 Thank you, thanks for the invitation and thank you guys for coming Let me quickly introduce myself so that you get a sense of what Where we come from and why I'm going to be saying the things I'll be saying for the next half hour Kernel analytics is a company that develops tailor-made algorithms that automate and optimize decision-making so we use data from our clients to develop Predictive models and then we deploy algorithms that help them in their daily operations basically adding more intelligence to current business processes Today I'm going to be making a case That is while all these developments individually in marketing in supply chain in pricing in Asset deployment while all these things Individually make sense and add lots of value Today I'm going to be making the case for the next step Which is that it allows coordination across departments creating some additional synergies that cannot be Obtained otherwise So right now I did a spoiler of my entire presentation. So basically this is my claim for today So artificial intelligence will allow extreme day-to-day coordination So not high-level coordination in which the two, you know, the CMO and the CEO talk to each other No, no, no, I'm going to be arguing that artificial intelligence will allow to Make tangible in daily individual operational tactical decisions So that these two things are Coherent yeah, so basically coordination between departments in a complex and changing environments in And that this will be a Significant competitive advantage that can be sustained and built over time So basically I'm going to be spending the rest of the presentation Describing at a high level this idea and then providing concrete examples in different industries and different decisions to make To make the case. Yeah so What is a little bit the big picture of before and after? So before in the good old days Departments were operating pretty much very with you know autonomously Coordination took place at the executive committee but not so much between departments on an operational level and Companies had information silos that basically was reflecting Departmental management silos. Yeah, so each department would hire their own application to run their show And as a result all this would end up in different information systems So hence making it even more difficult to get these operational Coordination yeah, so what I'm suggesting is that in when everything that I'm describing today happens and Believe me, there's a lot of room for improvement We will be getting like highly coordinated operations between departments at a micro level at an operational daily level So the executive committee still of course makes sense to align strategies But then AI the data science or the data engineering will make it happen On a low level. Yeah, of course, there will be a unique data lake and that is something that yes, it is more and more mature in in all companies But artificial intelligence as a low-level coordinator. This is not So much or as we can see it But yes, we can see some examples that I will share today of Tied a wave that is coming and that will be seeing Unfold in the next in the next years All right So basically the examples that I came up with of course, you know different companies have different departments that are you know Leading but the cases that are they'll be presenting today cover these areas. Yeah, so marketing both above the line and below the line so marketing slash CRM operations mostly supply chain Pricing And then expansion is you know typically asset deployment Yeah, so network deployment or it could be even point of sale deployment Yeah, and then of course strategy and finance are interested in aligning all this be it from a budget perspective or be it from a you know coherence in strategy perspective. Yeah, so Okay, so let's jump into it Let me start with example number one Pardon me, which is flexible and personalized promotions. Yeah, so this will involve CRM supply chain and pricing So let's start with Individual applications and then we see how the combination of those create some extra value beyond the individual applications. Yes so Imagine that you are a fashion retailer. So you you know The the season ends the clearance period starts and you have certain stocks for For individual products. Yeah, so it is inevitable. We all know That there will be discounts in this period Yeah, and when you have higher stocks, typically this requires higher discounts and the other way around. Yeah, and Yes, there is one You know individual application that is only pricing which is answering a complex question Which is what is the itinerary of discounts that I should put in order to get rid of my stock with the best possible margin Yeah, but this would as such would not fall in the Talk that I'm giving today. It would be a talk on itself markdown optimization. Yeah Then CRM and loyalty programs allow us so different story now CRM and loyalty programs allow us to understand Extremely the tastes and the sensitivities of each individual customer. We don't do things by chance So we can look at what we buy what we click what we abandon in the basket and Based on this we know that you know pow likes, you know boring blue suits and somebody else likes more shiny clothes We know who is price sensitive We know who's not price sensitive basically by observing their behavior. Yeah, so we know they don't do these things by coincidence So basically we know we have and when we develop recommendation engines Basically what we're doing is we're trying to you know personalize CRM Customer touch points one-on-one and again, this is a you know an interesting project on its own That