💡 Download the complete guide to AI-generated synthetic data!
Go to the ebook
Episode 41

Synthetic data for impact: how to innovate in health insurance with Daniela Pak-Graf, Merkur Innovation Lab

Hosted by
Alexandra Ebert
In this episode of the Data Democratization Podcast, host Alexandra Ebert interviews Daniela Pak-Graf, the managing director of Merkur Innovation Lab — the innovation arm of Merkur Insurance. Daniela shares her tips and best practices for innovating with data in one of the most conservative and sensitive industries, health insurance. Tune in to find out how to accelerate innovation through effective data management, forward-thinking organizational decisions, and enabling technologies, like AI-powered synthetic data generation. We cover various topics, such as:
  • [00:02:02] Innovation in healthcare and insurance - the organizational challenges and solutions
  • [00:07:35] Customer-driven product development 
  • [00:12:03] Health insurance transformations - telemedicine and predictive healthcare,
  • [00:20:17] Data cleaning, data management, and accelerating innovation by doing data science in-house,
  • [00:25:09] Innovation success stories: automated, GDPR-compliant decision-making for marketing,
  • [00:28:00] Challenges of data anonymization and the synthetic data solution,
  • [00:34:32] Opportunities for synthetic data: machine learning, dashboards, data sharing, and collaboration with third parties, 
  • [00:42:56] Synthetic data for fair algorithms,
  • [00:47:20] Proactive preventive healthcare, 
  • [00:48:41] Gendered medication in healthcare,
  • [00:50:22] Tips for young female professionals in the healthcare industry.

Transcript

[00:00:09] Alexandra: Hello, and welcome to episode 41 of the Data Democratization Podcast. I'm Alexandra Ebert, your host, and MOSTLY AI's Chief Trust Officer. Today, it's my pleasure to welcome one of our customers on the show. With me is Daniela Pak-Graf, CEO of Merkur Innovation Lab, the innovation hub for all things AI, machine learning, and data for the health insurance organization, Merkur. With Daniela, I'm going to talk about innovation in healthcare, in which direction the entire industry is moving, and also the lessons learned and the secrets Daniela has from her experience bringing innovation to production and managing this change process, particularly in organizations that are heavily regulated, such as the insurance industry.

Plus, of course, we're going to talk about the importance of data, and all the challenges that come with getting access to it. Lastly, we have a big chunk in the episode dedicated to synthetic data, how synthetic data is used at Merkur Innovation Lab, what Daniela sees are the benefits and also the limitations of synthetic data, and which use cases they're already covering and plan to cover in the future. I'm sure there will be plenty of things for you to take away. Let us dive right in.

[music]

Daniela, it's so nice to have you on the Data Democratization Podcast today. I was very much looking forward to our conversation about innovation in the healthcare and insurance space. Before we dive into all these interesting topics, could you briefly introduce yourself to our listeners, and maybe also share why you're so passionate about the work that you do?

[00:01:56] Daniela: For sure. Hi, Alexandra. I'm happy to be here. My name is Daniela, and I've been with Merkur Innovation Lab for three years now, but I've been with the mother company, Merkur Insurance Group, since 2018, right after my second maternity leave. There I started in risk management and insurance mathematics. In 2020, I got the chance to join Merkur Innovation Lab as managing director of a team of one, namely me.

[laughter]

[00:02:46] Alexandra: Startup feeling within a large organization.

[00:02:48] Daniela: Yes. I had the possibility to define a new business, to define a new business model, basically founding a small startup. Sometimes it felt as a real startup. No money, no people, but a great idea. I hope a great idea.

[00:03:14] Alexandra: Maybe for those of our listeners who don't know Merkur and Merkur Innovation Lab, what or what's the position of Merkur in the Austrian market, and also what was the idea behind Merkur Innovation Lab?

