2020-05-31-marco-andreoni.mp3

Marco Andreoni: [00:00:00] The most important lesson is that adding a little bit of creativity, you can share 99% of the concepts, creativity in how to communicate the idea. I think flash statistics goes together with creativity.

Harpreet Sahota: [00:00:37] What's up, everyone,  welcome to another episode of The Artists of Data Science. Be sure to follow the show on Instagram @Theartistofdatascience. And on Twitter @Artistsofdata. I'll be sharing awesome tips and wisdom on Data science as well as clips from the show. Join the Free Open Mastermind Slack channel by going to bitly.com/artistofdatescience. We'll keep you updated on bi weekly open office hours. I'll be hosting for the community. I'm your host Harpreet Sahota. Let's ride this beat out into another awesome episode. And don't forget to subscribe, rate, and review the show.

Harpreet Sahota: [00:01:24] Our guest today is a statistician and data scientist based out of Monza, Italy. He's earned a bachelor's and master's degree in mathematical engineering and machine learning, as well as a master's degree in mathematics and cryptography. Since 2016, he's been part of an organization called Quantyca, where he works with data, covering every part of the Data lifecycle from ingestion, storage, analytics, web applications, cloud storage and beyond. He started at Quantyca as a junior business intelligence consultants and climbed his way up the ranks to become lead data scientist and he's now responsible for projects involving data analytics in general and machine learning with a focus on models and on the industrialization aspect of the work. When he's not doing data science, he's out there running 10 KS and marathons, juggling and travelling. However, you may recognize him from the work he's done as flash statistics, where since September 2017 he's created over 50 amazing artistic graphics, taking his audience on a painted journey across statistics. So please help me welcoming our guest today, a true artist in Data science. Marco Andreoni. Marco, my man, thank you so much for taking time out of your schedule to be here today. Man, I really, really appreciate it.

Marco Andreoni: [00:02:36] Hello. Hello, everybody. Hello, Harpreet. Nice to meet you. Thank you all for this opportunity to be here.

Harpreet Sahota: [00:02:43] It's my pleasure. I remember when I first started becoming active on LinkedIn and I was seeing your paintings come up quite awesome. And I was like, wow, this is really cool how you're just distilling these concepts down into such beautiful works of art. Man, it was really, really cool to see. I think your work has inspired a lot of other people to learn statistics and create these type of graphics themselves. Let's take us back to the beginning. You know, talk to us about your journey, how you first got interested in statistics, machine learning, data science? What drew you to the field?

Marco Andreoni: [00:03:13] To be honest, when I started my university studies at Polytechnic of Milan, you know, 2011, I didn't even know statistics. I have to be honest, I've always been good at math. That's why you have chosen mathematics and engineering, so simple. Luckily during the five years of the of my university, the first three years and then the last two years or so, I was really lucky, I had the opportunity to meet amazing professors. They share their knowledge with me and they turn me on onto statistics. And that's simply why I've chosen my measure of my master's degree in applied statistics. Really I wish you'd meet a person like them on your way because they are able to explain complex ideas so easily and to share passion about their job. So that's where everything started.

Harpreet Sahota: [00:04:10] We all had those teachers that really just inspire us to pursue some, you know, path of studies. So it's really cool that you had that experience. So I'm interested in about cryptography. That's something I've heard about, you know, maybe I've seen them on some movie. I think there's a movie that like Alan Turing or something. I don't know that dealt cryptography, you know, but it's something I've heard about. But I have no clue what it is. Can you give us an overview of what cryptography is? 

Marco Andreoni: [00:04:38] Let's say, to put this simple. In fewer words, you can say that cryptography is a tool to mask, to protect a messenger between a sender and the receiver. OK, so imagine you want to write to me a message, and obviously you don't want anyone else to read this message. OK. So you want to hide your message to protect it. And that's in a few words. OK, maybe the most famous and the most auction example of cryptography is the Caesar Cipher. I don't know if you know it, but it is a simple way to mask it message in which you exchange a letter using another letter and steps above in the alphabet. So let's say with an example, if you want to use the Caesar cipher with n = 5 okay, you will switch the letter A with the letter A,B,C,D,E, so with letter E, okay and so on. And so obviously this is a really easy, easy example, but it is good to capture the idea think.