allows to get you know uplifts significant uplifts in in CRM management Having said this the idea that I would like to present today is when we mix the two. Yeah, so basically the idea is Okay, I know that I'll be doing 30% discounts two weeks from now when the clearance period starts What if I did some below the line offers to some of my clients directly saying why don't you avoid all the Hustle of going to the clearance period You know here's my 20% discount on these individual products that you care for if you go now Private sales. It's all yours. So basically I can I can You know do a first pre sales period Target it one-on-one that Helps me sell a bit better clothes that I know that in two weeks time There will be at a 30% discount and from a CRM perspective the customer says okay You know, I'm not sure whether this will be available in the clearance period I know that clearance periods is a hassle in any case So I might be inclined to accept that offer simply because of the convenience. Yeah And the uncertainty of getting the right products. So basically here what we're doing is we're doing we're mixing Personalized Interactions so personalization and pricing and we're again Defending the margin a little bit better with respect to just pure pricing In the sales period. Yeah, okay. So this is Example number one. Let me move to example number two smart shopping list and Stop food waste. So here the idea is so now we're moving to a food retailer or yeah, supermarket more in general Yeah, not only food The idea is that Extremely from simply from a CRM perspective When we have recurrent purchases and supermarkets is one of those cases that we know Clients come, you know every week or every two weeks and they buy lots of things So and lots of these things we buy Recurrently because we need to eat and clean and so on so here There's one idea which is again individual which is how do we develop a smart shopping list? So basically telling our clients you should stop thinking about doing the shopping list Let me help you with that. But of course that requires some intelligence on on the supermarket side meaning If you I bought, you know, two bottles of shampoo last week Please do not speak about shampoo until I don't know two months from now Because if you are careful enough to be looking at how often I buy these things I will How often I will buy these things you will see that I'm not gonna need any of this for the next yeah a few weeks Yeah, so Okay, so this is something that can be done and we can see what is the frequency and and this works for everything that is You know shampoos toothpaste and all the basics. Yeah, and then of course when we go to the supermarket We also have some indulgences and these are indulgences probably don't need to follow this Stable pattern. It's more like again recommender engine say okay if you like pesto sauce You might also like these other Italian sauce or whatever. Yeah, so okay, so this is for as much as for the smart shopping list Telling people to stop thinking about this and then they don't make mistakes then also Given that we can track with loyalty cards or with or online We can also predict store attendance. So when people are going to be coming to our stores And the idea is to do that the ideas that we can In fare when it would be a good time for somebody To come to our store because we have lots of interesting things for that person that day Meaning if I like strawberries and I like whatever a certain kind of shampoo And if those two items are on sale with a heavy discount that particular day Then it might be a good idea to say power Didn't you come to the store today, you know instead of tomorrow you were going to come in any case But today we have this so it creates a clear call to action for that particular person to come that particular day and again and the stop food waste thing is Dynamic pricing for Perishable goods. So the idea is that I have all these apples I know that they're gonna be off in one or two days So I need to you know decrease my price or start putting some promotion or some discount on it So this again, it's something that is interesting on its own. It's dynamic pricing to sell with the best possible margin the The the products that you have so you create yourself a reputation for avoiding food waste, but at the same time you are Yeah, so you're you're you're maximizing your margin So you're doing both things at the same time and what I'm claiming here is that these two things are interesting But you can also add an extra thing on top of it Which is mixing the two so saying when it is a good time For a certain customer to go to the store and get all these things Yeah, and the idea is that all these things can be refreshed, you know In real time once we know how sales are going and finally, of course the supply chain bit of it is you know to ensure that there's Product availability in in all the supermarkets and all the stores and all the products All right Change topics now we move to network rollout. Yeah, so if if you are running a gas distribution company or a telecom operator you will be Deploying or spending huge amounts of money deploying networks Yeah, deploying pipes or antennas or fiber optics and so on and so forth And you'll be investing a lot in this and this is key for customer acquisition because if these things are not in place Customer acquisitions is impossible. So basically you would like to know when Where with a lot of precision you should be deploying these networks hoping that? You know neighbors in those streets would sign up for the service that you're offering. So The rollout of these physical networks are is extremely costly and