[00:03:26] Daniela: Merkur Insurance Group is in one month, 225 years old. The oldest health insurance company in Austria. We are very proud of the use in the market. Merkur Innovation Lab is the small daughter, the small startup of a very old company. Our CEO had the idea, we have so much data, and we're using the data only for calculating insurance products, calculating our costs, and in the era of big data of Google, of Amazon, Netflix, there have to be more possibilities for health insurance data too. He said, "Yes, a new project, a new business, what can we do with our data?" Since 2020, we are doing a lot.

What I'm really proud of is we still exist, and yes, in the last two years, we have made quite an impact on the mother organization concerning the view and the importance of collecting data and working with data in different contexts.

[00:04:58] Alexandra: That really sounds exciting, and I have already so many follow-up questions, but before I start with them, could you also then elaborate on this point, what drives you? Because particularly for everybody working in a startup, you know you have to be passionate about what you do. You have to feel the purpose. What is this for you, Daniela?

[00:05:17] Daniela: What's driving me, it's the possibility to change things, to use something which always has been here, but no one had the idea to do more with it, to make an impact. In this, we succeeded, because we are now using data for various use cases. We are doing predictions of claims. We are doing predictions of retention rates, churn models. We are calculating focus groups for our marketing, or say the strategies.

Use cases, no one thought about three years prior to our existence. It's a small success, but a success, it is.

[00:06:15] Alexandra: Definitely an interesting and future-driven direction you're venturing into. What was the reason to have a separate subsidiary or daughter organization, as opposed to having it just as a department or a cross-functional team within Merkur AG?

[00:06:33] Daniela: As I said, Merkur is a very old company with people working in the insurance business for 20, 30, 40 years. There were a lot of perceptions. Do we need innovation? What is innovation? No one needs data for other purposes than calculating products and paying claims. Our CEO decided innovation is best left outside the company, because then you have an open mind, you attract open-minded people. New people, younger people who know innovation, is important for old businesses. The old business would also continue to exist in 10, 20 or 30 years.

That was the main reason we put it outside the insurance company.

[00:07:46] Alexandra: Makes sense. I could have guessed that this was one of the reason, knowing how challenging change management can actually be. If we zoom out a little bit and look into the health insurance sector as a whole, where is it heading? What's the most important, or what are the most important areas to innovate, in, let's say, the next three to five years?

[00:08:09] Daniela: As everywhere, it's the customer who drives the need for innovation. If you look at Netflix or eCommerce businesses, Amazon, whatever, they know a lot about who is the customer, what does the customer want. I as a customer, I'm used to getting recommendations. I'd like if Amazon is telling me, "Users who bought this, bought that." it's a different way of doing things. The people are expecting other services at a customer support. This was mainly driven by the big players, Apple, Amazon, Netflix, whoever.

I as a customer, I have a personalized experience. In insurance business, nothing is personalized. There is one product, and this product has to fit all over Austria, all over focus groups, all over genders, and that's far from the reality, from what a customer wants. Now we have other forms of communication the customer needs or that wants 24/7.

[00:09:39] Alexandra: And everything digital, smooth experiences, and so on.

[00:09:43] Daniela: Yes. Self-service platforms, chatbots. In insurance, it's a sales agent. You have to call the sales agent, you have to do an appointment with the sales agent, and one month later... just to sign a contract.

[00:10:00] Alexandra: If you're lucky.

[00:10:03] Daniela: [laughs] If you're lucky, yes. That's one point of driving the need for innovation. What's next? Value added services. If you look at Amazon, Amazon started as a book vendor, and now it's Amazon Marketplace. It's Amazon Prime, it's Alexa [laughs].

[00:10:30] Alexandra: A huge logistics company.

[00:10:32] Daniela: A huge logistics company. In insurance, where are the value added services there? I'd like to have a number, where someone is telling me, which doctor to call, in which hospital to go to. I don't know. This doesn't exist. The need for innovation is driven by customer needs, and right now, the old insurance companies are struggling. The customer needs this, but I am not doing that. There's a mismatch.

[00:11:12] Alexandra: Makes sense. If you say it's on the one hand, the personalization and the changed customer expectations, when they use services like Netflix, Amazon, and expect similar things from their insurance providers, then it's the smooth customer experience of the 24/7, everything digital and so on.

