Jonas Christensen 1:46
Okay, Felipe, we are back for part two of a very exciting interview series here. Because last time, we talked so much that we didn't get to cover everything that we wanted.
Felipe Flores 1:58
That's it. Well, thanks so much.
Jonas Christensen 1:59
You're most welcome. We heard about your ascent from a backpacker with poor English to a data science executive with perfect English and your whole journey throughout that. We heard about data driven leadership, and you had lots of sage advice for listeners out there about what to do and what not to do. One of the things we didn't capture is what you actually do day-to-day. You're very much a practitioner in this space and you are one of the executives of what is called Honeysuckle Health. I'm very interested to learn about Honeysuckle Health and what you do there. And I'm sure listeners are. So could you tell us about your role there and what you do and what the company does as well?
Felipe Flores 2:36
Yeah, happy to. And thanks so much. Thanks so much for having me back mate. Yeah, always a pleasure spending time with you. And yeah, very keen to discuss this part as well. And thank you so much for the kind remarks. Very, very generous. So thank you. Thank you so much. So before working in Honeysuckle Health, I'd been in banking and finance for about five years. And I found the challenges super interesting. The applications for AI were almost endless. The data sets were extremely rich. It's an industry that it's one of the areas that have overall sort of higher maturity compared to other industries like health, and many others. I think banking and finance are a little bit ahead there. But the whole time that I was in banking and finance, or maybe not the whole time, but increasingly over time, I felt like we were using this amazing technology to sell people money. And it got to a point where it felt like - it was like trying to find people online and say ''Do you want some money? Would you like would you like my - how much money would you like? Would you like more money?'' And I was like - I was like enjoying the technical side. But over time, I wanted to move into something different, something that ideally was more purpose driven. And in thinking about where that could be. And you know, that's one of the beauties of working in data science. You can move across industries quite freely. And yeah, like I have friends who work either in mining and not as data scientists, but working in mining or power generation. And those are sort of older school careers that are stuck, generally stuck in one track unless you retrain, but we're so lucky that we're able to change industries. So I'm thinking about what type of industry I would like to go into next. Healthcare came up as one of the main ones where it felt like it was primed for disruption, both through digital technologies and AI specifically, and it felt like something was bubbling up in AI. This is a couple years ago. So in 2019, as I started to think about it, I had been living in Melbourne at the time for about 12 years and I love Melbourne. It feels like home in Australia. I was getting to a point where the cold was getting to me and I wanted to move somewhere with a warmer weather, close to the beach and have a better lifestyle in terms of like a smaller place. And that was mostly for family reasons. So kind of like multiple, multiple things that led to change. And I was I'm lucky enough to come across the opportunity at Honeysuckle Health through an old colleague of mine: she was a main stakeholder in the past. So we worked together. She was head of marketing and I was head of data science. At the beginning, the relationship was quite difficult. And she was a person that was not very data driven. She felt like, you know, the CEO had almost like, forced this data analytic stuff into her domain. And then suddenly, like I was there, and they were like ''You guys work it out''. I remember like finding it so challenging at the beginning. And it was one of the relationships that I really invested. And, like I remember for one of the main projects that we got off the ground in her area that ended up increasing conversion rates by about 30%, which was like a huge, huge effect on the on the business. I remember to get that off the ground, I probably spent - I remember quantifying for the team - like I probably spent like 600 hours working with the stakeholder and like with other people, but like, the stakeholder was involved 600 hours, so the work to get this initiative of the ground and that man like ideating, socialising, getting people comfortable, working through it, and etc. And in the end was a success. And anyway, all I'm saying there is that I personally haven't always been so deliberate about investing in stakeholder relationships. And I think most people in our industry, sometimes we're like ''Oh, you know, stakeholders are a pain in the ass'' or ''stakeholders, this or stakeholders that'' and sometimes, like, if you double down invest, it pays off professionally and personally. So with this stakeholder, we became friends. And then like, years later, I mentioned - like we caught up and I mentioned that I wanted to move into healthcare. And I'm always looking for a seat change. And she was like ''You know what? I've got a friend who, we went to primary school together. He's setting up a new business in the healthcare space. And they're looking for a head of data science right now. Like I saw him on the weekend and we caught up on the Monday''. Like, amazing, amazing coincidence. So literally, we're catching up for lunch. She tells me this, she's like ''You want me to send in your CV''. ''Yes''. She texts him my LinkedIn. Before we finish lunch, he's like ''Book him in for an interview''. And a couple weeks later, I was here in Newcastle. Did the interview, got the gig and came back the next month to interview people for my team. And a couple months later, I was here when I