with long-term return So basically what we would like to know in this case is Sorry one extra thing and yes In Spain and in any, you know developed country and even developing ones We have very rich data about people volume of people type of people buildings Roads own networks Competitors networks. This one is a little bit less detailed, but everything else is extremely precise So we know exactly how many people are living where at a building level So we can get extremely granular and what we would like to do here, of course is learning from the areas that we cover and moving to or extrapolating this knowledge to other areas that are not covered with your current network But it allows you to forecast how many people are going to be signing up on year one year two year three and so On and this is the benefit. This is the this is the revenues that you can expect from Making the investment. So once this is done We can do you know the net present value of these deployments at a at a building level or at a street Level and then of course here comes the The so CRM is in charge of acquisition of customers Expansion is you know in charge of making it happen deploying the network and then the finance guys, of course This has heavy capex implications and and when we're talking about deploying Telecom networks also optics implications. So So they try to make sure that everything is aligned Yeah, so that the that they get the best return on investment on these on these on these things The the the gas distribution example is very easy to follow then if you translate it to Mobile telephony it gets more tricky, but the logic is the same. So When you do a gas distribution, you need to have a not only you need to identify what you know streets like parts of streets Are profitable you also need to link them together because the gas needs to flow through the the network. Yeah So then the algorithm the engine needs to be smart enough to know where it needs to go and And how to link all the profitable bits and in a non myopic way Yeah, so it needs to be able to make little investments to reach profitable areas when you do this same idea for a telecom operator The logic is the same. It's simply that antennas do not need to be you know Connected with pipes necessarily but but definitely you need to simulate adding these extra antenna in the network or upgrading The capacity or the technology from 3g to 4g of these antennas What is the impact on the quality of service? What is the impact on the customer experience? What is the impact on business metrics of customers living in that area and then is when you can close the circle and basically These two examples end up in some sort of like bottom-up simulator in which you can In which you can bring together on the same table the finance guys who will say how much is this going to cost from a CapEx perspective and what optics implications it might have the CRM Slash strategy guy who says well I have all these you know customers that are increasing their data consumption and I want to you know launch this new Price plans with all these data packages on it And then the expansion guys as well wait I need you know This is the dimensioning that I need for all this to happen And then they it basically allows to coordinate a coherent strategy between the three of these until they converge. Yeah All right, let's move on to another example I'm Pricing and personal so personalized pricing and bundling so now we're going to the airline industry And here One could say okay. What prices should I have for my flights? But more in particular here. I'm gonna be talking about ancillaries so ancillaries are all these services that you are offered be after the After you buy the ticket. Yeah, so the insurance the preferred seating the The Boarding and all these extra things. Yeah food sometimes so in the past These were used to be the same for every route. Yeah, so if you go from Barcelona to Madrid then you know, these are the prices But really we have a lot of information when people are making that decision we know what kind of reservation and Sometimes we also know what Exact customer is behind that reservation so we can fine-tune these prices to increase customer satisfaction and Revenues and net because these are typically services with a very low marginal costs But high value for clients. Yeah, so basically, you know adding an extra suitcase in the in the plane or you know offering Priority insurance or the priority seat boarding. It doesn't really cost me much as an airline but It it is something that people are willing to pay the question is how much and of course if you do You know targeted pricing you can you can you can increase these and basically it's an industry that is already used to Different prices. We all know that when we're sitting inside a plane We know that nobody in the plane paid the same price for that ticket and we're cool with it We're simply extending the same logic to the ancillaries so So basically the idea here is How do we define these prices? And of course, this is a long story, but it it has to do with you know estimating elasticity based on massive a be testing and And once you have this it is a system that continuously challenges current prices targeted segmented Dynamic prices and until continuously finds the more optimal ones This again is a pricing Exercise only but it has very interesting synergies with CRM and both batch and in real time so batch is okay This is the prices that I would you know otherwise give to Paul Agulio Based on his past behavior. So we know that he's you know Appreciates very much checking in the suitcase and not so much some some other thing. So the idea