One other thing I'd be curious about, talking with many of our customers in the United States, and also with insurance providers in the United States, we see this notion of turning healthcare from something that's reactive and health insurance to something that's more proactive. As opposed to just paying for your medical bills, how can the insurance provider actually add even more value to their customers in terms of nudging them to have a more healthy lifestyle, or maybe going to the check-up appointment with the doctor earlier as opposed to neglecting it for months and months or years on end. Is there anything happening in the Austrian market? Is this something that's discussed at Merkur as well?

[00:12:11] Daniela: Yes, we are beginning with the discussion. We're not the quickest, [laughs] but health insurance or a health insurance company has to make the transformation away from paying claims, only paying claims, to delivering services in terms of prevention, predictive healthcare services, preventing the sickness before it arises, and not paying for symptoms afterwards. It's Merkur's idea to use data given to us by our customers, to use this data and deliver services in terms of prevention.

I don't know which medication does work, which doesn't work, with which sickness, is the medication prescribed really often, or is it an old medication no doctor is prescribing anymore, what could be done in terms of training as a means of staying healthy. In using data we collected or that we've collected for the last, I don't know, decades, we could develop value added services for our customers in terms of prevention.

[00:13:50] Alexandra: Definitely. I would also say that there's vast potential with potentially even 200 years worth of data. [laughs] Lots to come yet. You mentioned it several times now, that Merkur is not the quickest, which is something I suppose holds true for the entire insurance industry, which is just very traditional, heavily regulated, risk averse, oftentimes also very conservative. Do we have some secrets or some tips to share with our listeners? How to make innovation happen in such an environment nonetheless?

[00:14:23] Daniela: It's been quite a ride, in the last three years. I don't think there's only one secret. Every company has to find its own way of doing things, but for Merkur, it definitely was putting innovation outside the organization, and only hiring open-minded people, new open-minded people. I'm the only one who transitioned from Merkur Insurance Company to Merkur Innovation Lab, and my whole team had no notion of insurance or how is insurance done, what's possible, what's not possible. For my team, all is possible.

[00:15:17] Alexandra: Okay, that's interesting.

[00:15:18] Daniela: Yes, but it wouldn't have worked if no one came from the old organization, because I know the processes, I know the people, I know who to ask if I want something really quick. I know the domain. I've been in insurance since 2009, so I know a bit about the processes behind the data. That's one secret. Another secret, and that's really important, is the support inside, the old company. With us, it was the CEO.

[00:16:06] Alexandra: It's very important to have C-level support. Otherwise, things die down.

[00:16:12] Daniela: Yes, because with C-level support, you can buy some time. You have to show results quickly, but if you have a supporter, or various supporters in the management board, you have a bit more time showing the results. That was our luck. You have to become a bit of a salesperson inside the old company. The sales pitch. You have to pitch your idea the same as the other is pitching its idea for financing for the first seed round, or A-series, or whatever.

[00:17:03] Alexandra: That makes sense. Otherwise, it will be hard to also get the support in the different business units. One thing I'm wondering, though, if you mentioned that you're the only person that was brought in from Merkur, the mother group, how does it work with innovating in a sense that you also solve business pains, or help specific departments to find innovative solution to common pains?

Do you have a very interactive relationship with them in identifying potential business cases, and then also focusing on the most relevant, most valuable ones, or are you focusing more on this type of innovation where you say, "It's maybe not yet solving the most pressing business pain one unit has today, but it will be huge in the years to come."? How to balance these things?

[00:17:52] Daniela: Yes, it's quite a balance because it would be interesting solving problems which are only important in two years or in three years because they are much more interesting, but what we are doing right now, we are basically solving business pains. We are solving pains in sales and marketing in our health insurance department. We have quite a lot of work to do, because now, the department leaders, the people are coming to us, "Can you help us? Do you have the data? Do you have a solution?" We are not searching for work. [chuckles]

[00:18:37] Alexandra: It's coming your way.

