 So good afternoon, everyone. I would like to start by saying thank you to the organizers and also to the sponsors for making this event possible. It's a pleasure for us to be here with you today to talk about business and to talk about how we can make business decisions based on the usage of mobile data and also mobility data. So let me introduce myself first. I'm Darifio Reglesias. I'm the big data analytics tech lead for Vodafone business. And here with me today is Javier. Yeah, hello, everyone. I'm Javier Perez, and the head of data at Carto. I think before jumping on the specifics of what Vodafone and Carto are building together, I think it's best to introduce also what brought us here. And we can start by the fact that we still see a lot of institutions and a lot of companies that base their decisions using census data, which is by no doubt a very important source of information. But as we know, the census only gets updated every so often. So what you do if you want to use census data today is that you take the latest publication, then you project it to the present, and then you use it. And with the past of the years, these projections start to become less and less accurate. If we look at some of the major economies in the world, we see that, in average, and also weighted by the population, census data is seven years old. So the last publication in the US was nine years ago. In UK, Spain, Germany, and Italy, it was eight years ago. So at the moment, it's quite old. At the same time, and I think we can all relate to this fact, 90% of the people have their mobile phone within one meter reach during the 24 hours of the day. So it's with us when we work. It's with us when we are watching the TV and we are looking at two screens at the same time. It's with us when we run. And it's next to us almost when we sleep. So it's basically becoming an extension of our bodies. And this is not going to change. This, in fact, is booming. So if we look right now, there's more than 5 billion people that is using mobile phones to tweet, call, text, navigate. This means that all the traffic that has been generated with these devices has grown 18-fold in the last five years. And if we look at the near future in 2021, it's expected that it's going to be 11.6 billion devices generating mobile traffic. So not only mobile phones, but also IoT devices. But this is by several billions more than the people that is estimated to be on the planet by 2021. But in order to make sense of all this data, so the data coming from the mobile phones, the data coming from connected cars, sensors, machine devices, we need to put them into context. And one of the key core components of all this data is location. Because at the end, as we like to say a lot in Cato, everything happens somewhere. So imagine that now all these institutions, all these companies, instead of using data that is seven years old, start to use and base their decisions with data that is updated on a monthly, weekly, daily, or even on real-time basis. With this ambition, it's why Vodafone and Carto entered into a partnership in 2017 to put into the market Vodafone Analytics. There is a data product and a suite of interactive web applications that allow public institutions and private companies to generate insights to take decisions using human mobility data. In this video, we're having, as an introduction, you can see the city of Madrid that has been gridified in cells of 250 by 250 meters with data aggregated over a month. And the user can filter by day of the week, or week, so weekday or weekend, time of the day, also filter by demographic profile of the visitors, look also aggregated the origins of these visitors. So this is just one example. And today and in the rest of the session, we're going to enter in details of how we have built this. So with the ambition of creating something differential that we can offer to our clients, one of the very first things that came into our mind was to create a global product. And being Vodafone a global brand and a global telco allow us to create these kind of products and services that we can offer to our customers and to our clients at scale. And in fact, the product Vodafone is already available in four different countries. And we are working on increasing the number of countries in the search. We need to consider that we have more than 100 millions of mobile devices connected to our network. And also, it's important to know that in most of the countries where we are present, we have a very good market share, allowing us to create these kind of products. But it's not just important to be a global telco, but we also within Vodafone, we have some of the main ingredients that allow us to create this product. And it's not just having a very good network, the best network as we do, but also having the skills and the tools to store all this data from the network and to process it at scale. And it's also important, one of the other ingredients is to have the ability to work in a multi-cloud environment. Later, we will talk about how we are using multiple clouds because business requirements and also, in some cases, legal requirements. And probably the most important thing is to have people with the right skills, like our data science teams, that are working to generate or to extract the max out of this data. We are working with global and local data science teams. And we are using one or some of the best tools out in the market to process all this information, like Carto. So Carto, we are leading the world of location intelligence, thanks to a team of more than 130 geospatial experts, engineers, and data scientists. Everything, in fact, started here in Madrid in 2008 as a data visualization boutique that then, in 2012, was turned into a product company. Now, we are headquartered in New York and we have major offices in Madrid, Seville, and Washington, DC. We are backed by investors like Axel Partners and Salesforce Ventures. And we have a customer base in the order of the thousands of proving this growing trend in the market on location intelligence. So our goal with our technology is to allow companies to generate business outcomes, business insights with the use of location data and following five key steps. So the first of all is how our technology integrates with the databases, the data infrastructures of our customers. Second is then how our customers can enrich their own data, so the data that they generate on their business, with data coming from third-party partners like Vodafone or OpenData sources. Third, how