Harpreet Sahota: [00:05:52] So how do you see machine learning and cryptography kind of inter playing in the near future?

Marco Andreoni: [00:05:59] I would speak about the future, but also about the present, because I know now privacy here with GDPR. And more in general, privacy in general  is crucial in data science because a data scientist will like to use every possible Data, but obviously not each single Data is available and is privacy compliant. Ok, so you need to just give you an example. I know about some projects where some data scientist starting with NLP models, but then they had to stop their project because the fields used by the Modeler were  protected by privacy. So they couldn't use these information in their model. OK. So I think the cryptography is an actual topic, not a topic for them for the future. You should always pay attention to mask your Data, to protect your data, and to understand which data can be put into your model to produce a valuable outcome I think.

Harpreet Sahota: [00:07:21] What's up, artists? Be sure to join the free, open mass much slack community by going to bitly.com/artistofdatescience. It's a great environment for us to talk all things Data science, to learn together, to grow together. And I'll also keep you updated on the open biweekly office hours that I'll be hosting for our community. Check out the show on Instagram @theartisticdatascience. Follow us on Twitter at @ArtistsOfData. Looking forward to seeing you out there.

Harpreet Sahota: [00:07:52] Just in case anybody in the audience is not familiar with the GDPR. And if you're not, you really should be. Do you mind giving us a overview of what GDPR is?

Marco Andreoni: [00:08:01] Yes. Let's  call it a set of rules involving customers data, especially here in Europe. But this involves also companies which are not based in Europe, but with customers which are based in Europe. In a few words, this is a set of rules which aim to protect customer data and to give customers the very all of their data. It's a long, long set of rules.

Harpreet Sahota: [00:08:36] Yeah. But the general gist of it, I think, is definitely something that every aspiring data scientist, every data scientist should be very cognizant of, especially in, you know, if you're going to e-commerce companies, things like that, most definitely need to be well aware of that stuff.

Harpreet Sahota: [00:08:53] Talk to us about the genesis of flash statistics? Talk to me about what was your inspiration for creating it?

Marco Andreoni: [00:09:01] This is a good question. And I know it sounds maybe strange, but I really don't remember the exact moment when I got the inspiration. It seems as many other cases, I started without a precise goal. It was like challenging myself to find a way to make statistics accessible for more people. And since I like drawing, the recipe was quite immediate. OK. Drawing plus statistics. This is the genesis of flash statistics.

Harpreet Sahota: [00:09:35] Talk to me about what you think the mission of a flash statistics is. You know, what's the one thing you want people to take with them when they come across your work?

Marco Andreoni: [00:09:44] If I had to compare Flash statistics to a company, maybe flash statistics is like a startup. It's in a starting phase. But I think Flash statistics could have several goals. Generally speaking, it aims to make statistics modifiable, but it can also be something more. I believe that seeing a flash statistics episode in the LinkedIn feed can be something a eye-catching and somehow relaxing because it is something so different from the stand of the LinkedIn feed. So you can just sit back, relax and watch an episode. 

Harpreet Sahota: [00:10:23] When you're creating the flash statistics episodes, and I know like sometimes when you're posting stuff out there for the entire world to see, for example, like on a LinkedIn feed, I may feel some internal hesitation or, you know, some type of fear. Did you feel any type of internal hesitation or a fear with creating the content? And if he did, how did you overcome it?

