 Hello, my name is Gabi and first of all I want to thank Big Data Spain for having me here. Today I want to talk to you only about three things. I want to tell you about people's data and I want to tell you that I think that we as an industry, the fashion industry, we're going to be able to achieve something that we haven't achieved until now. We are going to be able to capture clothing habits from specific individuals. Then second thing I want to tell you is that with this data that we capture from specific individuals, we are going to be able to give it a structure. We're going to be able to understand it and build taste profiles in the same way that today Spotify perfectly understands what music I listen to. We call Closet Graphs to those taste profiles. Then third point I want to tell you about is Personal Assistance. I want to tell you that in my opinion Personal Assistance for your clothes, for your closet, for your outfits are going to come as a result of owning and understanding people's data. No one is going to create, no one is going to invent a fancy technology and then on top of that Personal Assistance are going to pop up. We're going to be using existing technologies. What is going to change is the availability of specific taste data. Before I tell you about these three things, I want to tell you a little bit about myself. One minute. I am the co-founder and CEO of Chiquísimo. We are super small. Three things I'm going to talk about. Before I do, a minute about myself, Chiquísimo. I'm the co-founder and CEO of Chiquísimo, which is a small technology startup. Some of you might know Chiquísimo because of our fashion app. We have a fashion app that helps women decide what to wear. But actually 100% of what we do is focused on building the infrastructure and the operations to create Personal Assistance for your closet. We want to know what you have in your closet, what specific clothes you have in your closet, what you take out of your closet and wear, and all the things that might be relevant to you, such as what occasions you attend to and how that affects your clothing habits and also how do you describe to yourself anything that has to do with your clothing needs and habits. We do build the infrastructure and the operations to make this happen, but we are company oriented towards the consumer. Everything we do, we ship it to the consumer as soon as possible. People's data. If you think about it, I'm going to use the example of music to describe what I think is going to happen or it's happening actually in the fashion space. If you think about it 25 years ago, when we listened to music, we were just listening to music. Today when we listen to music, we are sending our behavior, our specific behavior to some companies. And those companies, as a result of that, know what we listen to and have been able to build our technology. We have been able to build our taste profiles. Like 15 years ago, some people, me and some of the people, one of them is right here, we started building plugins and mechanisms to capture the behavior of people in the music space. So what we wanted to do is to understand what songs you listen to. So if you are listening to streets of Philadelphia, we would get that information through an ITUs plugin or other plugins. But then we soon learned that that information is very relevant, but there was another type of information that was also extremely relevant. So not only the item, but also the sequence in which you listen to music, the sequence in which you consume content. So when you scale these to many people, you start understanding the strength of the correlations that people establish among items when they consume those items together. Okay, this was 2003, 2004, so it was very early in the days of recommender systems. This was new. But there was a super interesting learning for us. Not only what you do, but also the sequence in which you consume content. And with scale, what we ended up building was what we call the matrix of association, like a huge brain that would understand what and how people or different subsets of people would consume this type of content. What the company was called at the time, music strands, we didn't make any money out of it, so we sold the technology to Apple, and now part of that technology is powering Apple music. We applied this concept of capturing data to the finance space, and we built FinTech in 2007. But then, when we got to the end of the world, we started another group of people. We started looking at the fashion space, and we thought, so there was an industry that had no idea about the specific behavior and taste of people, and a few years later, everything has changed, and the industry understands our behavior. Not the industry, but some specific companies within the industry. So that same shift happened in a number of segments. So the question we had was, will it happen in the fashion space? Will someone be able to create a capturing mechanism? So that's why we built Chiquisimo. We built Chiquisimo to create those capturing mechanisms. So what we wanted to do is giving an outfit, which is how you express your taste and your behavior. So given an outfit, how can we capture what specific clothes you have? So we built a website and a few mechanisms, and people share their content. What we learned when we got this information is the same as we learned in the music space. An outfit is a playlist. So when you get dressed every morning, you are establishing correlations among different items. You are saying that it makes sense to work two items together. So we not only started capturing specific items, but also the correlations that people would establish among those items. At scale, you start to understand how people combine clothes, different clothes, and brands, and anything. The result of all this is, again, a brain that understands how people combine their clothes, what people were, and how that evolves. Not only people, different groups of people, total and subsets. So we