you can run geospatial analysis on top of the platform using the combination of all these multiple data sets. Then, build interactive web applications, dashboards, and visualizations, normally on maps, using the results of those analysis. And finally, how this integrates with other tools or as an enterprise software solution in the day-to-day operations of our customers. Carto, as a technology, can be seen as a full stack for location intelligence. So it can run on the cloud or installed on premise in your customers. And then everything starts on a database that is geospatial by design that is connected to third-party data and third-party services like the calculation of routes, geocodes, and isochrons. Then a set of APIs that allow you to query and run analysis on top of the database, an organization of libraries to build the visualization that then end up being one of our solutions or one custom application for one customer, or, for example, the Vodafone Analytics app that you just saw. So if we start talking about actually about the product, one of the main things that we needed to take care was about privacy. So one of the first things that we did when starting thinking about creating Vodafone Analytics was to start engaging with the privacy and security teams, both globally and locally within Vodafone, to ensure that we are taking care of the privacy of our customers by design since the very beginning. It was quite easy because protecting the privacy of our customers is on our NDA. So with their help, we defined a set of mechanisms or procedures that we apply in order to ensure that we are keeping this privacy under control. The first one is anonymize all the data we use, all the data that is extracted from the network and it's ingested in the platforms we use to analyze and to extract patterns and to extract business insight out of there. It's anonymize, meaning that there's no customer identifier within the data. Second thing is about offering the right to our customers to opt out from this analysis. And we do it in a very simple way by allowing our customers to configure their preferences on my Vodafone app. So it's an easy step. And the third thing is to ensure that all the insights are generated from aggregated and extrapolated data, meaning that we are not analyzing individual behaviors. We are not analyzing individual patterns. We are always analyzing aggregated behaviors. And one of the important things about that is that, in fact, we were awarded because of this privacy approach. We were awarded by the International Association of Privacy Professionals, which is an honor for us. So now, if we start talking about the data we use to create this product, Vodafone Analytics, how big is the data we use? So these are just numbers from Spain. We are working in this product globally. But we have more than 30 million mobile devices connected to our network, which is around 20% of the total population here in Spain, generating more than 10 billions of geolocated events every day. Those devices are connected to more than 200,000 network cells. And all this information is projected to more than 8 million locations, 8 million locations like municipalities, provinces, or grids. So to give you an idea of the size of the volume of the data we use for an average project, it usually requires us to process more than 300 billions of geolocated events, more than 300 billions for a single report. And from where are these geolocated events coming? They are coming from for different data sources. CDRs and XDRs are probably the most well-known in the industry and the most common ones. That's what we usually call active events, because they require the user or the customer to generate some actions, like, for example, sending an smh, starting a voice call, or starting a data session. We also have a different data source like NetPerform. NetPerform is an app tracker that allows us to get more granular information in terms of location, but with a lower sample of the population. And the last one, and probably the most important one, is network groups. They are generating most of the billions of events that we collect, and that's what we usually call passive events, are passively collected out of the network and stored into our platforms. So how we process this incredible amount of information? As I said at the beginning, we are working on a multi-cloud environment. And this is not just because it's fancy, but also because it's a business requirement in some countries that are different business needs. So as we are treating Vodafone as a global product, since the very beginning, the product was designed to work on a multi-cloud environment. We are using Vodafone Private Cloud, and also two of the main public clouds out in the market. And within that clouds, we have developed our own execution framework. That allows us to perform all these geospatial aggregations and transformation at scale. It's called Falcon, and it allows our data science team to develop code on PySpark and also on Scala. It's really good for them because all the data access layer is encapsulated, so they don't need to take care about how to read data out in a particular platform. They don't need to take care of these kind of things. And also, it's really good for us because it allows us to incorporate new features in a really easy way, and at the same time, it's easy to debug and to test. And this is a really super simplified architecture that we are using. Starting from the left, all the data coming from the network is flowing into the different platforms in real time by using Kafka. Kafka because the network tools has the connectors to extract the data to Kafka topics. Within each platform, we use a set of, let's say, simple or well-known applications or services, like EMR, Dataproc, or even within both of them private cloud data cluster. And every time we receive a request from a customer, we execute a set of batch processes to transform, aggregate all this data and to extract them into the BodaFone Analytics visualization layer that is powered by Cartoon. So once BodaFone had the data anonymized, aggregate it in a space because all the data comes from some sort of polygon. It can be a grid or it can be like a postcode. And extrapolated not to represent the BodaFone sample but the overall population, it arrived to Cartoon. And one of the first things that we did back in 2017 with the team, between the team of Data from Cartoon and the big data team at BodaFone was to design the data structure, so the data model for us then to