Marco Andreoni: [00:10:44] Well, I comprehend your point. But not really for me, because I started without any fear, because for me, that was just a game. Okay. So it was quite easy. But instead I was not sure about this thread of flash statistics. It's okay. I started it was a game and I didn't have some precise goals and something like that. But luckily I met Kate on LinkedIn. And not totally by accident I met her and she helped me a lot because she shared my content. And from that moment, everything changed. That was absolutely the turning point. From that moment, I had many subscribers to the group, the LinkedIn group and everything started.

Harpreet Sahota: [00:11:34] Just go for it, right? Not to be afraid, not to be afraid to build your brand and to create content and share it, you know, and share while you're learning. I think some people are very hesitant to post on LinkedIn. At least I know some people in our audience might be as well like has thing to share something because they don't want to look stupid, all go stupid I mean, is that the right word to say. But I'm glad that you

Marco Andreoni: [00:11:56] Also you are an example of sharing content in maybe in a different way because now podcasts are growing. But I think it's essential, because if  you don't put content or you don't create your own brand, you are just one in the crowd. It's a good opportunity.

Harpreet Sahota: [00:12:19] So what would you say is the most challenging part for you when it comes to creating the content for a flash statistics?

Marco Andreoni: [00:12:25] Probably this can sound a little bit boring, but it's all a matter of time. I mean, I usually take some notes about possible topics, and when I decide to create a new episode, I just sit down and I look for an inspiration. And then these can come from LinkedIn feeds, YouTube videos or lessons to learn from the past and so on. OK. So I need to convert an idea into the into a drawing, but then after the after their inspiration, the drawing part, let's say approximately takes between three and four hours. OK, maybe I split it into couple of evenings but and then I get the results.

Harpreet Sahota: [00:13:21] So do you have a personal favorite graphic from the archives?

Marco Andreoni: [00:13:26] Yes, sure I have. I think it is the episode number 12 or something like these relating to correlation and causality. I don't know if you remember this episode, but it is about this story of baby and Stork's. And I'm really satisfied, because of the story the concept and the drawing also. So this is my favorite episode.

Harpreet Sahota: [00:13:57] Do you mind sharing that story with us?

Marco Andreoni: [00:14:01] Yes, the story tells us about, there is a legend here. But I think also in the rest of the world. And these legends say is that babies, when a baby comes to the to the world, that is a stork arriving over at the home and leaving their baby. Ok, so the stork flies in the sky and brings the baby  to the parents. And so this is why this legend was born, because in especially in cold countries. When a family welcomes a new baby, usually a stork calm and the seats above the house. And so people associate the stork with the baby. Obviously, this is not true. OK, we everybody know where baby comes from. But simply, this is not a causality. So it is not the stork carrying the baby, but because of the presence of the baby, usually houses are warm. OK. And there is fire, and there is a good  situation where this stork can come and have a rest. OK, simply have a rest, babies are always associated with storks, but this is just a correlation and not causality.

Harpreet Sahota: [00:16:00] Are you an aspiring Data scientist struggling to break into the field? Well, then check out dsdj.co/artists to reserve your spot for a free informational webinar on how you can break into the field. It's going to be filled with amazing tips that are specifically designed to help you land your first job. Check it out, dsdj.co/artists.

Harpreet Sahota: [00:16:20] It's very good example. So we talked about your personal favorite. What do you think is one of the graphics, one of the paintings that is an absolute must for all the statisticians and Data scientists out there to check out?

Marco Andreoni: [00:16:42] Well, apart from Babies and Storks, if I have to choose one now, I'd probably say episode 16 or something like this. I remember the episodes, as you know, in a puzzle. Episode 16 is the one about P-value. I really hope that the weather example is playing in the episode can clarify the concept. It's  better to watch the episode, clarifying episode. 

Harpreet Sahota: [00:17:13] I think that is at least from my perspective, It's a concept from statistics that's not really well understood.

Harpreet Sahota: [00:17:21] What would you say is the most misunderstood concept from statistics and machine learning? And why do you think people are kind of confused or tripped up on that. 

