 Thank you. Good afternoon. First of all, we would like to really appreciate to each of you to stay here today, to the very last session, especially considering the alternative with these drinking beers outside. So thank you very much. We will try to do it short. We will try to do it interesting for you. My name is Santiago Aguero and I'm the Data and Analytics Director at the Cotel here in Advanced Analytics. The guy who actually makes things happen in Advanced Analytics. First of all, I mean, today we are going to talk about artificial intelligence and travel, but we would like to start with a very short briefing about first travel, then advanced analytics, and then we'll go in both of them. When we think about traveling, when we think about tourism, probably you are not thinking on that. But it was traveling. There was a tourist last century at the beginning of the 20th century. This one was tourism. This one was tourism. This one was tourism. If I ask you to frame a picture of what traveling means for you today, most probably you are thinking of something like that. A very different image. Here we are in Spain. I mean, tourism in Spain. What do you think about tourism in Spain? Probably something like that, something like Benidor or La Manga or whatever. Or even if you are a more sophisticated traveler, you are thinking on South Asia, you are thinking on Latin America or whatever. Any of these pictures, any of these last three pictures that I just saw has something in common. It's tourism and traveling has become overcrowded. This is something that we are experiencing not in the last years, but in the last decades. Some figures just to frame that. Number of international departures over the last two years are multiplied for more than two times. Of course it has some implications. Traveling sector and touring sector are experiencing extreme things over the last years. One of the things that are not actually increasing in the tourism and traveling sector is number of workers by customers. Understanding customers. We have a lot of new travelers today, new countries, emerging countries, especially emerging markets as China and Asia and Latin America. But also we consider a new customer the increase of frequency of your trips. 20 years ago, probably you go on vacations once per year. Today you are traveling quite a lot more. Number of workers, number of capability to serve this new demand are remaining almost constant. Almost constant. So it means traveling sectors and touring sector are facing some challenges. First of all it's this new type of customers. As I mentioned, new customers from emerging markets like China and Africa even with new needs, new requirements. But also we each of us are a different customer in its situation. It's not the same when you go on a trip with a bunch of guys, a bunch of friends than when you are planning a romantic weekend with your significant other. You are a very different type of customer. And today not only new customers but also new players are addressing these new needs. Airbnb. We have booked it 20 years ago. We have nothing compared to booking. And these new players are taking part of new slides of the value chain proposition. Most of them or some of them at least are giving transparency and cost. Now we are very, very aware of how much it costs, how compared a ticket plane, hotel room, etc., in different channels. So there is a pressure on margins. And this is the real business challenge that the industry is facing today. But of course not everything is bad. What we have today that we didn't have, what we had 20 years ago, we have this. Trillions, bookings, millions of search, a lot of data that both in direct channels on the industry channels but also outside then are bringing more and more opportunities to gather, store and processing information. And this is what we try to lever today. And this is what we are trying to explain how we do it. Of course 2018, the name of the game regarding big data is artificial intelligence. And this is not something I'm trying to tell you. Today this is a Joe and BCG and MIT survey and you will see travel and transportation is one of the sectors that will be most impacted by artificial intelligence. In two ways, I mean they put two axes that I agree with them. It's offering, I mean, we are offering to our customers new type of things but also processes, the way that we approach our customers. And for us this new offering and new approach means three things. Customers experience, I mean it doesn't matter that we all of us go to the same destination in the same date every time of the year. We want a customized experience, we want to feel special when we travel. With our friends, on our own, with our capital, et cetera. Also personalized deals, I mean this is about overcrowding. I mean new and new and new customers are going into traveling and dreams. And you know what? I mean they are willing to pay exactly what are they getting for, right? If you get a bit more, I mean you are expecting a bit more. But if you are giving up some value, you expect actually the company, the hotel, the airline, et cetera, to offer you a better price, right? And this is something that we get used. I mean this didn't happen ten years ago, right? No one questioned that if I had to wait in the airport, lack. No, now we are expecting a return. We are expecting something, right? And customer experience, personalized deals, even if this is an overcome market, we cannot expect to use mass channels in order to communicate with our customers, right? We need tailoring communication. I have to communicate this