build the applications on top with the fact that we wanted these applications to be able to answer some of these key questions. So for example, the who. So being able to look into a postcode and try to understand the patterns from a demographic profile of the visitors. Also the when, so allowing the user to be able to understand the differences in patterns between weekdays and weekends, between winter and summer. Then also the where. So being able to cross reference the data that BodaFone provides with the location of specific points of interest like a concert venue or some place with touristic attractiveness. Also then the how, being able to understand in general what's the common travel mode between one origin and one destination. And then also the why, so looking in historical time and trying to see in this postal code. So in average how many people come here to for example commute to work or just to spend one day or to be there, normally they just pass by and some others tend to stay for three, four hours. So trying to give answers to all of those questions. So when we had the data model is when we started designing web applications on top of that with four key verticals in mind. One is retail. So allowing retailers to correlate and analyze the performance of the stores with what BodaFone provides on the data. So are the volume of visitors over time correlate with the sales that I have in this point. Trying to compare the behavior that we see in the visitors in my stores with the locations where I know a competitor has a store. Also then trying to infer from that what is the trade area, the catchment area looking at where most of the visitors come from when they go around your store. And then also trying to look for similar patterns in areas where you don't have a store and you are considering to open one to say, okay, so my top performance stores behave from the point of view of BodaFone Analytics on this way. Tell me in this other city in which areas I can find this type of patterns. Another vertical then is tourism. So in tourism is more for public institutions and regional governments that want to understand the patterns of tourism in the different parts of the region of their city and how it changes over time and how it changed if they did a specific event. Something like that. Then we also have outdoor, out-of-home, outdoor media to help marketers to be more efficient in the way they plan marketing campaigns with billboards, understanding the mobility patterns in the different areas of a city. And then mobility, also to help urban planners and cities to understand the impact that certain events have on how people move around the city. For example, if you put, you organize a big conference like this of how this has affected mobility in Madrid compared to last week or the week after. In this video here, what we have is one of these interfaces. This one is very much focused to retail. So you can basically look or load your stores and the stores of your competitors. Click on those, look for an address and geocode that address into a point and then create an isochron or a driving distance or walking distance buffer around it and then start to look what the data says. And we have integrated here data from Vodafone but also data from other sources like the sensors, financial data, points of interest and so on. And yeah, so you can then generate reports that characterize the surroundings of that location and also compare with another location. Compare with a location of a competitor or just click in an agency. Okay, so if I move the store from this number to 100 numbers going up, what changes I can see and also what changes I can see contextualized with the average of that municipality, which is always very important. So this is another example and this is the product for tourists in which we can see Andalusia region divided in municipalities and we are comparing the number of visitors against two different months, September and August, I think. We can see in this case a decrease in September compared with August near the coast. It is also possible and this is really powerful to focus or to analyze a concrete type of visitor. In this case, I think we are filtering just to analyze and to extract information out of the sporadic visitors, both international and national ones. And we can check the number of visitors on all the municipalities. It is also possible to select one of the municipalities, in this case Seville, and analyze the origin of the visitors to that particular place. Are they coming from another municipality? Are they coming from a different region? Even are they coming from different countries like Portugal or Morocco? And at the left, it's always possible to check the demographics of those visitors. It is also possible to analyze and this is really powerful to analyze the evolution of all of these metrics over the time by checking this historical series. So, as I said at the beginning, we have this product available in four different countries, Spain, Portugal, UK and Italy. And there we have a couple of the most important clients for us, for example, Junta and Dalucia is using this to analyze the tourists they are residing, like in the previous example. OBS is a really high retailer, a really big retailer in Italy. They are using this solution to analyze the visitors and the visits to their stores and also to their competitors. We have been working also with other public administrations like Transfer for London. JLL is an interesting one. JLL is a real estate consulting company and they are using Vodafone Analytics combined with older data sources and also combined with their business knowledge to offer their clients better results compared with older competitors. And also an interesting one is MTV. MTV used this solution to analyze the event they organized last year in Bilbao, the MTV Europe Music Award. So it's possible to analyze the effect of an international event like this one in a particular city. Also the collaboration between Vodafone and Carto, what allows our customers is to not only visualize data on maps, but also to move, to go one step ahead and start analyzing data using maps because it's not the same to answer the where. So in this city where tourists tend to go more versus trying to understand the why. So trying to uncover the patterns that explain why a group of tourists tend to behave in a certain way and try to be more efficient with that information. And this is very much related to the fact that it's not the same to analyze this type of data using