Marco Andreoni: [00:17:32] The topic which is related to P-value, which is in general hypothesis testing. It's the first thing that comes to my mind. I really don't know why. But  many learners, get confused about this topic and really matches this point.

Harpreet Sahota: [00:17:51] Would you mind trying to kind of clarify or demystify that concept for us here?

Marco Andreoni: [00:17:57] Yes, sure. The subject is a so large, many different hypothesis testing methods and I think maybe I can clarify a point that I care most. Okay. When doing hypothesis testing, you'll never accept the null hypothesis. So why? Because the null hypothesis is true and less proven otherwise. Okay. So it is the basis. Rather you check if there is evidence to reject the null hypothesis. So when you make a hypothesis testing you take the null hypothesis and this is true. And you look for evidence to reject this. And if this is not the case, you don't say I accept H0 of the null hypothesis. But the correct expression is I don't have enough statistical evidence to reject the null hypothesis. That can sound a little bit tricky, but it's very important because the null hypothesis is something that is true unless proven otherwise and the you need evidence to reject the hypothesis. This is an important concept of statistics.

Harpreet Sahota: [00:19:28] Definitely, yeah. And I think one way students can think about it, especially if you're in the USA. In the USA, they had in the court system, you are innocent until proven guilty. So the same concept right, the null hypothesis in that situation will be that the defendant is innocent unless we have sufficient evidence to claim that they are not in fact innocent.

Marco Andreoni: [00:19:49] I don't accept you are innocent, but I  do not have enough evidence to say you are not innocent.

Harpreet Sahota: [00:19:58] That's a great point for hypothesis testing. I think a lot of people don't realize or maybe they do realize, at least in my experience, people maybe don't appreciate the fact that even A/B testing is just a simplified version of a hypothesis test, right? Yes, exactly. That's exactly that. So you'd be testing as just a hypothesis testing. And even then when you have these hypothesis testing that you're doing, there's this really serious errors that you can commit as well as type one error, type two error. So those are very important considerations to make when you're designing an experiment. I would like to get your point of view on this.

Harpreet Sahota: [00:20:35] Do you think it's important to learn all the formula and equations even though we have advanced software that doesn't work?

Marco Andreoni: [00:20:43] I don't think so. But let me explain better the idea. I don't like the philosophy of, OK, just take this model as a black box, push the button and see what happens. OK.I strongly believes that you have to know what to use. So you don't need to memorize every single equation. Every single formula and everything. But you must know the underlying idea. This is the important thing. It's the support.

Harpreet Sahota: [00:21:15] Do you have any tips or any good ways for somebody to learn the underlying idea behind what the software is doing?

Marco Andreoni: [00:21:23] Someone can start from flash statistics, if he really wants to learn the basics because let's say machine learning is an evolution of statistics. OK. So I think you have to start from the basics to understand the concept. Then watch all the available software libraries and tools and so on. Obviously, see the documentation, watch some tutorials. But you don't, this is important for me. You don't have to take the library, push the button and see the outcome. Start from the documentation, understand what's beyond and go ahead.

Harpreet Sahota: [00:22:10] Yes, the excellent point is just to look at the documentation. Because if you go look at the documentation, they'll also cite references and they'll cite other papers and other works you can go look at to get a more in-depth idea of what's going on. Then also, you just look at the source code and kind of see under the hood what's going on.

Harpreet Sahota: [00:22:27] Do you consider Data science and machine learning to be an art or purely hard science? And why?

Marco Andreoni: [00:22:34] Good point. To be honest, I hope there will be soon consider more as a science in terms of reproducibility and scalability. I mean, a so-called data scientist needs to be an artist in a sense. Okay. Because first of all, he must be competent, but also creative and communicative. OK, so I can ask this. However, if he doesn't follow rigorous and the scientific approach also, that are not only from science, but also from common delve up idea. He will always work on experiments or proof of concepts instead of production projects. I think in this period data science is used more in art, it's not science. It's quite strange because it is called Data science. It is not so science. And it needs to become more scientific, more rigorous, more with more reproducibiliy and scalability.