joined this healthcare company. And I'll tell you a little bit more about it in a second. But when I joined, I was employee number eight, based on when I started and now we're over 100. And it's been two years, just over two years. The growth has been awesome, has been really, really exciting. In the company, what we're trying to do is help people live healthier lives. And the focus there is smack bang in health care, both on the preventative health side. So primary health and the wellness area on how do you get people to either stay healthy or be a little bit healthier over time. And then we've moved into secondary tertiary healthcare, which is around like - hospitals doing work with hospitals, analysing data from surgeries, clinical practices, quality of health care, what is the value of healthcare: are all things that we're measuring, and working with different partners across the industry to help feed back some of this data, where in some cases, at least in Australia, for the partners that we work with, in some cases, like it's never been looked at before. In other cases, like the data hasn't been aggregated at an industry level. So it's been difficult to get benchmarks. And in other cases, that depth hasn't been there. And a big part of this change is assisted by a big change in strategy that we're seeing from health insurance in Australia, where traditionally health insurers have been payers, and in the words of some of them, they feel like they were dumb payers. Like essentially, like if things came in, they would pay the bill and a large number of health insurers, they're wanted to change their strategy to move from a healthcare payer to a health care partner. And that means getting a lot closer with the members, with individuals and understanding their health needs and their health journey a lot better and be able to provide recommendations. That's where Honeysuckle Health sits and where we can help. So our mission is to help people live healthier lives. The way that we're doing that, at least in my mind from a data science perspective, it's almost like taking the playbook of the big tech companies in the US and what they did to advertising, which is obviously like bad. But taking those learnings and applying it to healthcare but for good. So what I mean by that is like - take such aspects of personalization of finding people at right time and offering them a message that will motivate them to action. And in this case, the action is not buy something, but an action is like: develop a better habit or like do some exercise or like go see this particular physician or get this type of treatment. So, like the personalised recommendations delivered to people at the right time, with a big focus on the preventative side, over time, we'll be spending more and more of our data science efforts on the behavioural science component and how to speak to motivate different people differently to get better health outcomes, which we measure and obviously feedback to all our models. It's super early days to do this in healthcare. There's a number of very large challenges that need to be overcome. And we as an executive team, we discussed these openly before everyone joined. And we sort of said like ''This is gonna take a long time''. We kind of agreed to give it about 10 years and see where we get to. And that was really nice, because at least from a data perspective, the main challenges, the way that I see it is: There's a lot of data that needs to be captured in order to support everything else we want to do. There's a lot of connections that need to be created within the industry. So that could be like GPs and health insurers, especially in getting data flowing, especially like measurement and outcome data, but there's a lot of linking and connections that need to be created. And then overall, there's a big optimization piece that needs to be done in order to help people find and have better pathways to better health. And what I mean when I say ''a pathway'' is that in healthcare, it's quite an obfuscated market in the sense that there's not a lot of price transparency upfront. There's not a lot of transparency about outcomes and what similar people have gotten as a result from a particular doctor or specialist and the treatment. So we're like collecting all that data and helping optimise the adventure that people go when they interact with the healthcare system. If you think of it as like "Choose your own adventure" and everyone goes to different paths. And the last thing I'll say here is that I'm particularly excited by that because the majority of the applications of AI in healthcare have been about diagnostics, and primarily computer vision imaging. So it's been like taking X-rays or brain scans and doing the diagnosis or what conditions and there's fantastic companies doing excellent work in that space. And I think that that's the area of AI in healthcare that gets the majority of the attention, because that's where you can apply a lot of the cool algorithms, the core deep learning algorithms, you can apply them readily to that space. But on the other extreme, there's companies like Honeysuckle Health, where we're taking this broader view of the patient journey, and having a person-centric approach and looking at the entire adventure and "Choose your own adventure" and wanting to optimise that for people in a way that's evidence based and whether the data has been captured, linked, and then the recommendation has been optimised: like the journey being optimised with personalised recommendations, and then a bit of behavioural science. That's where we're hoping to go, it's gonna be a long journey. We're right at the start. We've had a few little wins, a lot of setbacks, a lot of big challenges, and there's more yet to come. But I definitely feel like it's got a lot of what I was looking for in terms of a more purposeful application of analytics and AI. And that's made a really, really great journey.