is okay We screw up with power will you we lost his luggage Last time he flew with us so we can use this also To give favorable conditions or even for free sometimes some of these things because it has long-term implications in terms of customer value management and this is where the CRM bit comes in and of course pricing and CRM they need to be Coordinated on this front so that we're not charging assume that I'm a business traveler assume that business travelers get High prices assume that they lost my luggage and they get a high price for something else that I get extremely upset so the logic is that they you know that they can compensate and and And yeah, and they can so that so that from a CRM perspective they take care of a you know valuable customer like myself At the same time it could be in real time like sorry, mr. Agulio We are late and we're you know the flight is not going to take off for the next four hours So this again is something that can be applied in real time and again It needs to be consistent because if I try to you know Hire something else in that in real time for whatever reason then This is something that they can play around with but it kept it can be it can be also applied to the hotel industry With the ancillary services that hotels offer to their guests beyond the room themselves. Yeah so Okay, so this is idea number four. I think so let me move to promotion optimization. So I Single idea Promotion measurement has greatly evolved in the last year. So now companies consumer good companies know much better What is the profitability of their? of their promotion so basically answer answering a very simple yet important and Tricky question which is when we did the 50% discount on the second unit? Did we make money or did we lose money and this has to do with you know? somehow Reconstructing the counterfactual of what would have happened if I didn't do the promotion and then try to see whether the sales Uplift more than compensates the loss in margin the cost of the promotions and maybe some Cross product cross time cannibalization. Yeah, so all this is good And it you know, it's more or less where the state of the art is so this helps companies to understand past Profitability and it allows them to negotiate with retailers with better information So that they they know what are the implications of their of their promotions future promotions for their profitability Again, this is a single area, which is you know trade marketing. Yeah But of course there's another department that is marketing like above the line mark marketing So these are the guys that you know make ads on TV and these you know, they invest heavily as again And it this also has an impact sometimes sometimes bigger sometimes smaller on sales So it's good to not do double accounting and that the trade marketing gets their own benefits And the marketing guys get their own benefits Yeah But then you also have below the line. So this is CRM people Getting agreements with the retailer so that they can do coupon recommendation for their clients Yeah, so this is where the below the line thing comes in so basically getting the net profitability And what is the best way to do ads on TV? trade marketing for everyone that comes to the supermarket or personalized targeted coupons to the To the customers of the supermarket how all that fits in and how is When are we being too generous and when we're being not enough generous? So what is what is the combination of all these parameters to maximize overall efficiency? And of course once you've done this there are you know upstream implications for supply chain to make sure that you have enough product in all in all the stores. Yeah And also of course to optimize the advertisement Investment. Yeah, so again here. We're mixing marketing trade marketing supply chain and CRM all in one same cocktail mixer Let me move to assortments So here the idea is Imagine The company here what they're doing is It's a flash sales company. So here the idea is that we have Products Available for a for a limited amount of time. Yeah, so Imagine that you're selling clothing. So You know relatively high-end brands with hard discount, but they are available only for five days So you need to so customers need to make a bit of an impulse decision, which is Should I buy this now that it's 50% off or not and And so basically what these companies do is like they put new campaigns Which are essentially brands lots of products within those brands and a certain discount and then based on this they Yeah, they sell basically until they run out of the stock big brands try use this online outlet So to speak to again get rid of the stock from from previous seasons So this company Has developed an engine that helps Personalize interactions with each individual customer. So basically this We know perfectly how much each customer bought What they bought what they clicked what they abandoned in the basket As you know as far back as we want so years Yeah So basically we know everything that has interested you in the past and this of course helps us making these recommendations with a lot of precision and when you check it with a control group or champion challenger so the non personalized Group and the personalized group you see an uplift a Significant uplift that can be you know between 10 and 20 percent now the question so again This is a single idea. It's pure CRM. It works nicely and it's and it's and it's fantastic. Yeah Now the idea though is so this is more answering a commercial question an assortment question from these Online outlet company and the idea is what how should I what calendar? Should I have for all the different