. Who are you for now? Next, and who are you not for? By building just one set of analytics tools that did one job, and then slowly, job by job, building out more and more analytics tooling, we built the most profitable of the four major record labels. And we ended up using this analytics tooling to totally transform the business in every single part of the business. But we did it very simply. Well, in the same way that you eat an elephant, how do you eat an elephant? One bite at a time, how do you build analytics tooling that makes a difference in an organization? One job at a time is the same lesson. And that's how we totally transform DMI music. Focusing on jobs to be done, digging deeper into real underlying needs of our stakeholders and solving their problems for them. Same lesson we learned about audiences, but applied internally. Lesson three. Here we're going to talk about machine learning models and production models of all different kinds. And in the music industry, this opportunity is much clearer than it might be in other industries. In the music industry, what's very, very clear is that decision makers, artists, or even managers or people who work in labels are, shall we say, not that familiar with machine learning models. They're not going to be interested in tables of coefficients, they're not that interested in tables of models. They're not that interested in graphs or numbers at all. What they want are ideas and lessons and insights they can act on. Like for David Guetta, there were no numbers in that example. Just simple insights he could latch onto. Although, by the way, David Guetta, total nerd, fascinated by graphs. He kept elbowing me out the way to get closer to the graphs on the page and then asking questions about them. Total nerd. Most music people, not so much, focus on simple insights and lessons that they can carry around in their heads. So when you have lots of task specific models in production, let's think recommending the next part that should go with a parts purchase maybe, or recommending service times or website optimization, I don't know, task specific pricing models maybe that you've built. What's an incredible opportunity is to think about each of those models and say to yourselves, or something, say to your stakeholders in the business, if that model could talk, what questions would you ask it? Let me give you an example of that. How we take numbers and models and we put those to one side a little bit and we tease out from it the essential ideas or the insights that stakeholders and decision makers can carry around in their heads. At the end of the day, numbers are only there to add weight and support to ideas. How do we get to ideas from production models? Well, here's an example of a model, single purpose model that we built in the television industry. I worked at the BBC and we were the biggest producer and seller of TV content outside Hollywood. We would make loads of shows for us for the BBC in the uk, then we'd sell them all around the world. We'd sell them in France and Germany and Spain, sometimes to Netflix and sometimes to local broadcasters. So we had this constant pipeline of TV shows. We need to sell them in each country usually. How much should we sell them for? Are we getting a fair price for our content? No idea. So we built a predictive model which looked at the essence of a TV show, the genome of a TV show. For example, Sherlock, one of our shows. Anybody familiar with Sherlock? Yeah. Such a good show. My God, I wish that would come back. For example, Sherlock. And we broke the show down into what we call its genomic elements. So whether it's mood or genre or themes or plot, characters, locations, all the elements of the show that we thought would help it to resonate or not in individual countries. We build a database database for all of our shows on analyzing their genomes. And we built a predictive model for each country, for example, Italy, for every show to say, how much should this be worth in Italy? And we got great prediction power from the model. We did a reasonable job of saying, no, no, that show's worth much more than you sold it for, huh? You got scammed or good work selling it for that much because you really shouldn't have been able to. So we did a really good job helping with the pricing which was the core purpose of the model. Single purpose model. But we also said what can we learn from this model? And one thing we could learn from it by looking at coefficients amongst other things is what features drive value in different countries. In Italy, for example, shows that contain the genomic elements fear, technology, law and others over indexed were more valuable. Could some sell for more excellent things? Shows that contain themes like romance, people and culture, surprise and business sold for less. The more observant of you in the room might question why romance, which is a famous attribute of Italian culture, doesn't sell well in Italy. The double more observant of you will realize this is British romance being produced out of London and apparently it's not up to Italian standards. So don't try and sell British romance in Italy. What we were able to do was take the single purpose model, tease out these lessons, these simple insights from the model and give them back to the teams. If teams are making a TV show, they were very welcome to use this information, information