Jonas Christensen 13:39
Brilliant! Oh, there's so many questions I want to ask you now, after that long explanation. Thank you for that, Felipe. I should say one of the things that listeners might have picked up on is that it can actually get cold in Australia. You mentioned that Melbourne is cold. Most of Australia is typically scorching hot, but we do also get cold pods in winter. And we don't ride the kangaroos to work either. So that's another myth busted there. So for all you international listeners, that's how it is here. Now, Felipe, I'm hearing sort of a few types of projects in what you do. There is some that I would call maybe "traditional analytics consulting" as in you take a data set, you find some patterns, and you sort of report back the insights and they can be used to make more strategic decisions. But it also sounds like there is stuff that is really, in the nitty gritty of - you say personalization, right? - so that necessarily means that you have to pick some really key points in a customer-client journey and optimise for those and really analyse them from all angles and work with the subject matter experts, right in the frontline to really figure out what's going on. And if there's a gamification element to, in the whole prevention piece, right, because you've got to motivate people to get off the couch and go running or stop smoking or wherever the leaves are that you're pulling. Two questions from me before many other questions. Where are you getting this data from? And how are you interacting with sort of the frontline professionals, maybe carers and so on, that aren't directly in your organisation, but you're going through another organisation to get that? How you develop the things on the ground?
Felipe Flores 13:40
Awesome questions, right. So on the data side, the data comes from our partners, and we're taking an ecosystem approach to the improvements that we would like to see in the healthcare industry. So that means that we've started by working with what we're calling the ''Coalition of the Willing''. So we went and knocked on a lot of doors and said, like our vision, what we would like to see as changes in the industry. And we got a lot of doors closed in our face. And then we got some that invited us in to have more of a chat. Or we realised that that's just the starting point, and kind of like the first pass, but we're working with this coalition of the willing, and all the data that we get is from third parties. The majority of the data, sorry - I should say that we do offer some support programmes directly to members that we send to patients, that we get paid from the insurers, so then it's free for the patient. And that was one of the early pieces of advice from our parent companies that they said ''Don't just be a data science company, but have some arms and legs that can help you execute on the recommendations''. Because customers like being organisations, they will want to buy the insight, the recommendation and the service on how to capture that value. And it was fantastic advice. So we got a couple. We got two other divisions or business lines that help us do that. So we have the brain and the AI, and then the arms and legs to help us make that happen. But the data side largely on partnerships. And as you can imagine, in healthcare, a lot of privacy, a lot of information security, a lot of the identification, and we've done a lot of work to get that to a top notch standard and coming from banking was surprisingly helpful in that side. And then when it comes to the recommendations, and I guess the link between the data and the recommendations there, is that today, a lot of the data that's required to not only provide personalised recommendations, but also to do it in a responsible and ethical way, is not available. So we're having to capture a lot of it. And last week when you and I were at the Advancing AI Melbourne conference, and when we had the panel on AI ethics, one of the panellists Jeanine, she mentioned this case about health care in the US, where a company made a predictive model to estimate the benefit that different people were going to get from a particular intervention. And that benefit was being predicted from a financial perspective, used as a proxy of healthcare benefits, but essentially, say ''if I was going to get $600 worth of benefit from this programme, and you're going to get $100 for benefit, then the programme should be offered to me first, before you''. Later - when you think about it in the first line, it sounds good - but then what was found later on was that that algorithm was discriminating, inadvertently discriminating against lower socioeconomic people, because they hadn't spent as much on health care. So then there wasn't as much to save, compared to people that could spend more. So that kind of like hit the news and made everyone very, very afraid. And one of the things that that I realised after that is like, today, we don't have the data to do a better job. So yes, we're using finances, which