Harpreet Sahota: [00:23:39] That's an excellent point about the reproducibility and, you know, being able to put the models into production. I think a lot of people in their first starting out learning are working out of their notebooks, they don't have much experience on the other side, putting it into production and then monitoring and evaluating post-production. 

Harpreet Sahota: [00:23:57] With the experience you've had working in industry, what are some things that we should be cognizant of? What are some things that we should be aware of when we had deployed some model into production?

Marco Andreoni: [00:24:10] Even before deploying a model into production, the first point that a good part of data scientist needs to know is that CSV is not the source of everything in the world. So in real context, it's not always possible to extract some Data and stuff from a CSV and then put it into a notebook and so on and so on. So that's a first point. Second point is that there are, I think, a couple or three aspects that are important. First of all, you need to know which model, which version of your model isn't currently in production. OK. And there are some tools such as ML Flow or something like that, or AWS Sage Maker works with this idea, with these tools, you can know which which artifact, which version of your model is currently in production. Because if you don't know which is working for you, who is working for you? You can't understand it. OK. So maybe you you start from version zero and then you upgrade your model, but you always need to know which version of them that is in production, first point. Second point, a data science project that is composed not only by the code, but also by the data and the Data in a real company, in a real world, Data are continuously changing. So you need in some occasion you need to version also your Data. And there are a lot of tools and a lot of ideas, on how to version. This simplest one probably is DVC, but there are also other tools. And the third point and in AWS Sage Maker, you can achieve these with the CloudWatch. You need to control your model, you need to have some metrics, you obviously the model changes during the time. So you need to monitor it and to analyze the performance and maybe they work and so on. When you started this project there, it will never end because you have to iteratively start again and improve and so on. It's like that the science and the lean manufacturing goes together. It's the concept of continuous improvement in a model.

Harpreet Sahota: [00:26:44] If you don't mind. I'd like to dig deeper into some of these get your expertise and your insight as to this, because I think a lot of our audience would definitely enjoy this as well. So we mentioned the three aspects. We need to version our model and version our Data and then have sufficient evaluation procedures in place once it's kind of released. So can we talk about each point in terms starting with the model versioning...

Harpreet Sahota: [00:27:11]  Why is it that it's important to the version our models?

Marco Andreoni: [00:27:15] Because simply, here we are touching again at reproducibility. OK. If I put a model in production, but I don't know which version is currently in production. And I get the results, how can I say OK, this is the expected result or not. I need to control my code  such as a standard a web application. OK. I need to know. OK. I have a repo with the code. And now I push my latest modification and now this code is taken, it's compiled, I can build an artifact, put it into production and then. OK. This is a new version of my model, so from now on I will get predictions from these new version. OK. So probably prediction. From now on will be different from the prediction of previous day. And so I perfectly needs to know when a version of the model is in production and when these changes, these can be quite dangerous sometimes.

Harpreet Sahota: [00:28:20] This is the idea of concept drift right? Like, you know, if you create a model, you put it into production, the model has some effect on some downstream behavior. So when you when you create the model, initially you were modelling some type of Data generating process. And then just by the fact that you have implemented a model, you've altered that Data generating process. Right? Now you need to come up with a new model to now model this new thing that you've created.

Harpreet Sahota: [00:28:47] So then versioning Data, what's the importance of that?

Marco Andreoni: [00:28:51] Yes, the idea is that maybe you can have a very nice model and maybe you can have a very good model. And yes, suppose that we are in the same data science team in a company and I will try you in an experiment, I prepare my code. And I say, OK, hey, take my code and watch my result. OK.Then you take my Code, you clone the repo, and you run the experiment. Then maybe you would come to me and say, OK. But why you told me that your accuracy, let's say was 99%, I run the experiment and the accuracy was 80%. It's strange, the model is the same. OK. Because also data are changing, OK. Especially if you don't work with CSV for a proof of concept. OK. If you work on real data. So you need to control your data, and to know, OK, I have these snapshot of data and I will run my experiment on these snapshots and then I'll get some results on these snapshot. So my experiment can be reproduced by anyone else. And this is science in a sense. That's the previous point. OK. When  a scientist prepares an experiment, he has to describe every condition. OK. So that the experiment is reproducible. And that's the same concept I think.