to help them to think about how it might be as valuable as possible in as many countries as possible. We never asked them to make TV shows that were optimized in that way because it's a creative process. They're very welcome to take into consideration the marketing teams. However, we did ask to absolutely use this information to emphasize the correct marketing elements of each show in each country to maximize its reach, its value. So we went from again, a single purpose production model, task specific model, and we said what strategic insights can it contribute? And we used it in that way. A good friend of mine, Samir, taught me that, persuaded me that this is actually a general lesson. And his rule of thumb is that every task specific model should contribute structure strategic insights. I want you to go away after this and think about all the task specific models or predictions or analytics that you do and ask yourselves, am I teasing out strategic insights for decision makers from this? Can I? Maybe you can find a way. It's an incredibly powerful lesson and for those of you who are worried about crushing creativity by telling or encouraging TV producers to make shows that do or don't emphasize certain genomic elements. My experience of working with some of the most creative people in the world is that great creatives absolutely recognize that at the end of the day, what they're doing is making creative leaps by making content or art. They're absolutely making creative leaps all the time. But the best creators I've worked with also want to know what the lay of the land looks like. They want to know roughly how big the leap is that they're taking. They want to know how high that is, like the risk levels of this leap. They want to map the terrain so they can understand where they're from, where they're heading to, what they're going to be landing on. And they want to understand the headwinds that are against them in this leap. Doesn't mean they shouldn't take the leap. But the smartest creators use that data wisely. David Guetta, for example, I sat with him in Paris. I made a mistake as an analyst. We've all made this mistake. We're looking at different genres of music and popularity of different genres. And I laugh at one genre. 90s music is a genre which English people think absolutely sucks. We had some terrible boy bands in the 90s, so 90s music has a bad rep in England, particularly, particularly compared to the rest of the world. I laughed at that, saying, don't go anywhere near 90s music in the UK as a genre. And David had this big smile on his face. I said, why are you smiling? Something's going on. And he looked at it and said, I'm going to bring that back. He saw the massive headwind against 90s music and thought, creatively, I can find a way to go against that headwind to take a creative leap and make that work. The best creators I worked with with aren't driven by data to produce art that's generic and that goes with the flow. They take creative leaps, but they want to know how the landscape looks. And by the way, I think that lesson about creative leaps, while it absolutely applies in art and culture, I think you'll find it applies in every business decision that gets made. There's some element of creativity in products and services that you produce, some element of art there as well. We just need to acknowledge it and celebrate it in the way that it's explicit in the music industry. So Samir's Law, as we might call it. Every task specific model should contribute strategic insights to the business. It's a good challenge for you. We can't do five lessons without talking about the hottest topic in the world right now, AI. I'm going to give you just one perspective on it, one perspective on AI. And let's talk about language models first and bit of background. So this is how I think about language models. This might be helpful. First, they've ingested vast amounts of knowledge from an incredible array of different subjects. They're able to then use that knowledge in really interesting ways by finding pieces of knowledge you never otherwise would would have thought about using and combining it together to give you novel insights. I think of language models with a macro tool as being unlocking the analytics of knowledge. I think of them as doing the same job that SQL commands do for data. Maybe in the world of numbers, we're very used to sorting, grouping, amplifying by multiplying or dividing numbers. We're very, very used to that. Now we can do exactly the same thing with ideas or with knowledge. Group it, sort it, segment it, amplify it, make these ideas bolder, reduce it, divide it by something, make these ideas simpler or smaller or less bold. We can do the same analytics we're very familiar with doing on numbers, we can can now do it on knowledge. Incredibly exciting. That's my framing for language models. Now let's think about your stakeholders who you're helping to make decisions. Business decision makers. I think of business decision making as being a bit like a journey that people are going on. And our job to help them along that journey is to give them. Traditionally what we do is to give them guidebooks to, to help them. And if you think about dashboards or reports or analytics, they're somewhat like guidebooks for a decision making journey. Relatively strategic, quite high level, maybe a little