[00:18:38] Daniela: It's coming our way. In the time we have sometimes left, then we are solving pains of the future.

[00:18:50] Alexandra: Understood. Since you mentioned solving business problems, I think this also ties back to what you said earlier in terms of you have to show results rather quickly. Of course, C-level support can help, but particularly when we focus on machine learning applications, AI applications and data analysis, here it's oftentimes hard to predict when exactly will there be certain results, plus oftentimes with legacy, infrastructure and data collection practices, there's quite some data preparation and collection that needs to be done, and sanitization of the data.

What's your experience in the environment that you operate in? Is it a big problem to get data in a quality that you can operate with, or is it something that you have solved in the years that Merkur Innovation Lab has been operating?

[00:19:41] Daniela: Yes. In the beginning, we thought it would be easy asking our IT, our data warehouse administrators or whoever, "Please give us the data, and we're doing our magic with the data." This didn't work.

What we are doing now is we are working along the whole data value supply chain. With our own data warehouse, we build it from scratch. We are extracting the data we need from the source databases. We are cleaning it, we are combining it, and in the end, we are doing analytics, but 80% of our work is getting the data, cleaning the data, understanding the data, because behind the data. There's a lot of business logic you have to understand. We are basically not only a data science department or company, but a data management company, but it's a lot quicker if we do it ourselves than asking someone for help and then waiting for two months.

[00:21:00] Alexandra: I can imagine. Then, of course, if you have this pool towards the data that you actually want to use as the motivation to have a faster turnover in terms of preparing the data, I assume is also something that comes easier. Thinking back to the several years that have gone by since the establishment of Merkur Innovation Lab, are there some special innovation projects that come to your mind where you had some valuable learnings, or where you learned some best practices that you could share with our audience today?

[00:21:28] Daniela: One of my big learning was that, yes, there's a hype around machine learning AI, and everyone has to use some machine learning model, I don't know, XG Boost or NL deep learning, but more often than not, it's only statistics, which is enough. We're doing a lot of advanced statistics. It's not easy, statistics, it's not only median and mode or whatever. It's advanced statistics, but very rarely we need a sophisticated machine-learning model. Our end users are quite content and they're happy with us. That was my first big learning because as a mathematician, I thought, yes.

[00:22:37] Alexandra: Finally.

[00:22:38] Daniela: Finally, new project, fancy algorithms. But no, no fancy algorithms, but only solving business problems using mathematics, but not very difficult type.

[00:22:56] Alexandra: I think that's a good approach to have, because what we see with some organizations is that they have their innovation department playing around with AI solutions, but not being aligned with the actual business problems, and developing something that's of meaning and of value to the parent organization. I think it's not about the technology. It should never be about the specific technology, but rather about the problems that you can solve, the value that you can add. Therefore, I think it really sounds sensible to approach it in that manner.

Coming back to the statistics that you apply, is there, of course in an anonymized manner, maybe, or can you walk us through a tangible story for our listeners about one specific project, product that was introduced or developed or co-developed by Merkur Innovation Lab, and how you manage them to also go through this change process, get the business units on board. I think that's always very interesting to hear, to pick some learnings from.

[00:23:53] Daniela: Yes, we are doing a lot for our marketing, for our marketing team. In Merkur, it was always the problem, which audience should we speak to, which audience should get a mail, which audience should get a push notification, or a letter, and in terms of GDPR, for which audience we are allowed to do what.

[00:24:24] Alexandra: Sure.

[00:24:25] Daniela: Now we've automated this decision using all the data so our marketing team can decide in seconds which campaign for which audience group, and with all the GDPR complications behind it. We are no longer targeting the wrong people in terms of audience group or GDPR.

[00:24:57] Alexandra: Infringement, so you basically have a way to get the right message to the right target audience, and at the same time, ensuring full compliance with GDPR, which I can believe is a huge time saver for your marketing team.

[00:25:09] Daniela: Yes, and another success story would be a machine learning success story.

[00:25:16] Alexandra: Oh, tell me more about that.