Harpreet Sahota: [00:30:26] And I think in this case, that idea that is here is called the Data drift right? So that's something that we need to be aware of. 

Harpreet Sahota: [00:30:33] So when we're looking at evaluation metricss and stuff like that for model post-production, what are some that our audience should do research and go learn more about?

Marco Andreoni: [00:30:44] I think that monitoring the model is like drawing a car with all these lights that you can see in your car. And obviously, you want to know if you are running out of fuel. And if your engine gets some problems and something and so on. The model is exactly the same idea. If my model is really important for my business and for my company, I needs to monitor it and to analyze how the model drift in there, how this score changes over time. Because only if you measure something, you can control something. Otherwise we are again on  the black box. OK. You put the model in production. And good luck. See what that best, it's really, really important.

Harpreet Sahota: [00:31:38] Thank you for that. I know we went a little bit off script there, so I appreciate you sharing your insights. I know our audiences are going to enjoy that.

Harpreet Sahota: [00:31:46] What role does being creative and curious play for being successful as a Data scientist? And how can someone who is who doesn't see themselves as creative actually become creative?

Marco Andreoni: [00:32:00] The answer is so easy, I think. Creativity is crucial. I think there are at least two moments when it comes into play. First, the communication of results and secondly, the choice of their model, more generally of the approach, not also of the mode. Because if you think about it, this set of available algorithm is not infinite. OK. Be sure that not every data scientists would choose the same algorithm. So you need the intuition and creativity, all these together. And if these are not your characteristics, I think the best advice is observe the experts. And if you don't know one of them, you can just scroll the LinkedIn feed, open some links and go with the flow. As we said before, LinkedIn is a good opportunity because you can find experts. Maybe it's quite if you think about it, we are having these interview and we are in that so different places. But we can get in touch and communicate and exchange ideas. And so it's a very good opportunity. So  if someone doesn't see himself creative, it's not the problem you have just to watch, try and repeat, there are lots of experts on LinkedIn. It's an El Dorado of knowledge.

Harpreet Sahota: [00:33:31] So I'm reading a book right now by somebody called Chase Jarvis. And he wrote this book is called The Creative Calling. And in there, he describes creativity as simply taking two old things and then putting them together in a new way. So how does this boil down to your work as a data scientist? Well, you could take, for example, a research paper somebodies approach. You could take your problem statement, draw parallels between the two and combine, you know what they've done their approach, take elements of that, take elements from another approach and apply it to what you're doing. And all of a sudden, you know, you don't have to reinvent something from scratch the ground up. You do research, right? You create a new solution and effort. I forget who said this quote, but it was, if you copy from one source, it's called plagiarism, but if you copy from multiple sources, it's called research.

Harpreet Sahota: [00:34:28] So you mentioned another couple of interesting points that I would like to touch on. One of them was the fact that when we have our results, it's not strictly just from the model that we've employed, it's the result from the entire process, I guess. So that that metric that we observe, it's not just because the model Data performance is actually a reflection of every choice made in the process. Right?

Marco Andreoni: [00:34:49] I really like a quote which says that I think it is from Bill Walsh, which is a football trainer coach, and the quote says that; focus on the process, the result takes care of itself. So the point is that I don't have to focus on the results and say I need to do push a model to the top, but I have to focus on the entire process, which is what they can control, obviously, and then controlling the process probably won't get resolved.