campaigns? To maximize overall sales and of course here the idea is that we might be Losing sales for two reasons reason one is what we call it coverage So basically there is a group of people a group of users in your in your in your customer base That finds nothing of interest that particular day Yeah, so this would be a mistake that the assortment people have done Because there are certain people that like I don't know sports clothes and that particular day There's nothing for them. So this is something that we can anticipate because we know through the engine We know exactly what is it that they that each customer cares about so we could simulate You know what would happen and we can see things like like this that that some that there's coverage issues The other one is a the opposite meaning this is a bit of an impulse Market therefore if there are two products or two brands that completely overlap It could be that you indulge yourself with one purchase, but not two in a given day So following this reasoning having the I don't know Adidas and Nike Campaigns on the same day would be a bad idea because they would be cannibalizing each other sales. Yeah So these are two things that we could discuss But we don't need to discuss because we have data to test these things So the idea is to try to understand how overall sales relate to the Coverage that you have so whether each user finds at least one or a few campaigns relevant for that particular day And secondly whether there are campaigns that overlap each other and they cannibalize each other sales So this is something that we can go to the data and understand and based on this We can simulate what is the best calendar for all their campaigns and and and when they should be launched. Yeah, so here the idea is That it allows me to better plan assortment definition yeah, so So how these things should be combined and of course this has CRM implications, but also operations implications And finance because we know when these things are coming in. All right Last example availability planning Okay, so imagine that you are, you know, the COO of Uber or Cabify or global on all these transportation companies The tricky what is core business for these companies is to make sure that The supply equals demand for transportation in space and time Yeah, so basically that you have the adequate but not too much but not too little like not too few drivers available Again in space and time so in the right areas at the right times of the day if you have excessive drivers You're creating frustrating in your own fleet and If you're and if you are short of drivers, then you're creating frustration on the customer side So basically your job is to balance these two things continuously and then if you leave the Markets so to speak to work on their own these two things do not match Magically it would take too much work too much friction too much frustration Yeah, until drivers learn when they should be available or when customers know that they can rely on this You as the COO know this better and and you can and you have some Marketing levers that you can use to make these two things match all the time. What are these things so? Well, these two things are incentives really so One thing would be Pricing yeah, so from a customer perspective if you're asking for a ride when it is more difficult to get drivers Then you know or there's scarcity say, you know 2 a.m. On a Friday night, then I'll charge you more so this is what's called search pricing and it's you know beautifully done by Uber and these kind of companies and and then and That's precisely the concept say okay, and of course there is an elasticity here at play Which is how people react to changes in prices, but experience says that there is some flexibility not infinite flexibility But some flexibility and be willing to pay a bit more basically because your alternatives are equally bad Yeah, there are no taxes available So I rely on Uber even if it's you know two euros more expensive or three euros more expensive at that particular time and day Okay, but then how do I make sure that I have enough drivers? And this is what extra benefits I can provide drivers to be available when I need them and this can be Again search pricing can benefit them as well. So no I'll pay you more But I can also trade with promotion. So have some sort of a CRM Aspect to it, which is I'll give you I don't know whatever points and then I'll give you preference for the Booked rides the one people going to the airport. These are good rides. They are booked in advance So I can you know assign them directly to individual drivers that you know took the nuisance to be available on a Friday night When I needed him so then I'm gonna compensate the driver later on so again, this is a CRM to drivers Dynamic pricing all this together in making an operations Which is making sure that we have the right balance for between supply and demand So again, it can combine real-time pricing But at the same time there needs to be some but the pricing can be real-time The CRM needs to be happening a bit in advance and that's why you need to make forecasts Yeah, you need to know how many drivers you're gonna need on a Friday night so that you can make the offer to the Driver so that they can make plans and not have a dinner with their friends that particular day because they need to be available Yeah, or a few of them. So again, we're mixing in Pricing CRM and operations on in one same layer and one same engine that is, you know, pure core business for for these companies so by now, I hope that I have convinced you that that this that Extreme coordination adds a lot of value in addition