[00:25:18] Daniela: We did a machine learning model for our claims department. They had a huge model, how they can pay claims if this and that I think there were 150 rules. It was quite a lot of work maintaining these rules. Yes, we succeeded in reducing these rules to five rules and one machine learning model with a better outcome than 150 rules and a lot of maintenance. The business department is quite happy.

[00:26:04] Alexandra: I can imagine.

[00:26:05] Daniela: We are doing a follow-up in June.

[00:26:08] Alexandra: Interesting. Very curious about that. Since you mentioned GDPR, one other thing that obviously comes up whenever you want to innovate with data is actually accessing meaningful data for your machine learning models, for your innovation projects. Has this been a challenge within your environment in the past? What are the solutions to overcome this privacy versus data innovation challenge?

[00:26:31] Daniela: Yes, it was an obstacle for some time. I thought, okay, we can close our small company because we can't work with any data in Merkur Insurance. I've learned you can't work against GDPR. You have to work with the restrictions GDPR gives a data science team. It's possible there are rules, and the rules are there for a reason. Yes, we found a way of working with all data in Merkur Insurance Group, MOSTLY AI helped us. Sometimes it was the solution, sometimes it was a bit easier, but sometimes only synthetic data was the solution. We found our way around it, and we are doing innovation with the most sensitive data there is, health data. Thanks to MOSTLY.

[00:27:44] Alexandra: Very happy to partner with you here, since Merkur is definitely not only the oldest health insurance, but I can speak as a customer, also one of the most innovative and customer friendly companies. I'm very happy that we are partnering here in this space. I'm very much looking forward to what's yet to come. I think that's exactly the point that you make, that GDPR is here for a reason. I'm personally also happy that we have these strict privacy rules in Europe because there are ways to reconcile innovation with privacy protection, and why do you need to neglect privacy if there are options to combine both? Therefore, I'm also very passionate about the synthetic data industry, and being part of MOSTLY AI.

Can you tell me a little bit about your synthetic data journey so far? Obviously, many of our listeners are very curious to hear case studies from the practice. What were your learnings so far? What was the reaction of your data scientists and data people? Any challenges that you encountered when introducing synthetic data to Merkur as the parent organization?

[00:28:46] Daniela: Yes, it was quite a journey until we found the possibility synthetic data is giving us. First, we started with plain old anonymization, aggregating data, but it was never enough, because when is a data set anonymized? Big question. Our data protection officer always said, "Okay, but". It was always a but. And I understand his view. Then I said, okay, no more. It's a lot of work, but it's never enough. There's still a risk. There's still another possibility of anonymization. Yes. I did a bit of Google research.

[00:29:47] Alexandra: Always helpful.

[00:29:48] Daniela: Always helpful. Then I found MOSTLY AI, and the possibility synthetic data is giving us. Then I read a bit about how synthetic data is generated, what are the pros, what are the cons, and what are the possibilities, and what are the limitations, there are limitations.

[00:30:10] Alexandra: Sure.

[00:30:11] Daniela: Synthetic data is not solving all our problems, but a lot of them. We're using synthetic data for analytics. It's also working for dashboards, because the distribution of the data is still there. The outliers are cleaned, but that's okay. GDPR. Now we have dashboards with real data and with synthetic data. The synthetic data, we can use for external use cases. We can work with third parties. We are using synthetic data for training machine learning models, because training machine learning models with real data is sometimes not really allowed and often frowned upon by the data protection officers. I made our data protection officer very happy.

[laughter]

[00:31:19] Alexandra: It sounds like Merkur Innovation Lab is making many people happy. The marketing team, the claims department, the data protection department.

[00:31:27] Daniela: We're using the synthetic data to train the models. Then we are using the real data for the finished model, because then we would need the real data, but it's working, and synthetic data is enough.