is terrible proxy. But we do have to capture a lot of new data in order to - about people's health, and their behaviours and their lifestyle, and even their emotions and mental health, we need to capture a lot more data in order to be able to provide people better recommendations and better value. And then the way to do that is your other point around working with the doctor community: the clinicians, sorry, the nurses, the humans, essentially, that are treating the patients. And today, it's all pretty much all of the work that we've done, if not all of it, has been with a human in the loop. And that human in the loop is a clinician that is, you know, interacting with our analytics or AI, also without technology, and then they're making kind of like the final decision to pass that recommendation or a different recommendation to the patient that they have in front of them. And that's great feedback for our systems as well. So like we totally welcome it. And your other point around, how do you get people in other organisations to want to take part of these changes? And like it's super hard. It's super hard, where healthcare is a super-fragmented market and everyone has different incentives. So everyone wants different things and to be able to get them get a particular group of people, say it can be like: orthopaedic surgeons or general practitioners like family doctors, if you want GPs or family doctors to be part of this coalition of the willing, you have to understand deeply what they care about, what they value, what their world is like. And then build some, generally some technology, some analytics and some AI to cater for them. And that will be different to the orthopaedic surgeons and the nurses and the insurance companies. Like, everyone will have different and has different needs. And having that kind of like deep Product Development at such a specialised level across an ecosystem has been particularly challenging. So we have some initiatives that are at a very early pilot stage. And then we have some initiatives that are more mature. Nothing super mature, but things that are improving. And lots of ideas that are still yet to come when we get the kind of like the right partners to make it work. But for example, we've been linking GPs, so family doctors with health insurers, so then the health insurer - everyone's health insurance in Australia typically offers a number of support programmes for free to their members, and the members don't know about it, and the insurer is unsure of who should get what programme. So one of the services that we do for the insurers is like, let our models tell them who should get each individual programme and when. And that's been helpful. But the data that we're using there is claims data, so it's largely payments data. So we're doing that as one avenue. But on the other side, we have software that GPs are using, that shows them the support programmes that are available to the patient that is sitting in front of them, and then they get a quick description. And the GPS is able to offer and refer those programmes straight to the patient sitting in front of them, and they can enrol them in these programmes. And then the GP gets feedback about how the patient is going on those support programmes. And they're an additional to the care that they're getting so far, like from the GP and specialist and anyone else that is in their care team. And having that information at their fingertips, that two way communication, and then some measurable outcomes being fed back to the GPs, are ways that we're making progress on that on that side. But they're still across the industry. There's many, many areas that are yet to come.
Your data collection exercise is massively complex and multifaceted. And what I'm imagining here is this is really, I don't know, 10 year journey to really get a good data set here. This is going to take a long time, because it's very fragmented. But also we all know the joke of doctor's handwriting being terrible, and all that stuff. But sometimes there's actually where you're starting, like they might be writing handwritten notes or, or quick things that they print out and give you a referral for the next thing. And that's sort of the standard of technological advancement in these practices. So I imagine that you're actually having to build workflows and software and systems to actually make it so easy for people to capture data that they almost want to do it without actually, as in, it's almost easier to do that than what they did before. And you kind of have to flip it around to be the customer experience of the doctrine in that sense. So you can tell me whether that's right or wrong. What I'm also interested in here is for listeners to understand, Honey Circle Health is not just your average startup. You've got some financial, technological and IP firepower behind you to actually do this and play this long game. So maybe you could talk a little bit about how you constructed in ownership and all that stuff as well.