built what we call the social fashion graph, and this is a summary of the process that we had a couple of years ago, because now things have changed quite a bit. So given an outfit, we capture the correlations in that outfit, and on top of that, we built this graph of correlated items. And this system, this data platform, allows for a search and discovery experience. So given any input expressed by someone, I need ideas to go to school. I have a wedding in winter and I'm going to wear a red dress. With what should I combine it? So given an input, the graph does the job and then delivers an output. So this allows us to create an experience where people continue giving us further data about their needs and habits. Through the search engine and through other mechanisms. This is the consumer app. In case you don't know it, it has something that we call MyCloset, where you can store your digital clothes. You can tap on one of your items and it will give you ideas on how to work that item. It will tell you how other people are wearing that item. And it also tells you how you can combine that item with the rest of the items in your closet. So the consumer app is obviously a delivery mechanism. And it is also a capturing mechanism. But one of its main important roles is to serve as a tool to understand what interface is the most useful to capture whatever someone has in their head. So I have a need, I need to express it, what's the best interface to capture that need. Just a little bit of background to give you an idea of where the number one outfit is up in the world. It will be featured by Apple in 60 countries. And this is the distribution across countries. So it's basically Europe, the US, and Latin. And this is important because when we capture data, we capture data in different forms. And culture and language affect the type of data that we need to process. So first of all, we think that data from people is going to be the most heavy-capture, okay? But the fashion space is actually super complicated because it doesn't have a structure, okay? When you express your taste in music, and if a lot of people express the taste in music, they're going to use words and descriptors that the industry also uses. The type of music that you like, pop, rock, whatever. So in general, the taxonomies or the ontology that people use is the same as the taxonomy that the industry uses. So whenever somebody expresses an input, it is very easy to capture it and to understand it. In the fashion space, that doesn't happen. So when we looked at the data, we thought it was, we thought we were super cool because we had millions of unique, different descriptors, you know, describing people's taste, habits. But we understood that those, that chaotic and unstructured data was useless because our algorithms could not act upon it, okay? It was too complex, too unstructured. And what we did next is something that's been done in many other sectors, so it's nothing new, but it's actually quite new applied to the clothing segment. So we understood that the 200,000 descriptors, normally, sorry, the 200,000 descriptors cover 90% of people's input. So it's a huge long tail, okay? So we focused on those 200,000 descriptors. And looking at those descriptors, we understood that most of them refer to the same concepts. What is this? Someone could say, I need ideas to go to class. I need to decide what to wear to school tomorrow. Someone in Spain would say, I need ideas to go to El Colegio or the Instituto. So there are different ways of expressing the same concept. So we ended up creating a list of 2,500 concepts, and these concepts are a summary of 99% of the needs women have when they express their clothing needs or clothing habits, okay? So it grows little by little, but it's going to grow no much more than 2,500. So this is what we call our ontology, okay? The ontology is the backbone of our data platform, of our social fashion graph, and it is in charge of helping us convert and structure data, which is the problem in the fashion space, to data that we can understand and that our algorithms can work with, okay? So we have a lot of data, then we have a lot of clean data, so we have billions of structured data points, okay? And then one of the roles of the social fashion graph is to automatically attach the scriptors of the ontology to the outfits, to outfits and to people, okay? So as individuals communicate input to the device, okay? The ontology is transforming that input into understandable input and attaching it to the outfits and to people, okay? With this we can encapsulate what you've expressed in taste profiles, that's the way, like a spot if I would call it, or in closet graphs, which is how we call it. So a closet graph is a list of the clothes you have in your closet with very rich metadata, and that metadata defines your clothing trades, your clothing occasions, and whatever the scriptors are relevant to you when you get dressed. So, well, this is chiquisible. A data platform that receives data and forms a clean data set of correlated descriptors, outfits, and people, or bags of descriptors, that communicates with delivery mechanisms, with the consumer devices, okay? The consumer devices have an interface and content, and then they have incentives, you know? They have a number of habit forming technologies, and both the data platform and the consumer app talk to each other through the algorithms. So basically the algorithms, in our case, we have nine relevant algorithms. In our case, the job of the algorithms is to understand what descriptors are needed to deliver the right content for specific people input, okay? So, given a question by a user, the algorithm goes to content, goes to the social fashion graph, gets a number of looks, and chooses what descriptors of an outfit does it need to collect to return the right content for you. So basically I need ideas to go to a winter wedding with a red dress. So the algorithms bring you that content automatically. So the social