special deal for you with this special experience for you and then I have to use direct channels. I have to use personalized channels. What we call here, I mean tailoring communication. It's not about TV and radio, right? It's about receiving the exact message and the exact moment. Okay, I tell you, name of the game in 2008, sorry. What is preventing us to use advanced analytics or use artificial intelligence? It's not about the lack of knowledge. I mean you are hearing in an event, I mean hundreds of people talking about that, talking about artificial intelligence. You open up newspaper, everybody is talking about artificial intelligence today, right? So why are not the companies jumping into that? And it's not because of the technicians. This is because of the top executives, right? What we have found over the last years when we go to a company and ask the CMO or as the CO, what do you think about artificial intelligence? Let me show you a picture and summarize that. This is what they think about artificial intelligence, right? Something like a black box, right? Okay, you're not really sure where the value is. And this is not something that I just came here and so this is another survey from McKinsey, these guys do a lot of surveys, to see level executives. What is preventing your company in order to apply artificial intelligence? And let me highlight the top three reasons, right? Talent and knowledge, I don't have the people. I don't know how to do it. Artificial intelligence, technology, maturity, remember the crystal ball, right? And top management and clear about a value. Basically, not sure if it applies to my company. Yeah, I know my competitors is using, but we are different, we are special, right? So, and I will handle the ball to my colleagues here. We are about to bust these remits. We are about to talk about three use cases that we have actually implemented in three different clients in order to overcome this kind of barrier. So, Fran? Thank you. Yeah, so as Santiago said, we are going to try to make things happen. And we are going to try it through these three cases. First, how we can, in fact, I will present them as three disruptions, because I really think that AI changes the way we are doing these three things currently in the sector. So, first is how we can answer to Santiago's challenge about how to tailor our communications. Then we'll see how we can personalize our deals, redefining our pricing strategy. And finally, how we can customize our product, but also our service to respond better to our customer needs. So, let's start for the user-based marketing disruption. And here, I would like to start with this number, which I found, I really think it's quite describing the potential of this use case and why I think it's a disruption. It's a figure from Google Insights saying that 60% of the passengers, I think it's from serving the US, but I think it applies also in Europe. 60% of the passengers would be willing to go on an input strip if you impact them correctly. If you choose, if you find the right offer at the right moment and using the right channel. So, let's go on how we should build this to make it happen. So, here, which would should be the vision? The vision is very simple, I think you can imagine. It's using all the information, all the data. We have right now, as Santiago said, it's enormous. It's a lot that we have about our customers. So, knowing better our customers and using this data through AI to first know which of those customers, in fact, are part of this 60% are really willing to travel with you at that point, because this 60%, in fact, it's something that it's moving. One day, you are willing to go on a travel, but maybe next day, you are not. So, it's something that has to be changing. So, first, you have to identify those who are willing to travel with you, that's important to. Also, to then identify which is the offer that they are willing to accept, which is the special deal that you should push to them. And finally, how you should push the channel, but also the message and even the picture or even the complete communication. So, here, we'll see how we can build it technically. But here, it's important to have in mind three principles. First, as I said, it has to be user-based. What happens, what works for one person, it doesn't work for another. And here, I'm really talking about a user, and I think it's a principle. Because in the last trends on marketing online, we can even personalize at segment level. It's something really common. But with AI, we can go to user level and it really changes how you do things. Then, secondly, you have to be very smart to impact those who are willing to go with you and save the money and don't impact those who are really not going to go with you. And here, when I'm saying going to go with you, it's on the short term, but don't forget about the long term. Because here, they are willing to go with you today, but maybe they have a high potential in the future. And you should also impact them. And finally, it's what I was saying about you can use a segment-based technologies. But as you learn, and it's important to learn from yourself and your errors and your impacts in the past, you can switch. Because the first time a customer is going to go to your site, you're not going to know this customer. And you'll be forced to use lookalike strategies, something to say, OK, I don't know this customer, similar to other customers. But you have to be very fast on this and learn to switch and to move from something like segment-based to really this user-specific user. So how we could