interactive, not that much. Often pretty much tailored to one set of needs, not really customizable to too many different sets of needs. Often they're a bit like guidebooks. And I think what language models unlocks is the ability to go from giving people relatively static guidebooks to stories that are very much personalized to their tastes. In the same way that a bedtime story can meander and wonder according to your specific needs, the language can be adapted to meet your needs. And if you want to go off on a tangent, you absolutely can. And explore specific subsets and dig in and dive in in the same way that bedtime stories play that role. So analytics and reporting can do that now with language models attached to them? It can and should. So in the old world, what we did was to produce dashboards or guidebooks for our business decision makers. And in the new world, we can offer them personalized tour guides in the form of language models. This is a good friend of mine, Mac. He's a personal tour guide in Tokyo. Anybody here from Tokyo? Not a lot of farming in Tokyo, I guess. He's the best. He's actuari by trade, mathematician, statistician, investment banker for a period and now a tour guide of Tokyo. And unlike a guidebook with Mac, if you have any wants and needs during your day, any personalizations you require to the narrative, any thoughts or ideas about the kind of things you want to explore or dig into, he's there at your side to help you to do that. And that's exactly the role of language models when they sit on top of reporting. If any stakeholder wants it simplified or dumbed down or explained more clearly, or implications work through or subsets of data analyzed, they can very, very, very simply do that. Now, totally transformational for the a role of insight reporting, dashboarding and guiding. Which brings us to lesson five. Then we're going to reflect on the nature of creativity and then we're going to have probably 20 minutes of questions. So I hope you've got them ready. See people at the back there lining up for questions. Hopefully I see the mic at the end. Lesson five is the importance of moving on in business models. And I talked at the start about the business model disruption I've been part of in my time. Let's talk about the music industry business model disruption as an illustration of the need to move on. This is quite a difficult topic, I think. Here we're looking at music industry revenue in the US over time up to about the year 2000. We'll get into the colors in a minute, but for now, what an incredible story. First of all, wow, then it was not a very good story for about a decade, which was when I decided, hey, I'm going to go work in the music industry. Never mind the parties of the late 90s and the extravagance and the stories you hear. No, I'm going to join an industry after a decade of solid year on year decline. Yeah, that's fun. That's rock and roll. So I joined the music industry in about 2010 at the bottom, what looks like hopefully the bottom of a declining market. Let's narrate this story a little bit. Of course, you've got vinyl making up a lot of the industry that goes away. You've got cassettes making up a lot of the industry that then goes away. Then CDs. You've got downloads which were never as big and we're already declining on this graph because there's a little bit of streaming in there, but not very much. Now I inherit an analytics team who not only doing audience analytics like the David Guetta stuff we saw earlier, but also supply chain analytics, pricing optimization for CDs. And if you think about CDs, it's a very large number of SKUs and very sparsely populated on shelves. They have one or two of each album on a shelf in a music store. Supply chains are not that fun. And so pricing and supply chain optimization for that, forecasting consumer demand, really interesting problem. Really interesting problem. And so I inherit this team. We're very good at that. Probably one of the best in the industry at that, I would say. But even though CDs more than half of our revenue at this point in time, and managing that in a declining market is incredibly important because cost is a factor as well. Even though we're talking about more than half of our revenue, we took the really difficult decision to reallocate quite a lot and then all of that resource to have them work on that small green section at the bottom down there. Streaming, which was just starting. Now, streaming is an incredibly controversial topic in the music industry. We were talking about it earlier this morning over there. It was even more controversial at this time because the CDs dying, the revenue that's paying for what parties were left is going away and. And what's coming is pennies. So we're giving up pounds for pennies. Incredibly controversial artists are holding out their catalogs from streaming services. Managers of artists are furious. They're not getting on board with this new technology. It's not obvious what the streaming services are going to do, whether they're going to charge more or less or free or. None of that's obvious. Spotify wasn't in the US at that point even. We took our analytical resources from the core of our business, from CDs, and we applied it to streaming. We had the absolute best understanding of what was really going on in the streaming market. We could see early on that it was bringing new users