[00:31:48] Alexandra: To train the machine learning models. Absolutely. What I oftentimes hear is that particularly when training machine learning models, it's anyway, rarely that you care about the one individual that has the super most unique disease in Austria, and more about generalized patterns, down to a granular level. Those are definitely well preserved with synthetic data, and therefore many of our customers say it serves as a drop-in replacement for the production data. Plus it's GDPR, CCPA compliant, and then they can actually use it.

You mentioned that when you did your research and now obviously you also gathered a lot of experience with synthetic data yourself, that there are pros, cons, opportunities, limitations. Super curious to dive deeper into those. What would you say are the opportunities for Merkur Innovation Lab and Merkur Group when you look into synthetic data?

[00:32:42] Daniela: I think one of the biggest opportunity is working with third parties. When speaking to other companies, not only insurance companies, but companies working with health data or customer data, there's always the problem, "How can we work together?" There are quite complex algorithms. I don't know, homomorphic encryption. No one is understanding homomorphic encryption, and it's not something which can be done quickly. Using synthetic data, it's a quick fix if you have a dedicated team who can work with synthetic data. It's not as easy as it looks because as one says in data science, garbage in, garbage out.

[00:33:40] Alexandra: Sure. The original data needs to be good in the first place.

[00:33:46] Daniela: Yes. It's a lot of pre-processing. You need a dedicated person who's good at working with distributions, checking, understanding the business logic behind. Working with third parties.

[00:34:05] Alexandra: Is a big one.

[00:34:06] Daniela: Yes.

[00:34:07] Alexandra: Curious to also better understand this point because we know whenever anybody wants to do machine learning, regardless of whether production data, synthetic data is the source or the training material, you need to have. You already mentioned it earlier, this 80% of data cleaning, collection, pre-processing, to have something meaningful in the end, and actually be in a position to do machine learning. Would you say that there's a lot which simply comes due to using synthetic data in terms of additional pre-processing? Or is there a big chunk which actually has to be done regardless of whether you use production data or synthetic data to have everything in a state where you can do machine learning?

[00:34:46] Daniela: It's both. You have to do the pre-processing if you work with real-world data. That's the same problem.

What's additional with synthetic data is you have to interpret the results, you have to check the result, because not all the business logic remains in the synthetic data. This we have seen in a lot of use cases, I don't know, we have some products which can't be sold to children under 18. The synthetic data algorithm doesn't know this.

[00:35:29] Alexandra: If it's a small sample, then this could be a problem, and requires to actually ditch the results where one product goes to. Make sense. What else would you say are skills that your team members need to effectively work with synthetic data? Did you do any training when you introduced synthetic data?

[00:35:47] Daniela: Yes. We did a lot of training. MOSTLY AI helped us with this a lot, with a monthly course, because sometimes we have questions and problems we don't know how to solve, then we need a bit of input. We did a month of training, we did a lot of testing, trying, failing. We failed a lot in the beginning, but now we've succeeded in setting up an automated pipeline, where every day, for certain use cases, our data is synthesized, and we have a data collection and new data warehouse, which only consists of synthetic data.

Now if we have a use case...

[00:36:48] Alexandra: ...we have the data.

[00:36:48] Daniela: We don't need a lot of pre-processing, but we take table, I don't know, 'customer', whatever, and then we can start with the algorithm the same as if it would be the real-world data.

[00:37:05] Alexandra: Makes sense. Significantly faster time to data. How long did you have to wait to comparable data prior to introducing synthetic data?

[00:37:17] Daniela: One month.

[00:37:22] Alexandra: That's a difference in contrast to everyday fresh synthetic data. Really cool to hear that this is implemented at this stage already. Coming back to you also mentioning the different limitations and benefits, what else would you see as pros and cons of synthetic data? Also, what are the problems where synthetic data is not the solution and where you need something additional?

[00:37:44] Daniela: Yes. For our management, for the department leaders, you would need the real-world data, because sometimes the distribution is not enough. You need examples, you need one customer with, I don't know, his claims ratio above, I don't know, 200% or whatever. For doing the analytics on a certain hierarchy level, you would need the real data.