build this technically, I'm going to show you this through a remarketing example using only one data source to make it simple. But of course, you can do it more and more complex. Here, for example, the data source is your own navigation data. That, for example, there are a lot of technologies, but you could track it with Google Analytics. Then from your own navigation data, what you know is what they are doing right now on your site, which products they are looking at, how much time they are spending on your site, these types of things. Those are your variables. But you can also go on your historic data and look and say, OK, which were on the past the good customers and the bad customers, or maybe better, the good users, the bad users. And the good users, for this example, it's very simple. It's the one which booked, for example, if it's a remarketing model. And the bad ones is the ones that didn't book. So you give all this to the machines to learn. They will learn which are the patterns, and then you just have to give him the current users that are navigating on your platform. And they will say which ones are going to book and which are not going to book. But the way of talking, of responding to the machine is by using probabilities in this case. But you can use these probabilities to segment your audiences and impact more the ones that have the highest probability, or even don't impact all the ones that have the highest, highest probability because they are going to book with you anyway. Of course, don't impact all those who are very unlikely to book with you. For example, for instance, and this is a real number, by applying such techniques, you can have figures like this. You can reduce your remarketing investments by 50%. That's why I was talking about a disruption. So next case, I would like to go through a pricing revolution or a disruption. And here, I'm just quoting the last research saying that you can increase your revenues by 11% and later research, I'm talking about, I don't know if you know these guys, Santiago Gallino, Junli, they are very well-known people in the revenue management world. So they are serious guys. And here, why I'm talking about a disruption if it comes from the revenue management world? But let's see why. So here, the vision, the start is the same. You know a lot better your customer than you used to know him. And you can use it to really know how much every customer is willing to pay. How does it change the game? It changes the game because on the past revenue management, how it works in a very global way, it's comparing total capacity of your products. For example, your airplane or your hotel. It's comparing your capacity to your demand. Maybe you are not using total demand. Maybe you are segmenting it in business travelers and tourist travelers, maybe even something more sophisticated. But at the end, you are comparing demand in segments to capacity. With this, you have still to do this because it's important to compare total demand with total capacity, this is important. But you can, on top of that, modify personally, user per user, a little bit, the each price, to go a little bit higher or a little bit lower, adapting the price to the willingness to pay of each of your clients. And here, there's a second component which is really important. And it's that understanding that you have to use AI just to estimate the willingness to pay of the customer, it's not the full picture. In fact, there's also a strategic component because this is going to interact with humans, not with machines, and when we are booking a trip or even a product, it's very general. We have a decision process, and it's not the same that I offer you 100 euros, for example, at the beginning, and then I go down to 80%, and I offer you this 80% with a specialty for you. Maybe you are going to accept this 80%, but if I do it with another strategy, you come to my site, I put 80% at the very beginning, maybe you are not going to accept this. So you have to really develop our strategy, understanding how the humans behave and even a little bit more complicated, how each human behaves to maximize your total profit. So the first one, it's really the same as before, it's user-based. The second one, the second principle, it's this strategy that you really need to develop our strategy, which is the best strategy for each user. And finally, and here, it's even more important, it's as before, that you have to self-learn, but here it's very important, because in fact, when we are learning on strategy, and here I'll talk later on, you can use, for example, reinforcement learning methodologies at the beginning, if you are applying it to topics like pricing, that it's really impacting your P&L, so you have to be really, really careful. You want to learn a careful strategy, something that you're sure it's going to work, and it's not going to decrease your revenues, so you're going to learn from your analyst on the past the strategies they developed, but what deep learning and reinforcement learning in general has proved in other sectors, but also in this one, is that when we learn from humans, when we learn strategies, even more important strategies from humans, we are not maximizing the total earnings it's even better to learn from the errors of the machine and let the machine even improve those strategies, so here it's really important to let the machine learn from its own errors. How can we build this technically? Here the very and most important thing is to gather a lot of data because to really estimate the willingness to pay at each time and also know at which point