in, it was exposing them to new music, helping artists to get discovered, and it was growing the piece, albeit in a declining market, which you can't see on aggregate. With clever statistical tricks and experiments, you can see it's growing the pie, actually the declining market, it's growing the market size. We were able, using that analytical resource, to convince some of the biggest artists in the world and their managers, who make quite a lot of the business decisions. Hearts and minds are important here. We're able to convince them that they should get on board with this and support this. We're able to convince the rest of the business, who's reliant on CV income, to get on board with this and support this initiative and we were able to work with all the streaming players, Spotify and Apple, et cetera. Because we had more analysts looking at this problem in cleverer ways than anyone else did. We were able to partner with them and build tremendous relationships with them. What happened next is, I think, relatively well known to everybody, but it's worth looking at it anyway. Streaming is all of the green stuff turned out to be a good idea. Excellent. It looked bad. If that was the wrong call, right turned out to be a good idea. The music industry in the US in aggregate is bigger than ever. Many say that money is not going to the pockets of artists. It's a distribution problem, not a macroeconomic problem. And I would like to think that if we hadn't diverted resources from the core of our business, from CDs to streaming, a lot of the major artists wouldn't have been on there, the business wouldn't have been behind it, the platforms would have made less smart decisions on pricing and distribution and marketing. And we might not have this picture in front of us right now. The lesson being, if we're here to help our business leaders to make decisions, and if decision making is like going on a journey, different decisions to new lands of opportunity, then our job sometimes is to help them to leave the old world behind and take resources from old established businesses in order to grow the new ones. Sometimes you can't have that new growth without taking resources away from the old world in order to do it. Really tough decision. Very controversial. Paid off for us, I think, is a really important lesson. If I try then to bring all of that together into the macro theme as I see it, the macro theme, what have I learned in my time in these industries and with these transformations and these challenges? For me, it comes back to a phrase I call informed creativity. And I think that's data and creativity interacting together with each other. And that's clear in the music industry and in creative industries. I think is as true in your industry and your decisions as it is anywhere else. It just needs recognizing, celebrating, and optimizing for so informed creativity. The combination of creativity and data together and good news, companies that harness creativity and data in tandem grow twice as fast as companies that don't. That's not me saying that, that's McKinsey. So maybe I'm not totally crazy. And if we reflect on some of the lessons, I think we see it throughout. At the start of my career, I thought that business decision making was very mechanical. The right data, the right analytics, the right decision gets made. No, I don't think that's ever been the case. It just looked that way for a while. Wasn't a helpful frame. Creative industries help me to realize that and help me to optimize for that. Where people do not make decisions in rational, clear ways. And while transaction data, loyalty schemes and purchase data are incredibly important for lots of things, who's bought a teddy bear and who hasn't, they're actually not going to give you what you really need, which is deep underlying needs. In the same way that in the music industry, people buy music and engage with music for deep underlying needs. Our farmer Olden Farms for exactly the same reason. Something much deeper than parts or machines. So digging deeper is such an important part. What we must always do versus taking data at face value and assuming it tells us the full story. Think about David Guetter and growing his career as well. Then think about task specific models and challenge yourselves. Are they contributing strategic insights to the business? Ask your stakeholders what questions would they ask these task specific models if they could. Our job is to help business leaders to make decisions. That's like going on a journey. Our job is writing guidebooks sometimes. But a much better way to help business leaders to make decisions going forward is to give them personal tour guides like my friend Mac or using language. Models and journeys require reaching new destinations, but also in doing so, leaving the old world behind just like we left CD analytics behind. Very painful decision without which we couldn't have had the new world of a bigger music market and streaming revenue. I think all of that. The common theme for me is informed creativity, to which I hope you're inspired but also somewhat challenged. And I also hope you've got lots of good questions for me. Thank you very much. I think we've got about 15 minutes for questions. I think so. I understand there's mics at the back. I can just about see people. I really do hope you've got some. I can see some people moving.