fashion graph, consumer app, in the middle of the algorithms, and then a number of tools that we use and give us basically control, okay? All of this is patented in the U.S., the data capturing mechanism, the interfaces, and a system to tag fashion items with shopable products as well. So the way we see it, again, we're going to be able to capture data from people. By creating specific ontologies, we are going to be able to provide a structure to that data that we've captured. And as a result, we're going to see soon through personal assistance. And I'm going to describe now how I think this is going to happen. So if you think about digital clothes, this is an asset that is extremely recent, okay? Almost yesterday, brands didn't have their digital equivalent. And now, well, these garments, they do have their digital equivalent. And in most occasions, these clothes are listed as catalogs, are lists of clothes in the e-commerce website. And the role of the team there is to make sure that when somebody wants to find something, they can find it. So how can I add the correct metadata to my products? How do I build the browsing experience? What do I do with the search box? Okay? The problem is that there is a number of companies that have thousands or tens of thousands of different clothes, garments, in their website in any single day. Think of, I don't know, any large retailer, H&M, for example. You go to the website, you might see 200 garments, but actually behind all that noise, there are 20,000 garments. So what we're going to do, what I think we are seeing is some attempts from a few e-commerce companies or technology companies and brands to shift, to move from list of clothes to graphs of clothes. So what they are trying to do is to create, to offer the user mechanisms to go from one item to another. And these mechanisms are very different to the traditional browsing methods. So what we've seen, the technology that we've seen attempts these are recommended systems. If you're looking at an item, it shows you similar items. Okay? And a few companies employ these systems a couple of years ago. Okay? It didn't work for them. Now what we are seeing is the same concept but with different technology, basically MS recognition. So given an item, give me similar items. So what they are trying to do is they are trying to help you dig deeper into the catalog. So you not only see 500 items at a time, but you see, I don't know, 1,000 but are more relevant to you. What I think some companies are trying to achieve and some of them are closer than others is when someone is looking at a piece of garment and looking at a red sweater. The way I think about that red sweater is that it's comfy. It's that it is warm. It's ideal for the weekend. And it's, I don't know, tumbler style. So the way the company has classified that is as a red sweater. The way I classify that is as tumbler, warm, whatever. Okay? So somehow a few companies are enabling people to express the input, those descriptors, and that allows people to dig deeper into similar items according to the way I describe this item. So how can you, how can I find more clothes that are warm, that are tumbler style, and that are casual or for the weekend? Okay? So in order to do that, again, I think you need a couple of things. You need to own and to understand an ontology, so a list of descriptors that people use and that can be attached to your content. And you need a system for people to express themselves. Okay? So digital clothes are a new asset. Most of them are listed as catalogs, but some companies are trying to establish correlations among those items, especially if they have large catalogs, and especially if they have a discovery problem or a personalization problem. What I think is coming very soon is the combination of the graph of a shopper, so my closet graph, so my list of catalogs, my list of clothes, the list of clothes that I have in my closet, with the catalog of the e-commerce. Okay? So if both items, if I understand the ontology of both graphs, or if I've created the ontology of those graphs, I can easily match those graphs, so when someone comes to my e-commerce, I can tell them what from my e-commerce, from my catalog, they're going to like. But also as a user, when I go to a website, I'm going to be able to see what from your catalog fits my existing closet. So this is an example that we are not going to see in the short term, because it depends on, you know, a number of companies shipping it, and it's not going to be easy for them. But if you are looking at a piece of an item, what you want to see is how an item combines with what you have in your closet. Okay? Or you're going to be told how an item, how you can build outfits with what you are seeing and what you already own in your closet. Okay? So the two former use cases are going to take, I would say, some years. But there's another use case that could take a few months or maybe some weeks, which is the use case in which you go to a store, you scan a product, and this is not going to happen in all stores, you scan a product, and in real time, you are told your device tells you how that item combines with your clothes. So what specific outfits you can build with this item that you are looking at. And this is the beginning for me, for us, of personal assistance. So I think everything is going to change. I think we are in the fashion space where in the same moment as music was in 2006, there's lots of things happening, a lot of noise, lots of technologies, but there is, you know, we might see in the future a few unicorns with technologies being adopted by a lot of people. Technologies being adopted by a lot of people because they power very useful use cases. And what we've learned from feedback from millions of users is that the single biggest problem that people have is how to combine their clothes. There are different ways to express this. How to combine my clothes, how to combine colors, how to decide my looks more easily. I want to feel