of the decision process each of your customer is really complicated, so you really need to know your customer, integrate, I put here five data sources, but really if you can't imagine of another one, put it, use it, then you have to integrate this, you can use leverage AI techniques to estimate the willingness to pay, and here I just put the example of three customers, but it's not just how the willingness to pay is changing depending on the price, it's also changing depending on the time, taking into account the time component. Finally, you still have to do your total demand against capacity comparison, because if you have some users, if you have less capacity than demand, you are going to offer a very high price to some users because maybe you don't want to let them access to your product because you have other clients that you can accept, and finally you offer a price. Here, which is really important, I'm going on the down part of the slide, it's to have a jointly strategy on your prices, because here what's going to happen is that first, you need to do it in real time, for instance, that's why I put a technology like Data Prep or Data Flow, you can use another one, it's just an example, but what it's really important is that you have to be consistent. Maybe you strategically want to define that you don't want to offer the same price to the same customer on different channels, why not? I could understand that, but you have to be consistent, you have to have a strategy, that's why, for example, you have to use a CRS, here is how we build it, for instance, ourselves, but this is really important. Okay, so I'm going to go through the last one, how you can personalize your customer experience based on AI, and here the figure, the number, it's really impacting also, it's that 36% of the travelers would be willing to pay more if you iterate your services and even your information, taking into account their past feedback, and you have your data, you have the data to do this, they are, even more in the travel sector, your customers are giving you a lot of feedback through service, through claims, through social media, but also in other sectors, data is there, so it's not only about increasing customer satisfaction, this is the focus, of course, but you can also increase even your revenues because they will be perceiving more value, and Santiago said, if they perceive more value, they are willing also to pay more. So here the vision, it's more or less what I already went through, it's about taking into account all the feedback that you have online or offline, all the opinions of your customers in the past, to put your customer in the center and to use this data to improve your product, to first respond better to their needs globally, maybe you can learn from what one customer is saying to improve your global product for everybody, but also to offer special offers or special deals, or maybe not special deals, special services, I would say special services for special customers, for example, if you identify that you have a special customer, maybe a business customer sitting on the front seat of your airplane, that always is asking for a glass of wine before the departure of the flight, you can proactively offer the glass of wine to him, so you increase a lot of his satisfaction. So the principles I think are really simple, you have to be client oriented, don't think, of course, as I said, you can increase your revenues also by using this, but it will be a consequence, you have to put here the AI working for the client, this is really important. Second, and it's a little bit more technical, but it's really important in here, you have to take into account the context where you're client, it's giving you this opinion, it's not the same how I'm writing or speaking in Twitter compared to how I would talk to the hotel director, it's really different, so taking into account things like the tone, the words we are using is really important to do an overwrite, and finally, all this will be used to offer recommendations to your hotels, to your service, to your agents, and each recommendation you do, it has to respond to an NPS, the satisfaction, the maximization of the satisfaction of your customers. It's like putting, when you take each decision, you have to take into account that it has to maximize the satisfaction of your customers. So how you can build that? Well, as I said, you have to join every single piece of data you have gathering the opinions of your customers and here what's complicated, of course, it's non-structured data, it would be text data, maybe even voice data, so you have to try to join your silos to have a unified vision of each of your products, services, but even also users. Finally, you have to do things like translate your commands if you have minor languages that you want to take into account, and you have to estimate the sentiment of each of their opinions. Maybe you want to do things like detect the global topic and maybe estimating that your clients are talking about the reception as a global topic because you want to know if, under reception, if there are a lot of problems, for example, in the reception, but you also want to detect the special insights, so something much more specific than reception, maybe you want to detect that you have a problem in the time, maybe the check-in, it's too slow, for example. And finally, as I said, you have to recommend some actions to the agents for improving the product but also to give special services to your clients, for instance, the second example, the one I went through before. First one, you are