Yeah, now that sounds great. So to your first point around getting the user engagement, and in this case our users are largely doctors and insurers, basically, and then bit bit of government work. But to get the users engaged, then it needs to be a process that's a lot more seamless than what they had before, where we're reducing friction, being able to pre populate a lot of information as much as we can and making it a couple of clicks instead of a form, for example. And for that we've done a lot of integration with different government agencies that has helped us do a lot of that heavy lifting. And now we're still like capturing a lot of data, but doing a lot of technological development in sort of apps and connectivity to be able to link the different parts of the healthcare system and bring them together and get that information flowing with obviously a bit of analytics and AI behind the scenes. And then to your other point around Honeysuckle Health. So the company started at the beginning of 2020. So in January 2020. The company Honeysuckle Health has two parent companies, who are both health insurers. They each own 50% of Honeysuckle health. There's one based in Australia and one based in the US. In Australia, the insurer is NIB, which is about the fourth largest health insurer in Australia. So they own 50%, of Honeysuckle Health. And then we have a few people on the on the executive team who have come from NIB. And for example, my boss, the CEO: he was an NIB for 14, over almost 15 years. And for the last seven years, he ran the biggest division in NIB, which is the Australian resident's health insurance business. So that's the majority of their book. The biggest part and he ran that for for seven years. And then the other organisation is Cigna, which is a global health insurers health insurer based in the US, they are massive. So for example, in Asia, and in Europe, and in America, they're they're massive. In the US, they're also about the fourth largest health insurer, and they're global. And I would like to mention that for scale, for example, Cigna has about 1500 people working in data analytics and data science. Like 1500 is amazing. And then they have an additional about 1000 people working in data engineering. The companies are massive and a powerhouse. So we've got both of these companies came together, largely through the efforts of my boss now. And the CEO of NIB, Mark Fitzgibbon, that they initially came up with the idea of having a data science company in the Australian healthcare space, and then wanted to find a partner that could help them speed up their learnings and their progress. And then they went to the US and they met with five or six of the largest health insurers. And the NIB already had a bit of a relationship with Cigna. But they still met with kind of like all the major health insurers, and there was like different levels of eagerness to move forward. And Cigna was like ''Yeah, that sounds great. Let's do it. Let's move''. And it was literally like, from my understanding - this was before my time - but from my understanding, it was like, about a year from first conversation to Honeysuckle Health being formed. So that meant that in 2019, essentially the conversation, the contracts, the agreements, the payment, and everything happened, which I thought was super fast, essentially. And yeah. Then Honeysuckle Health launch in January 2020. I'd had sort of my interviews and everything before that, but then I joined in Feb 2020. And it's been, Yeah, awesome, awesome so far. And yeah, so working with with insurers, and with hospitals and doctor groups and a little bit with government, and we see government as an area of growth for us. We think we can add a little value to.
Jonas Christensen 27:49
So we've established that you have firepower to do the long game and to play the long game, and you need to do that. And so, Felipe, I'm interested in what is this long game as in where do you actually think that we are going to be using data to improve health care and our general health in society? So this is more of a societal question, I suppose. And Honeysuckle is a big part of that. Where are we going to be using data to improve health across society in the next sort of 10 years or so?
Felipe Flores 28:19
I love it. Thanks for that question. One of the things that I'm really passionate about is called ''Patient Reported Outcome'' measures. It's been widely used in other parts of the world. And one of the examples that I came across from Europe actually was around prostate cancer and prostate surgeries. And then through this example, you can see the differences in capturing and tracking the right data, analysing it, and feeding it back to the system, and then opening it up and what the benefits of that can have. So I'll spend a couple of minutes sharing this story. In most cancer surgeries, the data that is tracked is something that you can capture from a clinicians perspective, and it's easy to capture, and it's around mortality. It's kind of like the main thing. So five years after somebody had a surgery for cancer, what is the survival rate or the mortality rate for that cohort? And most hospitals around the world have spent a lot of time improving their techniques and their treatments in order to increase that survival rate over time. And measuring it at the five year mark is is one of the key components there. That is largely done and that improvement is there. That's tracked. What's not tracked is these patient recorded outcome measures. So they have to come from the patient about how they're doing after the surgery. In this case of prostate cancer and then prostate surgery, what you ask the patient - because these outcome measures are very specific to the procedure that they had done - so in this case, you ask them whether at one year after the surgery, whether their experience incontinence and or erectile dysfunction. So these are things that are super personal, kind of like quite, like a little bit embarrassing to talk about, and not