more confident when I go out to the office or the school or, you know, I know how to combine my clothes, but I want to innovate. I want to change, I want to risk a little bit. But combinations of the same problem, how to combine items. I mean, machine learning has solved this problem decades ago. The problem in fashion is that we do not have the data. We do not have the item, the described item. So that's the problem that people have the way we see it. But then the biggest opportunity for the fashion industry is to be able to build a mechanism to capture and to understand people's clothes and people's habits. And we think this is going to happen in five years. So close to data is going to be captured. Personal assistants are going to help us make decisions. And this is not something that is going to be owned by the industry. It's going to be owned, we think, by a few tech companies in the same way that the music consumers own also by a few companies. So it's a very large market. It's a very large market market. It's now owned by a number of companies, but maybe five years from now the ownership of this segment is changed. We think that. So a number of players that are looking closely into this space, what are the technology companies that want your attention because they want to sell you data or they want to sell you ads or they want to sell you clothes. So these tech companies, they're fighting, they're building the foundation to own the attention of people. Then there is obviously a number of companies that are very interested in understanding trends, but not trends as we define them now. So there's a number of companies who want to know if there is a brand in South Korea or in South Korea whose sales have multiplied by three in the last week, they need to find out as soon as possible. So that's what we refer to when we say trends. And then another type of companies are hardware manufacturers, like Samsung and others. A lot of people have talked about smart mirrors and personal assistants, but what these companies are doing is giving an input, giving a garment that you might buy, I'm going to tell you what other garments which are similar to these you might also like. So the software that is powering those smart mirrors, it needs to evolve. So a clear problem is that there are a lot of companies that we think is going to be solved by technology because there is a huge opportunity at stake. So this is what I wanted to tell you today. People's data encapsulating taste and then serving people automatically. If there is anyone kind of building similar things, I would love to talk later. So if you have any questions, please. So this is a question there. So the question is we are going to build this for men. I mean, eventually the industry will offer this to men, but today the sources of data, the interest doesn't lie on men, it's what women is. And then for men it's like a copy paste. This is another question. Which was the first value proposal of the app before gathering all this data? How did you obtain the data in the first stage? Yeah, so one of the things that I said before is that we've always thought, so in the past we've done lots of things focusing on the algorithm. In this project we are focused on shipping because we think it helps you learn quite a bit. So the first thing we did was we built a WordPress and we emailed people asking for their outfits and they would send us their outfits with a description of what they were wearing and what it took off. And the first technology that we built was we hire a developer from China and we paid $80 for a voting mechanism. So the voting mechanism is also a way to express your taste. So it was super basic. So as soon as we had the idea, we shipped it. Thanks. There's another question here. Hi. Yes. So I see two problems or potential problems here which is one that people don't have a lot of clothes. I mean we joke about it but we don't really have that many clothes to really know that much. Now on the other hand, taste and clothes changes over time probably very quickly. How do you deal with this? How do you want to deal with this? Yeah. So we ask our users lots of questions and one of the questions that we ask and we love the data we get is what do you think about your closet, about your physical closet, about your clothes? And I would tell you that maybe 5% think they have a very little amount of clothes. So most of them think it's boring. A lot of people think they have duplicated clothes. Clothes they don't use. Very few people love their closet. Okay. So what we found out actually a couple of years ago is that clothes and brands are very important that describe in your taste. But there are many other elements that are critical to describe your taste. As I said before, normally related to occasions, like special everyday occasions. Okay. And different descriptors how do you describe yourself? Okay. So for many people going to the office five days a week having to choose five different outfits, you know, and those outfits kind of send something to their peers. It's a problem. But it's a problem full of descriptors. And then about taste, about how taste evolves. Also we've seen it before in other sectors. I mean, in the music space what we saw early early on. You might like ACTC, but you might also like Beethoven. Okay. So I mean, there is two different items that you are kind of interacting with. But the context in which you interact with them is different. So evolution is also tracked. Because one of the pieces of metadata is time as well. I mean, time meaning you like this today. You express an interest for this today. You buy this item today. And we see also how, I mean, I was, there has been a lot of, the brand Nike has been in news lately quite a bit. And we looked at how Nike and Adidas had evolved over time. And obviously they've evolved a lot. One of them has gone down. The other one has gone up. So yeah, things evolve. But you track that. So any other question? Well, thank you very much.