BI tools, business intelligence tools like Power BI, Tableau, these type of things to a pure location intelligence platform like Carto. Related to this, very recently we announced the launch of Carto Frames and our new data observatory. So Carto Frames is a Python package to use Carto from Jupyter notebooks or from Python notebooks. So this allows you to explore the data that you have on Carto, explore also our data observatories, explore also all the catalog of third-party data that we can offer and request a subscription. Then it allows you to enrich your data with these other datasets that we can offer, run geospatial analysis and then create interactive maps without leaving the Python environment. So using these tools, and I'm going to explain a couple of stories now, one of the projects that we have done has been to try to model and try to look for similarities between different cities. So what we did was to gridify the cities like the grid that you saw before on Madrid and enrich each of the cells with variables, with features coming from different data sources. Then with that, we modeled a similarity score, a similarity index, trying then to understand which areas are more similar between the different territories and also taking into account different cavities that the fact of using data from multiple sources generates. So we have one example here. So we started with a location to a cell in the Goya Street in Madrid, which is an area that is characterized for being quite residential, but especially with a high household income, but then also with very high density of visitors, as we see from both of our analytics, and also a lot of commercial activity, so having a number of transactions quite high. So we started from here, and we wanted to then look for similar areas in Barcelona, adding sequentially more and more data sets into the equation. So when we start on the first one, using the maps, the brighter the color, the higher the similarity score. So we started by only looking into demographics. So when you look only in demographics, you see that the twin areas are more spread on the city, so you have the center, you have the area of diagonal, but also you have neighborhoods on the north, like Sarbia, that I mean, due to the high household income that we see as well in Goya. So then we start adding variables for both of our analytics, road traffic, financial data, points of interest, and you see how the similarity score and the high values tend to cluster around the shambles carras or where like Balmes Street, Montaneribao, that if you know the two cities, it does make sense. Another project that we did using cartoframes and data from Vodafone Analytics was to use this mobile data to compute, to calculate catchment areas, so trade areas. This is very important for retail and real estate, because basically it says how big is the area around that location, around those stores that generates most of the business or that captures most of the visitors. There are many different ways to calculate them. I mean, if you have your business data, that's the idea. And they ask your postal code when you purchase. I mean, they can then compute the trade areas, the catchment areas of their stores. Other more basic methods are just to compute a fixed distance buffer like a circle around the location. Another one is to use walk-in or driving distances, so what we call isochromes. But in this case, we wanted to use Vodafone Analytics data. We created a function, we created an index that tried to reflect the attraction power between an origin to a destination that depends mainly on the amount of visitors that we see from Vodafone and the distance. So then, so we start from there. Also, what's important, there was one caveat, one characteristic of the data from Vodafone that is that the origin of the people aggregates in bigger areas when you go further away from the destination. So if you are in the same municipality, you have cells. If you have municipalities on the same province, you have municipalities and then provinces outside there and then countries. So we had to find a way to cut that data at municipality level and bring it also to a cell that is all explained in one of our blog posts. But then, and then sequentially we started to first create an isochrome to capture the 80% of the visitors and then using that index, we started reducing it until we got the 70% of the visitors to that location using the data from Vodafone. And what is very, very powerful from this is a sense to the human mobility data we can create tailored isochrones because now we can see the difference in visitors between a weekday and a weekend. And then we can try to understand the catchment area of allocation on the weekdays and on the weekend. We can see here in this example that is for allocation in CV how people on the weekends tend to travel longer distances. You see the one in the middle that is only for weekdays. It doesn't capture the Dos Hermanas municipality but when we look into the weekend it does. And we can do this also looking at other characteristics that are provided with Vodafone analytics so we can create catchment areas for a specific demographic profile as well. So very interesting area of research for us right now. So before finalizing I would like to also add some words about how we are evolving the product and the things that we are currently working on. And this is an example of how we by using aggregated data during a whole year we are able to analyze the speeds of the movement of people. Okay. Here we are filtering people moving to the north and the brighter the color means the more speed. Okay. So we in this particular map we are seeing a portion of Barrios Alamanka in which we can differentiate clearly Velazquez going to the north because you can drive to the north by Velazquez and also Castellana. But the other streets are not that bright because the average speed to the north is just people moving by walking not driving. Okay. And we are also working on another dimension that is just to focus on the direction of movement and how it evolves during the week during the day at certain periods because at the end what we want to do is to combine those two dimensions to offer a better product to our out of home clients. Okay. Combining those two we could analyze the average time that for example an app is shown to a particular target profile on a third time street. Okay. So that's all from our side. I think we have some minutes for questions in case you have some. And thank you for listening.