Go ahead. All right, so I think I speak for a lot of us here when I say thank you for presenting on this really great information here. We take a lot of our analytics to make decisions internally.

Right.

So we have internal customers we're trying to make decisions for. I wanted to ask if you had any guidance or tips on how we can tell what our audience internally wants to do. Because a lot of what you've been talking about has been, you know, market research and sectioning of an audience. Sometimes we don't have an audience. We have a stakeholder that wants to make a staffing decision or wants to try and make production in their warehouses. How do we drive that using this framework?

That's such a good question. And I think you hit the principles here, but not the methods. So. So let me first give you an example of when I totally screwed that up, and then I'll tell you how I would have fixed it. In the music industry, one of our stakeholders, actually the person who signed Coldplay as a band, he asked me to test some music from an upcoming band because he wanted to know what the first single should be. So I tested the music with loads of audiences. One song stood out as a clearly more popular than all the others. And I went back saying, hey, this is your first single. He looked at me and said, well, that's not the first single. Then I said, what? It's clearly the right answer. The analytics say it's clearly the right answer. And he said, we're not trying to get. What we're trying to do is not pick the biggest song. The band need to practice, build up a repertoire, build up a following and gel. So they need the smallest song first and then eventually they'll get to the biggest song. So it's good to know we've got that in our pocket. So I was actually answering the wrong question, completely the wrong question, because I hadn't done my homework. I hadn't asked him what question he's trying to answer. Now. I've left corporate world and I advise businesses now. And so now it's easy for me to go into a business and ask dumb questions. Like, my favorite question is to a decision maker is what keeps you awake at night, what worries you, what stresses you out? Questions like that, like dumb questions that you're kind of supposed to know the answer to when you work inside a business, but you can ask as an external person questions like that give you the real insight into what it is that you can do to help that person, what decisions they're really worrying about. And in my experience internally in businesses like that music example, it was very hard for me to ask those kind of dumb, simple questions that actually unlock the real needs of the decision makers. And in my experience outside businesses as an advisor, it's now really easy to ask it. So I think the lesson is just ask. Ask dumb questions. Don't be embarrassed about really asking the fundamentals what is it you're trying to do, what keeps you awake and like, what stresses you out. And then you can build analytics to help them. But without that, if you just pretend you know or you guess, then I've made Terrible mistakes.

Awesome. Thank you so much. Yeah.

As a follow up to that, one of my questions is what are your favorite methods for digging deeper? You know, a lot of in his room are very adept at their SQL query and pulling data that we already have. The question is, what are your favorite ways to gather that other data?

That's such a good question. I mean, I think that the simple answer from a kind of organizational point of view is that you need a range of different skills involved in analytics and you need like hard skills of data science and data engineering, but you also also need skills that are more akin to qualitative research that understand deeply, like consumer behavior. So in Harrods, looking at transaction data, we were trying to identify bulk purchases. They'd bulk purchase and then resell. We wanted to stop that.

Right.

Because taking our limited supplies and reselling them. And one person had bought some enormous quantity of this incredibly expensive perfume. Oh, it must be a bulk purchaser. And our business expert who really understands consumers said, no, no, that's quite common in the Middle east because a person has five palaces, each palace has five bedrooms, so they bulk buy perfume. So there's one in every bedroom in every palace. Okay, cool. But without somebody who really understands the end user and what it is that motivates and drives them and the different segments, without that person in the room, the data scientist is never going to work that out. Never know. So I think a range of different skills is the first thing. And then just recognizing that you need to involve that person or ask that person or ask those questions before you make too many assumptions.

Yeah, I love that. I have always found that, like, what are we going to do with this? What are we trying to achieve? All of those being really helpful questions there.

Yeah, it's hard, I think, is the answer. And you'll constantly make mistakes. All you can do is keep trying and involve a range of different, different people and be careful not to jump to conclusions too quickly. Very quiet bunch. Oh, here we go. Yay.

When it comes to, when it comes to trust and stakeholders, when you have a business decision or an insight that runs completely contrary to what they originally believe or what they may perceive, how is it that you kind of communicate and get them to trust that whatever it is is actually actionable, even though it's completely out of box for them.