to the individual value that each of these applications You know Cause in in in individual department, so I'm not saying that these individual Advanced analytics applications don't make sense. Of course they do but what I'm saying is that there is this extra Value to be realized to be made once you bring the two together And I came up with some examples of pricing Network or asset deployments marketing CRM and operations and of course all these things And if you're familiar with this kind of Initiatives, you know that there is one part that is strictly analytical, which is you know again artificial intelligence predictive models But then there is a business layer to it That is does this make sense are we creating value with this and this is the strategy finance guys that make sure that we're investing in You know CRM for drivers is more than compensated with the revenues that we make for for customers So This is it for the talk. This is all I wanted to say we have five minutes for questions if anybody has a particular question Well, thank you Thank you Congratulations Paul for the presentation. I was late from other sessions. So maybe you talk about that before to start however Many Companies who are deploying Customer or package CRM ERP and so on are talking right now about putting these models inside the CRM Inside the ERP and one of the problem of this is to manage the lifecycle of the models and the versions the control and so on I think so to get extreme Coordination maybe the best should be to put inside the CRM inside the ERP inside the The the kernel applications in the companies How What do you think about to put it inside or outside and coordinate an extra information from these systems to to make Available the models and so on Okay, typically it's a very good question and it's far from being resolved I think I think different companies are doing different things, but our opinion is You know, you need some Flexibility to be maintaining and evolving these models. So models do not you cannot forget about the models that you're running You need to keep an eye on how they are performing all the time So basically we what we suggest is okay, you have here your data lake Then you have an an environment in which analytics takes place There is a you know the analytics toolbox and analytics sandbox in the toolbox is where you have all the deployed models that are running our experience is that If you don't need real time Probably it's better that they sit in this, you know analytical toolbox We call it so that the you know, you have full flexibility to manage the algorithms And then the results are plugged in the ERP or the operational systems basically because Yeah, you don't want to slow down the the you know operations But again, it depends on the flexibility that the ERP offers in terms of having those models run inside So typically we favor being outside and then plugged in. Yeah, so typically there's an ERP and This is a bit of a analytical bypass and then it goes back in and then it's all the but but again They're you know pros and cons so I had a question about surge pricing. Yeah so it seems like a very efficient idea in terms of Managing supply and demand from a customer point of view if there's a high variation in the amount of money that you pay You could get negative emotional reactions from the customers I just wanted to know how big an issue this is and how companies are managing that. Okay. Well the First of all, it's something that they know in advance before they book So so you're you know, I would like to go from here to downtown Madrid And then if I'm you know making this request at the wrong time to am then they say, okay It's more costly, but I can still make the call. So it's not it's and it's not unexpected Exposed pricing that I get it's upfront so I can I'm still in control of the decision and Of course, there are limits to this so and I would say That it's within certain boundaries is something that is well accepted I think that it can be You know counterproductive if it goes out of these certain ranges or in certain occasions say that there is I don't know a fire and then you know Your search price algorithm if nobody is looking actively and knowing why is it that there is so much demand You could say oh, hey great great demand, you know increase prices and then you you can get like extremely negative reputation out of this But but of course, this is something that can be tested and that can be And so you you know what are the limits of this? So if you have high prices said say that you go too far in that you can see people, you know making a request and not booking so so you have all this information and you So the short-term impact you can measure very accurately the long-term implications. It's a bit more difficult to measure Yeah, so basically if this uncertainty makes me not even ask for a quotation, then it is and again This is something that can also be measured to see whether people go unactive when they have been exposed to extremely Diverse prices. So the beauty of this like data natives companies is that you can pretty much Model and measure all these things But so all this to say that it can be measured it can be optimized from a business perspective My impression is that within certain boundaries is something that is well accepted as it is for flying as it is for hotels Actually, I would argue that the variation in this service is a lot In in in rides is a lot lower than than than hotels and airplanes Okay, I think we ran out of time if anybody has additional questions I'll be happy to answer you can find me at the at the booth of kernel analytics. So thank you very much you guys