Location-Based Intelligence for Smarter Marketing with Klustera - #18

Key Takeaways

[Music] all right everyone I am here with Paulo Martinez and Carlo Rodriguez from cluster uh live once again from NYU future labs and the AI Nexus accelerator uh guys say hi hi hi hi to everybody hi everybody um so let's just jump right in and have you tell us a little bit about cluster what are you guys up to sure well we we started as a retail analytics company like enabling Carl and I worked at a marketing Innovation firm okay and and we were wondering what could happen if uh Walmart had the same data that Amazon has how could they grow their businesses and and the pursuit of that uh response took us to where we are now we track people's movements on physical spaces based on their Anonymous Wi-Fi signals okay and and we got some Traction in that space But then we found out we realized that that wasn't like where the value resided right like just by tracking the people what we understood uh like working with some of the major players in the market in Mexico such as uh aner Bush Walmart and valaris that a very large Airline there there is that the real value resides in um delivering relevant advertising to the to the users based on their behavior in the real world MH and that's uh where clera is now at so you're not no longer focused on the Wi-Fi movement element of it or is that still part of the still part a building block it's a building block but instead of using the selling the data as a dashboard or something like that we use that data to correlate the behavior of the people with their profiles Bas on the data that we get and then we target those people okay and so the the general idea then is to allow so-called brick and mortar retailers to have more of the kind of give them you know what with a an online retailer you know exactly what the customer does from you know their page path as well as you know maybe even heat map ey tracking that kind of stuff and so this is is trying to capture some of that same kind of insight for the brick and mortar retailer yes to capture that Insight but also to activate it okay right we realize that uh for example Google and Facebook they're massive companies which their business model resides just in in one simple idea that is to make advertising more relevant to the people right and that's why they're so big but they are just like online to online the things that you do online they they get transformed into online advertising but no no company yet has uh really tackled really nailed the how to take the offline behavior of the people and translate it into online advertising which is important because adver online digital advertising is growing like two two digits per year since uh I remember so it's going to get just bigger and companies need that data to to better deliver their campaign mhm and so are you telling them you're telling them with the data that you've collected um who to Target um but what else are you telling are you telling them how to Target them where to target them or uh are you um and are you providing a an advertising Network or are you telling them which advertising networks to use tell me a little bit more about how a customer uses what you're providing sure well uh first I'm going to give you like a glimpse of how how the solution works and then I'm going to let Carlo like go a little bit technically deeper okay what we do is to we interpret the we hear the the Wi-Fi signals of the of the smartphones right they have a um all the audience must know that the MAC address is like the anonymous device ID we hear that and we can pinpoint where is that uh device uh located in a very very accurate um way like instead of uh Geo fencing that has a 500 M like uh accuracy we have a 10 m accuracy so so that's the first thing second as uh people well Gathering a lot of data from the city like the Google car does we also map out the whole city but the routers in the city the by networks in the city so we map the city literally with latitude longitude exactly where are the by SPS commercial office space and also residential oh so we're not talking about movements within the store we're talking about movements out in the world of both uh like inbound and outbound okay well or outside the store like on uh Urban level like at the city level but also inside the store okay and so you uh you are collecting all this this data based you're basically locating the individuals and then what are you doing with that data how are you aggregating it well we we aggregate it as analytics as car said dashboards like so our clients can understand uh when do the users arrive how much time do they spend how do they move inside the store but also as users naturally connect to Wi-Fi on certain locations then with the with with the optin like with the consent of the US we can match their their device IDs to their online profiles and in that way we can like make the the whole cycle and start delivering smart campaigns uh for example one of our clients is uh major Burger chain in here in the US MH and they're using our solution in for a lot of stuff but but one of the the main points is to to understand how people like uh they loyalty their recency how how often do they go and if they notice a pattern and that that's where one of the the the ways that we use machine learning and AI for for our product if we identify that the the user is going to churn like it's going to bail out for a competitor because we compare their behavior to a lot of uh past behavior from other profiles MH uh then we send an aggressive campaign to that to that user because we know that he's like in danger of bailing out of the of of the brand and so to to make that a little bit more concrete I'm uh I'm a you know Burger chain a customer if I spend more time in an area that you know you might surmise I'm