That'S such an important question. I would say, honestly, I really struggle with this. You've got a stakeholder who wants to make decision X. You know, they do your data points and direction, why it's the hardest thing in the world. I'm going to go back to the music industry again. And I think the reason, one of the reasons why my team was so successful in the music industry, in an industry that not only didn't use data at that time, but actually disliked the notion of using data. Actually, when I left the music industry, one of the old timers gave me a big hug and said, thank you so much. When you first arrived, I won't swear, but when you first arrived, I thought you were one of those in a suit. And I was so wrong. You really care and you have my interests at heart. And I think in there lies a lesson. One of the reasons my team was so successful is because we built those relationships of trust with the stakeholders so that they knew we were in it together and we had their interests, the business interests at heart. We weren't there to challenge or embarrass them. In the music industry, that was about spending an awful lot of time at gigs. I didn't really want to go to late at night missing my children and drinking a lot of beer, but it showed that I was there with the team, in the trenches, doing the thing and cared about the, the subject matter. With you. I don't know what that looks like, but I think frankly, putting the hours into building those deep relationships of trust is one of the absolutely fundamental parts of it. And whenever I've skimped on that, it's hurt me in the music industry, I did that. It wasn't so painful to go to lots of gigs backstage and drink lots of champagne. In the book publishing industry, I'm not that interested in books. I don't care. Never bothered to go to book launches. I never bothered to build relationships. I, I just thought, I've done this once in music, I can do it again in book publishing. I got the analytics, it's fine. I don't think anyone trusted me. I don't think they thought I had their interests at heart. I don't think they thought I cared. And so they didn't really trust me or involve me or let me in on those decisions. So I think trust is the fundamental part. And then lots of, once you've got trust, you can have lots of careful listening and you can better understand why they think the thing they think is the answer. And, and it might just change the answer actually, and they'll respect you. So another music industry example with actually the same person who signed Coldplay, he had another artist he was incredibly excited about. He'd signed them. They hadn't Released anything yet. And we tested their music amongst those audiences. It sucked. Nobody liked this artist, and he'd made all this fanfare inside the organization about how they were the next cold player, you know, blah, blah, blah. And so, because we had that trust, first of all, he believed me and trusted the data, didn't think it was. And secondly, because he believed that he managed that decision in his own way. So there was no explicit challenge. There was no embarrassing him in a meeting. There was no explicit canceling of their contract or launch. The release date just kind of slipped and slipped and slipped and then fell off the release schedule. So the right decision got made, but in a way that was acceptable to everybody, you know, So I think trust and respect and listening. And then in my experience, good things happen, even when it's incredibly challenging to the person's public. You know, public statements of intent.

Hi, I think this is a very.

Similar question, but have you ever had, like, ego get in the way of having, like, artists, directors, or writers understand the data? Yeah, absolutely. Absolutely. I've met some really dumb people. This is not going public.

Right.

So I won't name any artists, but there was one artist manager, maybe as sensible I can, one of the biggest artists in the world. And I was advocating for this phased approach of thinking about audiences carefully. Like the David Guetter examples. Exactly. That we proved it to so many artists. It was obviously the way to do it. And he just wanted to freestyle the release of a new album. And he said to me, why do I have to do all the planning stuff and all the thinking stuff that you're advocating for? I know it's going to be a hit anyway, and. And I said, but hang on. If I can make your release, like, 1% bigger or 5% bigger or 2% bigger, what's that worth in economic terms? He's like, whatever. Just didn't want to do it. So ego and ways of working, you just did not want to think or plan.

All right, cool.

You do it your way then. So, yeah, I mean, I've definitely met some very people who really want to do it their way, and that's it. And in the music industry, you let people get away with that kind of thing. Often in other industries, less so. But I think in my experience, almost everybody is in it for the right reasons. At the end of the day, they really do want the right thing. And so even examples like that, I have to believe if I could have built a relationship of trust with him, if I could have had him really understand what we were trying to do. And it was for the best. I have to believe we could have made that release 5 or 10% bigger globally and made an awful lot of money together. So I haven't met people who are genuinely, genuinely trying to make bad decisions. There's some miscommunication about what the goal is here or how we want to work, or some lack of trust, I think. And again, I've got lots of examples where I think I've messed that up because I just didn't bother building trust that person. I think I had tried, but I just had failed. But there's lots more examples where I didn't bother because I thought the case was so clear analytically, I didn't need to bother wooing them. And it failed. And I realized now I should have put more effort into the relationship. Not too many, I would say. Ask me again. I have a drink. So I have a question for you and maybe a softer question really into the music industry. Did you forecast that vinyl sales would come back and be as big and popular as it did now?

Right.