[00:38:27] Alexandra: And some reporting where it's about the absolute revenue that a company does. Obviously, you want to have the exact number and not something that's very close.

[00:38:36] Daniela: Or for marketing, you would need my email address to be able to contact me, and not a fake email address. What's our limitation? It's the business logic. It's really the business logic, because sometimes you lose the context within the data. I don't know, people below 18 with this product, it's not possible, or now we have, I don't know, 18-year-olds with a PhD.

[00:39:18] Alexandra: Not that likely if it occurs for more than one child.

[00:39:22] Daniela: Yes. It's okay. You would need a bit of post-processing, and then the problem is solved. You lose a bit of complexity sometimes, but that's intended.

[00:39:40] Alexandra: Sure. Sometimes privacy goes against 100% accuracy. The goal is more how accurate can you be while still complying with GDPR and protecting the privacy of your customers.

[00:39:50] Daniela: Yes. It's a bit of no pro without con. Ethical AI is a huge buzzword. Discrimination within an algorithm is a huge buzzword. The Austrian regulator, FMR, is currently doing a questionnaire.

[00:40:23] Alexandra: How do you approach it, and how do you avoid discrimination. I'm super passionate about this topic. I'm actually also advising the European Parliament on the AI Act in this space. I think it's great to see the developments that are going into ethical AI, fairness in AI. Sorry for interrupting. I just wanted to share how you use synthetic data for fairness in AI.

[00:40:42] Daniela: I think it could be a solution to doing machine learning, developing machine learning algorithms with less bias. I don't know, minorities, gender equality. We are now trying to do a few POCs. How to use synthetic data for more ethical algorithms and less biased algorithms.

[00:41:24] Alexandra: Yes. That's exciting to hear. Actually, one of our other customers, the health insurance company, Humana, from the United States, I also had the pleasure to interview on this podcast a while back. They were also looking into fair synthetic data as a tool to help them debias algorithms, detect bias, mitigate bias. Of course, fairness is such a complex topic that you can't just take one solution. I also think that synthetic data has the potential to help a lot with certain problems in the whole fairness dilemma. Very curious to hear then also the results of your POCs.

There's actually one other story that I wanted to discuss in more detail with you. You shared in our preparation call that there's also a collaboration planned with an external company where synthetic data will be the facilitating tool to make this data sharing and data collaboration possible. Can you share more on this? I'm not sure how much you're already allowed to tell, but I would love to learn a little bit more for our listeners also today.

[00:42:23] Daniela: Yes, I can share a bit. Merkur Innovation Lab is currently working with another studio startup called Stryker Labs. They are quite successful in professional football, soccer, I don't know which word to use. They have collected a lot of data, how to train if you are injured, which position to play if you are injured. They collected a lot of medical data from scientific studies.

[00:43:07] Alexandra: Is it basically a training management tool that helps coaches and players to decide how to preserve health, how to recover more quickly and avoid additional injury, or how to better understand it? Okay.

[00:43:20] Daniela: Yes. It's a bit of a recommender for trainers in sports, what to do and what not to do, and if you do it, what would be the result.

[00:43:34] Alexandra: A little bit already in this proactive health care direction, and preserving player fitness and health. Interesting.

[00:43:40] Daniela: The idea is to use their expertise and our knowledge about injuries, the results, the medication, how long with which injury you have to stay in hospital, what's the prescribed rehabilitation, and so on. The idea is to use their business idea, our business idea, and develop a new one where the prevention of injuries is not only for professional sports, but also for you, me, the occasional runner, the occasional tennis player, the occasional, I don't know.

[00:44:27] Alexandra: Hobby athletes. That was actually my follow-up question, because I was super curious if this is something only for professional football athletes, or if every enthusiast and athletic person can get its hands on this solution and then benefit from it. Very, very interesting.

One other thing, since we also talked about fairness and anti-discrimination, I'm also personally super passionate to see where the health insurance and health care sector is going with data analysis, with proactive, preventive health care, because one thing that has been, as far as I understand, please correct me if I'm wrong, but one thing that has been a problem for a very long time in healthcare, is that so many studies are predominantly performed on the male body, and therefore, of course, our knowledge about health and preventive actions one could take are not on the same level for female individuals as for male individuals.