detecting that you have a problem in the reception. So at this point, I hope that I did it quite interesting for you and that you are very motivated with these free cases or even with more. So I hope you are starting to wonder this question, to wonder what you should do to start with this tomorrow. I have seen that you have taken some photographs to your cases, so you know how to do it. Okay, but what did you should do more than implement what is here? I'll try to do it in four tips but I believe that are very useful. So first one, I think that it will be shared with most of you but it's something that we must not forget is that AI is another tool. It's a very powerful tool but it has to be aligned with your business strategy. It's not something that it's going to solve your strategy. It's not going, AI cannot be your strategy. AI can respond to your business strategy. This is very important because then it's when you are going to see the power and the return of the inversion in such technologies. Second, I think that all the data scientists in the room will agree with me is that you should invest in people rather than in technology. Why I say this is because there are a lot of cloud technologies. I used Google for this presentation but there are others that you can use and that will permit you to scale fast and cheap but what's really important and you really need to do is to invest in people it's really complicated to develop all these cases if you don't have a good data scientist which technically understands what the algorithms are doing not just that he can use a library but also understand what's happening beyond that but also someone that understands the impact of the algorithms in your business. These two are really important because it's not only how the algorithms are working it's also how you combine all the algorithms to respond to the business question. Then my third point was that you should move really fast it's something that we are seeing the ones that are winning the AI battle are the ones that move faster also that move smarter but it's something where you need to really to be agile and to move fast so don't embark yourself on a two-year project without seeing results very quickly try to do a one-two-month project see the value, see if it's worth or not and then maybe it's when you will be able to convince all your company that AI can really bring value to them and they will invest more and more on such initiative and last one it's to close this I think it's really important we have talked about the business impact we have talked about the talent of people developing these technologies and even the technologies, how to build it but we have seen it a lot you can build a very powerful algorithm responding to a precise business question you can even prove that it's working and that it's bringing a lot of... and the return of inversion it's more than proof that you are multiplying yourself per free for example you can really prove that it will work but then when you deploy your solution in a lot of times it won't work and it won't work because at the end in most of the cases in some you can automate everything but in most of the cases there will be a person or even a team working with such technologies so it's really important to not only transform technically your company but also transform the culture of your company and really share with your peers and also with the teams that are going to use such tools how they work, that it's a learning process that it can be wrong but if they teach it it can learn from them, it's really important such things are really important so then deploying such use cases will be a success in your companies so thank you very much and if you have questions we are here to answer Hi, thank you for the talk and interesting insights my question is actually around the price strategies weren't showing different prices to different customers or maybe different prices to the same customer at different point of time affect the trust the user has in your product? Totally, that's why I said you have to be really strategic and you have to take it into account it's the example I tried to explain about it's not the same showing to a customer 100 euros the first time he arrived and then decrease to 80 then do it the other way around and especially in this sector for example if you start to decrease the price at the end of your at the close of a flight because you have some capacity your customers are going to behave not the best way for you they are going to arrive all at the end so that's why I was talking about defining the optimal strategy rather than just estimating the willingness to pay and maximize the short term willingness to pay it's defining a strategy that maximizes your profits but not at the very short term you have to, totally, you are totally right and I mean building on that I mean as Ferran said artificial intelligence is just a tool I mean it will tell you what is the willingness to pay in a given moment of time for a customer but you have to define what is your strategy you were talking about different prices for the same customer at different moments of time that's one strategy also could be different prices in different channels and today customers are aware of that I mean they don't expect the same price on a physical store that on an online channel we finally get used to that and of course time to the service channel very different factors may affect the price and it really depends on how you want to play with that this will be a tool that gives you insight about willingness to pay how do you use this