something that can be measured clinically without that sort of personal engagement from the patient. So you have to ask them. And then a few years ago, there's a clinic in Germany that started measuring these patient reported outcome measures for prostate cancer. And they found that they have a much larger than expected at the beginning and as a baseline, they had a much larger than expected proportion of men having incontinence and erectile dysfunction at the one year mark after the surgery. So then they fed that back to the surgeons and to the doctors and practitioners. And then they started working on ways that they could improve that, so that less people would have those complications a year out. And then over time, they brought it lower and lower and lower and lower. And then today, when you compare with overall rates for Europe, in general - this is a clinic in Germany - but when you compare it to the overall rates with Europe in general, as of two or three years ago, there were about 10 times better when it came to having those complications a year after surgery. So what I love about the power of patient reported outcomes, and being able to feed that back to the system and improve things is that for everyone, if you have a family member that gets a condition where they need serious treatment, you would like that family member to get the best possible care. And you'd like to know the types of outcomes that other people have received from the practitioners that your family member is about to see. And then within that, as I was saying, traditionally, its mortality rate. But there's other more personal metrics that can and should be captured. And that will give us a lot better certainty on who should be seeing us and our family members at different points in time. So by doing this across the health care industry, and as you were saying is definitely a long game, we will hopefully help people answer the question of ''Who should my mom see for her knee reconstruction?'' Right? Or like 'Who should my dad see for his prostate surgery'' or your wife, like who should your wife see for breast cancer surgery. Like things like that, that is like so important, so emotionally charged, you want to have or maybe I'm biassed, but I'm hoping that most people want to have the evidence, right there at their fingertips. Not only the patients, but the clinicians as well to be tracking and analysing and feeding back the right information. So we can all improve together as an industry, and then we can be offering better and better services over time. And that's something that we're super passionate about.
Jonas Christensen 32:57
Hi there, dear listener, I just want to quickly let you know that I have recently published a book with six other authors called ''Demystifying AI For The Enterprise, A Playbook For Digital Transformation''. If you'd like to learn more about the book, then head over to www.leadersofanalytics.com/ai. Now back to the show.
I've reflected a bit on as I've engaged with different healthcare systems over the last few years how everything is research based, evidence based in its creation, but the application of healthcare is not very scientific at all, to be honest. Everyone gets some doses of medication, you know, in 25 milligram increments or whatever it might be, because that's what the packet says. And that's what we do. It's not individualised. Everyone gets a hospital bed in a certain way because that's the system that you're going into rather than the system adapting to who you are and what you're there for. Joining a queue of others, people who are on this conveyor belt of a hospital system or whatever it is, to be put through the sausage machine. And the professionals in there, nurses and doctors and so on, are trying to make it work, but they're sometimes a prisoner in that system themselves. So there's so much you can do with personalization that is not picking out darker marks and an X-ray and all that stuff, that you can use data to inform. A simple one actually that came to mind when you were talking is: I was speaking to someone who is an executive in a health care insurance company, just like the ones you've described previously, and they found that when they introduced care at home - so people go home and recover rather than being in hospital, but the hospital will send the nurse or the carer to the home to take care of that person- recovery is much better, satisfaction is much better and general quality of life is much better. And it was cheaper. So these are examples of personalization that are not sort of super technologically advanced, but they do the trick, right. They're the things that you've got to find in data by experimenting, and so on.
Felipe Flores 35:03
I love it. That's a great example. One of the challenges of experimentation in healthcare that I didn't foresee coming into the industry is that you obviously want to have a level of experimentation. But the stakes are completely different. When you compare to a website design and having, you know, an A B testing that we're just splitting people coming into the website. We split them evenly. So for a lot of the interventions that either we're creating, or that we're measuring the effectiveness, we're seeking to measure the effectiveness, we've had to do retrospective control groups and create retrospective control groups. So essentially because of the ethics that if you're developing an intervention that can help somebody, but you're saying ''I'm not going to offer it to them, just to see if they get worse, and then I can validate that my intervention is good'', when you get into health care, that's kind of like a little bit tricky, a little bit iffy. So yeah, the approach that we've done is like, if we're estimating that it benefits, then let's give people the option, either directly to them or to the clinician but let's make them aware and then give them the option to opt in. And then we get our treatment people that way. But then to create our control groups, we look historically. And