sitting in a burger chain B is that an example or are you finding deeper patterns that would suggest that I'm might be more likely to turn well there are uh three main like things that we look at in building the model okay first is as people connect to the places well generally they connect to the places they go uh to eat like to have food that we we can have like a sense of of the comp competitors of if you go to somewhere we can know if you have connect to some competitor M first second we can uh understand where people live and work based on their previous Wi-Fi locations M so we can understand the the uh socio economical status of of the person in that sense MH that it's like um they comply with with that and third uh we see the the patterns of the visits just like in Google analytics or in any like analytic web have Analytics suit you can see how often do the do the person goes if they go daily if they go weekly if they go two times a month and then we compare that and we see oh we have seen some similar behaving uh customers that after bail after not going to to the Burger chain in 3 weeks they we never see them again MH that's when we we know that we have to activate some campaign okay okay interesting thing tell me a little bit about your backgrounds and how you arrived at this problem yeah well as P mentioned before we met at the marketing agency I was the head of innovation my background is a telecommunication engineer I always been a geek since I was a kid I learned to Cod when I was 1 just for fun and and that thing is that put me all through this through college and everything I went to singular University and I was being I am a maker I build drones and stuff and I like to go that that's go that's okay sure well we we met there and at a marketing agency I was a BP of analytics okay and the all my whole career was in the marketing and analytics space I majored in well minored in marketing but I majored in applied statistics and then I studied a master's degree in analytical Sciences MH so I always like try to to solve marketing problems using uh analytics and data and Carlo as he is very strong with with technology and sensors it was like a perfect match for this Venture MH uh one of the things that I'm sure you get asked a lot is around the Privacy implications of what you're doing um how do you you know what are the the questions that you're most asked around that and um you know how do you respond to those sure well well they ask us um uh a lot of about that because it's it's like a sens sensible topic we know that we're not the first company to aggregate customer behavior and activate campaigns uh a lot of companies have done that before and we know that we're not the only company that tracks people's movements in with cell phones with their cell phone signals and but first and foremost we know that if we that we have to put privacy at at like the first of every decision that we make MH because um at the end of the day we are not interested in individual behaviors but on aggregated behaviors like that's why we are called cluster because of the cluster analysis that we do for the people so uh in terms of how the the two main questions that we get asked is is this legal and the answer is yes uh we we have a a very uh solid legal framework that that uh like allow us for example to gather aggregated data and make sense of it in terms of of dashboards of analytics without infring any any kind of Privacy Law or or anything nor in latam or the US MH and second is well if it's Anonymous how you can activate campaigns and and the the answer to that question is that when the the user opts into our platform they they comply that they we can use their their data to to send more relevant uh advertising and that's what Google and Facebook does they take uh for example your birthday and they send some special promotion because it's your birthday I I don't think that anybody can like argue or be mad about it and so are all of the all of the users opted in users and what are they actually opting into is it uh some app that you're providing or something that the customer is providing that they want to get some value from it's uh the Wi-Fi portal get Wi-Fi networks okay so and you you guys are running the the Wi-Fi networks or you have agreements with existing Wi-Fi networks to do this both both yeah we can go and install the just the sensor or like our Wi-Fi sensor it's like a very small router okay that goes up in on top of the existing Wi-Fi technology that all of our clients have or if there are some hotspot companies that we're partnering in Mexico where we just um like uh put our platform in their existing sensors and then we can match the user profiles I was going to ask about this earlier and um we we moved on to something else what the degree to which the infrastructure needs to be changed in order to support what you're doing and it sounds like uh it sounds like yes um so does that mean that in order to in order to grow you need to um like who do you have to who do you have to go after these Wi-Fi portal owners basically what are they but they're not the beneficiaries of the analytics necessarily that's the retailer so what's in it for them how do you make them want to get engaged with you guys well first we give them the ability to to better serve their clients without spending more like without because they give a very like simple dashboard simple analytics of the Wi-Fi and we told them oh if you if you put this code in your your portal then you will be able to to give a better product to the client that is paying you and you don't have to pay anything okay it sounds like a interesting and complex ecosystem so you got the portal folks that they maybe are their C their client is like Starbucks or some com someone that's hosting that's providing at their location uh and then you