as part of your business strategy maybe it's very... maybe you can review it talking about the offering price different prices on different social media for the same customer it might be much more acceptable for the end customer not customers, customer why you are offering the different price on this social media comparing to the other one like okay, you didn't accept our previous offer we are now offering the different price because something I think it's on strategy I think that for me the answer it's more or less about what Santiago said it's up to you to define your strategy I don't know if it's I think that you have to have an answer as a company to that question that it's really fair if you want to be that transparent I think that for some customers and from some companies it will work but I think it's up to the company to decide if they want to do or not what you are suggesting that I think it's a good opinion it's a good suggestion but I think it's up to the company to decide if you want or you don't want to do this which I think it's fair enough the question is why why a customer will accept think about this is traveling this is... maybe airlines hotel rooms or whatever but in a couple of weeks we will have the Black Friday you will receive an email saying apply this code to our website and you will get 20% discount with this actual it's actually dynamic pricing and it will be normal for you why? because you accept that this is a limited offer and this is just in this channel what is the web page I won't get it in another channel there is a cost saving for the company if you go through their website or their app that you cannot get you go through booking for example and customers get used to that how you train how you train your customers how they can expect what they can expect from you loyal customer, recurring customers will get used to your strategy that's true so you have to be careful with that and sometimes that's true but seeing for example the give best in the last 5 days they are referring to 11th of 11th the big price down for this event but the prices were going down all the time like if you are a customer and waiting for 11th of 11th and you want to buy something you will get some kind of lower price but then after 2 hours the price was even lower so this strategy like waiting for only 24 hours of flesh sell it's like marketing you are describing a bad strategy maybe they are not using artificial intelligence maybe that's why they have to start using that rather than a general strategy it's a strategy and it could be a good strategy it could be a bad strategy this is just a tool that allows you to build your own strategy you don't have to replicate whatever is doing other island or hotel channel you may want to build your own pricing strategy and what we are bringing here is a tool that allows you to do it in a smarter way with higher knowledge of course there are bad strategies if it has the point yes I have a question on the same topic and what about including fairness metrics for the reinforcement learning I mean to diminish the disparity on prices within the same customer through time and within groups of customers have you tried that not specifically but I think it's a really good idea depends for me in that previous case I said include all the information you want to in that case I think it could change your strategy so you have to have this reflection before you want to do it or not but it could be totally possible and I think it could be more fair then it's up to the company also if they want to be fair or just maximize the revenues of course but it's something you can do you can do a balance of both between customers you could also maximize the weight on both objectives of course so in the travel industry there are a lot of different players and they have a lot of sorry a lot of different constraints let's say you have airlines have to sell a number of seats airline also OTAs don't don't care about that much so how do you deal with these different constraints in this case for example OTAs and airlines dynamic pricing strategy OTAs actually are not much playing with pricing the range of money is limited most of the time that we have done these for airlines hotel channels we work with them through the strategy the pricing strategy across all channels and OTAs is a channel it's not a player itself as you say they don't have much to do with pricing the pricing is received they can play with their margin most of the time OTAs have agreement that they cannot go below a certain price so they cannot play so much anyway in terms of aggregators or in terms of other players that may offer different product or similar product for different airlines or different rooms for different hotel channels you can apply exactly the same it's pretty much the same the only difference is you don't have one product you have a set of products and you can play with them but most of the time we work directly for the final company airline or hotel channels not aggregators in fact what you can do if you want to work with an OTA they work with percentages of the total amount of booking so they really care also on the price not metas, it's true but if you want to work with those one thing it's really interesting it's to tell to the airlines for instance how they should change the price or the deal with the OTA to change how they are positioned on the searches one of the parameters so how they should optimize the parameters to optimize how they are ranked on the metas or OTAs, I think that one is really interesting that's it thank you very much thanks a lot