mostly in recent years, we look for people that match the treatment ones, from a data perspective as closely as possible. And obviously, there's limitations there, because of the data visibility and access. But when you're looking across a couple 100 variables, it does get to a point that it's quite fine grained. And obviously, you need quite a large number of people to do that. So we're lucky to have over a million individuals in our database with a bit of history there. And we've been able to find control groups or control people to match our treatment people with and then we can look at their health over time and estimate and measure the differences between the people that would have needed the treatment, but the treatment wasn't available at the time. And the people that did get the treatment. What are the differences in their health over time? And with that, we've been able to quantify the benefits of interventions in health care. One of the interesting things there is that we've had treatment people that got matched to a number of controls. That there's many similar people like them, sometimes 20 people, sometimes 100 people, and then we get to sort of average out the results. And that's one of the ways that we do that. Sometimes we did like random matching. So a few different ways. But it's been been an interesting challenge, measuring the effectiveness of interventions in healthcare, and something that the team worked really hard on to create kind of like a framework that is reproducible, and can be applied to different interventions. And that recreates this retrospective control group to do comparisons, and then get a quantifiable figure. And then we expanded from the financial side to look at loyalty and retention, to look at the outcomes, the clinical outcomes, and a bunch of experience, so NPS and a bunch of other metrics to kind of like round up the whole benefit measurement. And then we look at which part of the benefit goes to different providers and organisations that are in the mix. And in Australia, a lot of it by design goes back to the government, which is really, really interesting in terms of the the savings if you want from having healthier people. A part of it goes to the insurer, obviously the individual gets the main benefit. And some of it goes to the insurer, and the majority goes to the government, which is Yeah, excellent.
Jonas Christensen 38:49
Nice. If I take this conversation one level up, what we are talking about, really, to a large extent is a debate around model Centricity versus data Centricity to an extent. So for listeners, what that means is the last, say, 10-15 years, there's been a lot of advancements in the types of models, we can do. AI models, machine learning models and the techniques and they are both very good and easy to spin up if you have the right path and package and so on. And it's available to everyone - open source for free. So there's an argument, there by people who are much wiser than me, that that's actually not what we should be solving. We should be solving for the next problem, the bigger problem, which is we often actually don't have the right data to make predictions that are accurate, reliable, ethically correct and explainable and which is basically what you're describing here. You have to go and create the data. It's not a technology problem. It's a data creation problem. And the other thing is: it really screams here that data science is not about just the science of using data. It's so much more. This whole organisation of all these people with different priorities in different businesses and different systems is a way bigger task than actually finding insights in the data. And it's what you're going to spend some years, lining up the data set. It's a really fascinating example of what we can do, but the effort that is required to actually do that is massive.
Felipe Flores 40:13
Yeah. Nice. Thank you wish us luck.
Jonas Christensen 40:15
Yeah, I do do that.
Felipe Flores 40:17
We'll keep trying.
Jonas Christensen 40:19
Look, Felipe, I think we're towards the end here. It's been a blast. It's been such a blast that we had to do it over two episodes. And listeners I've talked about before the book, of course, ''Demystifying AI for the Enterprise'', and we were very lucky to have Felipe contribute a case study to that book, which is a brilliant case study from his time in banking. And I learned a lot from that case study myself, if you're interested in that, you can check that out and buy the book. And Felipe, other than that, any parting words to the audience?
Felipe Flores 40:48
Oh, mate, I have to encourage people to get the book. It's fantastic. And I've got a lot from going through the book, working through it. And also getting an opportunity to contribute was phenomenal. I remember my wife's cousin got married in Lord Howe Island, which is off the coast of New South Wales in Australia,. Tiny, beautiful island. Well, while we were there, I was writing and finishing the case study, and it has like beautiful memories of like being with family, being in such a beautiful, beautiful location, and then being able to contribute a really small part to an amazing project that you guys spent so much love and blood and sweat and tears into it, and it came out phenomenally so yeah. For anyone and everyone who's listening. Please check out the book. It is outstanding.
Jonas Christensen 41:37
I didn't know that story. Felipe. I was imagining you sitting in a dark office somewhere but you're actually sitting on a beach looking at the beautiful skyline, so that's good.
Felipe Flores 41:46
Yeah, beautiful mountains at the beach. It was amazing
Jonas Christensen 41:50
That must be why it reads like poetry. Felipe, thank you so much for being on Leaders of Analytics again. All the best with your long game and we look forward to following the success of it and feeling it in our personal health care around the world. Thank you so much, and all the best.
Felipe Flores 42:06
Amazing. Thank you so much, mate.