Harpreet Sahota: [00:35:24] I 110% agree with that. It's very important that you just focus on the things that, like you mentioned, are within your control and ignore everything that you have no control over. That's a very good recipe for a much happier life and less stress for sure.

Marco Andreoni: [00:35:40] Less stress. Exactly.

Harpreet Sahota: [00:35:41] Focus on optimizing the things that are within your control. When it comes to communication, I think, you know, a lot of Data scientists are very technical. We like to kind of do heads down work on our quantitatively rigorous stuff.

Harpreet Sahota: [00:35:57] Do you have any tips for data scientists on how to effectively communicate?

Marco Andreoni: [00:36:00] I understand why some  data scientists act like that, because if you think about it, the process of constructing a model, a good model is very tiring in a sense, because you have to take data, understand the problem and try and try. It is like in my case is like preparing a half marathon, you have a long journey and then you get to your result. And when you get to the result, obviously, you want to share with other people, your journey in a sense. So, okay, look, I decided to use these transformation and I pick these model and and so on and so forth. But you need to understand that, first of all, maybe the audience can be non-technical, even if the audience is interested in your journey, but simply the audience cannot understand it. And so it's useless, your explanation is quite useless. And the second point, the audience is more impressed about the result, then the process, which is not so good. I understand that we have said that the process is the important part. But obviously, they are businessmen, so they they need the numbers, they need the results. Try to clean your explanation and leave just their main points. I think if you are able to describe the process without going in technical aspects, but just transmitting the process, these can really help non-technical people to understand the work beyond the model, the result. They will be happier because then they said, OK, we have these results, but we can control it, we have the entire model, we can and maybe they will get curious and go on flash statistics web site, understand some statistics and so on. So  I think the best advice is start simple and describe the process but in a simple way, just the main steps of the process.

Harpreet Sahota: [00:38:12] I 100% agree with that. Yeah, don't try to confuse your audience into silence, so you don't have to have any questions asked. That's a smart approach.

Harpreet Sahota: [00:38:22] What would you say are some of the similarities and differences in the creative process for writing up a publication or reports, doing a data science project, or creating a flash statistics painting?

Marco Andreoni: [00:38:34] I don't want to say obvious differences because, OK, let's say statistics and not so scientific. OK. Obviously they are different. I'd rather point out the common feature I think between them. The key point is that you must be able to adapt your communication style to the audience and to the context. So if you are able to adapt yourselves to different targets, it's not important that you are doing a project research, a machine learning data science project, a flash statistics drawing. But if you can adapt yourself, you will be successful in all of these three fields. Again, it's a method of communicating and communicating in the right way for the right target, I think.

Harpreet Sahota: [00:39:24] Last question before we go into the lightning round. What's the one thing you want people to learn from your story?

Marco Andreoni: [00:39:31] The most important lessons. That's the lesson that I learned and so I'll want to share is that creativity makes it easier to pass both a simple and complex concept. So adding a little bit of creativity, you can share, I think, 99% of the concepts, flash statistics is an example because it deals with hard topics sometimes. But it's not a matter of appearance, creativity in how to do communicates the idea. I think flash statistics goes together with creativity, which is the key point.

Harpreet Sahota: [00:40:12] So let's jump into lightning round here. So, well, what do you see flash statistics becoming in the next two to four years?

Marco Andreoni: [00:40:18] Maybe I'm quite the dreaming. I see flash statistics as a comprehensive statistics platform. Okay. This means in an interactive place where you can learn and understand. So not watch but understand the statistics at a different level. Okay. I imagine there will be lot of flash episodes together with one or more books. I'm working on it. And except maybe on video lessons. I really, really hope this can help people to appreciate statistics. This is the mission.

Marco Andreoni: [00:40:58] I'm looking forward to that look, man. As we mentioned in our introduction, you love to travel, what's your favorite country to travel to?

Marco Andreoni: [00:41:08]  I really, really love Italy, for example, during the summer, Sardinia I think is a top, but maybe the last country I visited is London and I really, really liked it. It's an amazing city. I went there five days for a conference, amazing, simply amazing.