guys are trying to use the data for the retailer or just in the case of the retailer or small chains they already are providing Wi-Fi to their customers and in this case These Chains they also well we get on top of what they have and in this case this many a chain always have hundreds maybe locations and that's how we also okay okay makes sense um so maybe a little bit more on how you're using machine learning and AI to solve it what what kinds of you know techniques and approaches come into play to solve this particular problem well this as I mentioned before this this you see the part of the analytic is a building block towards the end product that ising out audiences great targeting H profiles for the people in the case of the analytics literally to measure people we need to identify which people is a visitor of the place and a passer by since from that perspective we need to classify those those those type of customers H the road data that we get is a very messy messy data because it's a lot of data is coming from cell phones is El signals from the cell phones and we use models to classify which one is in the store which was was passing in front of the of the store and then after that classify them in buckets like okay this is a newcomer is a returning visitor etc etc and also to predict future Behavior we we get that asked a lot like where is the AI and what you have because in some other companies is very straight forward right however in in ourselves uh we think like 10 years ago maybe it was or 20 years ago it was very uh hot to say you were an internet company and now it's it's very hard to say we're you're an AI company and we think that that it's going to like how do you say that like it's going to permeate in the in all companies so maybe in one or two years it's not going to make sense to say that you are an AI company because everybody is is doing that it's like you say I'm an electric electricity Energy company now right like just because you use electricity it's a tool you see it's a tool yeah yeah you are in the tool business selling pickaxes well you sell pickaxes in this case we use the pickaxes to build stuff right right and the folks on the that listen to the podcast like to hear about what kind of pickaxes you use and it sounds like clustering is a big uh one of the big pickaxes you you want me you want me to get really technical on this sure sure well you see we have many many many branches here first of all is that all the all the is coming for that of course we use the cloud otherwise we don't have any any any on premise stuff and then uh first of all all this Cloud go through data pip that this is run on spark on SP we use two classifications is a logistic regression and a decision TR and based on that was the first one H logistic reg logistic regression got it analysation Tre and and with that we get to the analytics site and and the first cluster let's say about that then we get the models for the clustering of the of the behavior of the People based on the profile and not on the behavior they they they have shown before yeah that's my favorite part because we started using like hierarchical clustering but then we we started pioneering with a technical not pioneering but using a technique called self-organizing Maps H that that it like takes the all the features of the of the people and like makes like like self organizing map it's like a technique uh used more for like Machine Vision okay where we're using that to to make sense of all the the information of the of the data of the users and gather them into very more more smart or smarter uh segments that if we use just uh plain clustering okay oh interesting interesting yeah and also we have different layers of information that we get from the people in the case what is just analytics well it's just an ID It's Your Mark address the MAC address doesn't say anything about that's why we are legal because it's not personal identified information the MAC address doesn't give you the name doesn't give you the phone number doesn't give you the email doesn't give you anything just the brand of the phone MH uhuh okay where how do you behave inside the physical place but in the case when the people op in in this in the Wi-Fi spots we ask them to put an email or in this many 90% of the cases H Facebook login social login okay and once we get the social loging we can we can know the gender the birday and the likes what pages do they have liked before and based on that we can literally create a really good profile of what about about you because okay I see you in a hot spot that is in a brewery or in a coffee place and then I see you another city etc etc is is literally reflecting your behavior right and then with with those clusters with those audiences is where the way we make business okay oh great uh well that's exciting stuff can you maybe share anything else you'd like to add or or maybe you can share how folks can find you and learn more sure it's um well that first that we're uh very thankful for the interview and we're very excited to be in this space and in in New York I think it's a very thriving scene and uh that for all our listeners that uh we would like to to tell them that yes there are a lot of uh there's a lot of hype uh in the space and like we we we were marketing themselves are yes the machine Learning Company but nobody understood that and then when we started solving the problem using all the pigs and and and tools and hammers that Carlos said that's when things started moving around so so maybe like have uh this change of of of um perspective about the AI in on startups I think that's that's important and we we're based in Mexico City so you can come visit us anytime you want and we are at clera k l u s d e r a.com okay great great well thanks for being on the show much thank you thanks [Music] for