If we think of wearables that some insurance companies are considering to provide to their customers, the data that we collect from a much more diverse population and on a much more personalized level, I hope, in the future, will also lead to more equality in healthcare and to more fairness in the quality of services. What's your point of view on this, or how do you see the future evolve?

[00:45:47] Daniela: I am of the opinion that by using new data, I think women will be seen more in predictive health or medication. What we see in our data is, it's only medication sometimes for men which is used with women. You see that the medication is not working the same as with a woman than with a man. I think there's a long way to go, but by using new data, by using synthetic data, there could be a lot more fairness in our future models, future services. There will be services, especially for women.

[00:47:01] Alexandra: Oh, cool. That I think would definitely be a very welcome and needed edition.

Daniela, I think we could continue this conversation for quite a while, but we need to come to an end. Before we close, I would also like to ask you something slightly different, because I've seen that you're quite active also in supporting other female professionals, and you yourself being a successful CEO in the health insurance space. Do we have some words and advice for our female listeners? Which actions should they take to advance in their profession, or what actions should also allies, male colleagues, workplaces take to have a more inclusive and woman-supporting work environment? Very curious to hear your take on this.

[00:47:44] Daniela: [laughs] It's a loaded and difficult question to answer. [laughs] First of all, I think women need female mentors. I've always had male mentors. It's quite difficult because when they say, "My wife was a stay-at-home mom. I had the possibility to work, and do my thing, and do my career." That's the old generation. This is how it is done today in the year 2023. Female mentors, that would be my first tip, my first advice.

My second one would be, women have to begin building female networks. Male leaders, they have networks, a lot of networks. Clubs or they're going to events, there are always men, men, men, and only a small proportion of women. Men can play the game of networking perfectly, but women can't. There's a little number of female networks, and we have to begin building these networks, supporting each other. Then we would be a force to be reckoned with.

[00:49:27] Alexandra: Sounds like a plan. I think it's also the question about not being able to, versus not daring to, and not emphasizing networking as much as it's recommended, if I understood you correctly. This makes sense. I've also seen that you are an active mentor with, I forgot the name of the network, but if you can maybe also share with our listeners where they can reach out to you, where they can find you if somebody's interested, if there's a possibility to become a mentee, where to start.

[00:49:56] Daniela: I had the possibility to be at the Female Factor. Very cool set up in Austria.

[00:50:17] Alexandra: Absolutely.

[00:50:18] Daniela: [laughs] Vienna is also okay. It's quite interesting because sometimes women, they think they can't become leaders because they don't know this, they don't know that. They need this expertise and that expertise. Men don't have this problem. They are the experts, and if they're not, they're faking it. Women, sometimes they underestimate themselves.

[00:50:52] Alexandra: So more confidence would be beneficial.

[00:50:54] Daniela: Yes. More confidence, take on new challenges, and if it works, it works.

[00:51:00] Alexandra: Makes sense.

[00:51:00] Daniela: If not, what's the harm?

[00:51:03] Alexandra: [chuckles] Makes sense. Well, Daniela, thank you so much for being with us today. It was a real pleasure to talk to you, and as mentioned, I think we could have continued for at least one other hour. Thank you very much for taking your time to be in the Data Democratization podcast and sharing everything about healthcare and innovation in this space. It was a true pleasure.

[00:51:21] Daniela: Thank you.

[music]

[00:51:29] Alexandra: See, I didn't promise too much. It was impressive to talk with Daniela, and I loved how many things we covered, and how many tangible stories she could share with us. As always, if you have any questions, comments, remarks, just reach out to us on LinkedIn or via podcast at mostly.ai. Very much looking forward to hear your feedback. See you soon.

Ready to try synthetic data generation?

The best way to learn about synthetic data is to experiment with synthetic data generation. Try it for free or get in touch with our sales team for a demo.
magnifiercross