Harpreet Sahota: [00:41:39] So what's the number one book, fiction, non-fiction or both that you would recommend our audience to read and your most impactful takeaway from it?

Marco Andreoni: [00:41:47] I will answer with no doubt, Serve to Win by Novak Djokovic. I don't play tennis, I play tennis with my girlfriend, but I'm not an expert. But I find Novak really inspiring because of his attitude. And I don't remember the exact words, but his book contains a fantastic sentence in which he says that every single moment of his day is completely focused on being the first player in the world. And obviously, this can be applied to everyone's passion or interests. It's a method of dedication, focusing on something, focusing on your passion and this book is really full of inspiration. And I recommend it to all the audience and especially to you. I don't know if you really know it but it's amazing.

Harpreet Sahota: [00:42:59] I'll definitely check it out. Like any book that my guests recommend I end up buying it. My library has been full recently. I'll definitely check that out. And like all these books by these athletes are really good because that mentality that they have, that mindset that they have cultivated for themselves and there's common threads that you pick up. I've read a few books by some athletes. It's just that mentality that they have of excellence. And that we can all.

Marco Andreoni: [00:43:32] And an important aspect that I learn from from athletes is that like, if you consider, for example, flash statistics or your podcasts, from the outside, that you always see the results. OK, also you see Novak Djokovich OK, he's the number one player in the world, he's a gifted. OK, but he's not only gifted, he has a very, very, very strong mentality. And also beyond the flash statistics that is a lot of work and beyond podcasts, there is a lot of work. And they have a very, very, very concrete, practical example of this. And they are amazing athletes and really inspiring.

Harpreet Sahota: [00:44:27] Definitely. And so if we can somehow get a magic telephone that allows you to contact 18 year old Marco, what would you tell him? First, you'll tell us eighteen year old Marco, where were you? What were you up to? And what would you say to him at that point?

Marco Andreoni: [00:44:40] I think I was here in Monza and I was studying and nothing special unfortunately, but don't blame me, I wouldn't tell anything because I strongly believe in the power of experience. OK. And so I would let him take decisions based on his current situation. And I don't want to cheat let say. And yes, I think I wouldn't tell him anything.

Harpreet Sahota: [00:45:21] All right. So if you could put up a billboard anywhere, what would it say and why?

Marco Andreoni: [00:45:29] Give do ways to things. That's not always so easy. And it's quite relieved to what you previously said but this really helps you to focus on what really matters and without losing energy in useless issues and something like that. So for me, my motto, let's say, is give do ways to things.

Harpreet Sahota: [00:45:56] I like that give do way to things. What motivates you?

Marco Andreoni: [00:46:01] I think working with skilled and qualified people interested in sharing knowledge, these really pushes me to give my best. I don't know if you know the quote that says if you are the average of the five people, you and I really like these this quote and this can motivate you a lot.

Harpreet Sahota: [00:46:27] What song do you currently have on repeat?

Marco Andreoni: [00:46:31] Let me check my spotify playing list, blinding lights by the Weekend. 80s style song.

Harpreet Sahota: [00:46:41] Okay. Nice. Nice. So, Marco, how can people connect with you? Where can they find you online?

Marco Andreoni: [00:46:51] Well, LinkedIn is probably my main platform, and they can always get in touch with me. They can also join the Flash statistics LinkedIn group, we are more than 3000 and so they are working. And on also the Instagram page @flash.statistics.by.Marco. Or they can easily check the websites; Flashstatistics.com, that's the center of flash statistics.

Harpreet Sahota: [00:47:19] Marco, thank you so much for taking time out of your schedule to be on the show today. I really, really appreciate you coming on and sharing your perspective and sharing your insights with everyone. Thank you so much.

Marco Andreoni: [00:47:30] Thank you so much for the opportunity. And this time was really fast and really nice. Thank you very much.