I mean, you're in an industry where things switch pretty quickly between mediums and kind of hinge moments happen every generation seems like, or half generation. How did you guys forecast that? Or did data forecast that change happening? Yeah, I don't think. No, I don't think so, no. Although it's a long, slow trend. And so I think if you were paying attention, you would have. You probably would have called it. I think what's more interesting to me on vinyl is to the lessons about digging deeper and underlying needs is why are people buying vinyl? And I think the reason in that old graph in the 70s, why were people buying vinyl? The answer is to listen to music. It's the only way you can listen to the music you choose to listen to. You have to go by the record. And so that part of the graph is vinyl sales, but for one reason. And I think that the current vinyl uptick is vinyl sales for a completely different reason. I think most vinyls are not listened to. They're physical artifacts. They're for collectors or people who want to feel attached physically to an artist. And I think if you look back, that's always been the need in music. I worked with Pink Floyd and they were really reluctant to do a big extravagant box set of their work because they thought they were ripping off their fans by charging a lot for big, big box sets. We showed them that actually their fans want that. And by not giving them physical, scarce artifacts that are intricate and, you know, Sophisticated by not giving them that, they were actually letting their fans down at that moment. They said, oh, okay, sure, we've always wanted to. We just didn't want to rip our fans off. So I think there's always been this need for physical artifacts. Vinyl also did that in the original days. You buy it for listening, but it's also an artifact in the world where downloads and streaming came along and took away all physical artifacts. I think it was inevitable. This is how I think we should have forecast it. It was inevitable that we were going to go back to some kind of physical artifacts. We got it wrong. I would say as well, we launched all sorts of interesting physical artifacts in the digital days at emi, but I don't think any of them really took traction. The answer was staring in our face do vinyl. But I think the need is somewhat universal. It's always been there. Obviously was unsatisfied and, and to some extent it was going to come back. I guess we should have been smarter about spotting it and riding that wave a little earlier.

We have time for one more quick question.

Well, thank you, thank you for coming. I really appreciate it. Lots of great insights and wonderful stories. We have an a number of diverse product teams making products and they're using agile and they're kind of moving forward with minimal viable products and producing wonderful things. And I feel like we at analytics are kind of the end of the line. And now they say, oh, okay, now measure how good was that? Or things like that. And we're kind of the people that surface o in making the minimal viable product. We didn't put in all the things we need. So what can the analytics practitioners in this room do to generate the need from the top to push down requirements that we make better, accessible, consumable data to produce what we need?

Nice. If I've understood your question correctly, and there's two aspects to it. If not, we can pick it up in the next session after the break. But if I understood it correctly, I think there's two lessons. One is when designing products, you've got to have a philosophy of basing them off deep underlying needs. Just sell it at the highest level. And it's the story of the skateboard versus the doing tricks and it's the story of David Guetta and the different underlying needs that drive the kind of music he should be making. So that's a philosophy you've got to sell in quality. On the process of building products and making sure that data is useful throughout that process and used throughout that process. I learned an awful lot here by when I went from working at the BBC which had 100-year-old process for making TV shows and I thought we'd done a great job of retrofitting data and analytics into that process. From the what type of show should we make to should I make this specific show through to how should I adapt this throughout the process, the innovation process. I thought we'd done a good job of retrofitting data. Then I went to San Francisco, worked at Masterclass which produces incredible content since it's a tech company. We started with a blank sheet of paper and injected the right data and analytics at absolutely every step in the process. And my God, was it so much easier to do product innovation and to create content. Create content when data is just baked in as a natural part of the process. So I think a good exercise would be to say, hey, instead of retrofitting analytics into an existing process, if we started with a blank sheet of paper, how would we do it? That contrast will be stark. I'd strongly recommend piloting starting with a new model instead of retrofitting the old model. So I hope that's roughly what you're going to.

It's wonderful idea. We'll have to see if we can figure out a spot to apply it.

Pilot it somewhere and then it will end up taking over the main process. So that's a great way of driving change. Pilot it somewhere non controversial and then it will take over. So in the EMI Music I ended up taking over the whole business, analytically speaking. I started in catalog back catalog pricing optimization or something silly. Ended up taking over the business business. But start somewhere, it doesn't matter where. Prove it and then you'll roll it out from there. Thank you so much. We'll talk more and yeah, I'm around for the Q and A session and then if any of you got any questions at any point, just reach out to me. I'll make sure my email goes with the slides and I'm very happy to kind of help and talk more about this and learn more about your business. Thank you so so much.