(inspiring music) >> Welcome back to The Cube's coverage here in Las Vegas. I'm John Furrier for re:Mars coverage. Two days of live action, a lot of things happening in space, robotics, automation, and machine learning. That's Mars spelled backwards, but that's machine learning, automation, robotics and space. Got a great guest, Jens Ortmann, associate director at Boston Consulting Group, also known as BCG. Jens, welcome to The Cube. >> Thank you very much. >> So tell me what you're working on. You've got a very cool project you're working on, 'Involved'. Take us through what it is, explain what the project is. >> Yeah, so I'm part of the data science unit within BCG Gamma and I'm focusing on solving business problems for the automotive industry. What I would like to talk about is actually a small internal site project that we were building. It's a conversion rate engine, where we built an advanced analytics tool that computes the conversion rate for car dealerships, at scale. So for every single car dealer in a market, we can compute the conversion rate. >> John: What is a conversion rate? Can you explain that? >> So a conversion rate is very simple. It's actually out of the people that come into your car dealership, how many do you, as a car dealer, manage to sell a car to? >> So, what's your sell, through monthly kind of- >> Per visitor that come into, so your walk-ins. >> So, physical? >> Physical, yeah. So this was for physical stores. It's actually a key metric for sales performance for car dealerships, or for the automotive manufacturers to be aware of. >> So I'm watching here in the show floor at re:Mars, you've got the 'Just Walk Through', which is Amazon's 'take whatever you want and go', are you seeing you're getting analytics on like people coming in, you can see them, there's a drop off rate? Take me through how it works, the challenges because I don't envision like, "Oh, so they walked in and they left but they didn't leave with a car." It's not take and walk out, it's not grab and go. But the concept of using computer vision, I can imagine it being a popular thing. So how do you measure this, people coming in? >> It's actually a big challenge that we learned when we were doing this project. Traditionally, people were measuring it with like these laser sensors but the signal is very, very messy. Now when we wanted to do it at scale, we partnered with an Israeli startup called Play Sense, who aggregate mobile phone data. So we used mobile footfall data to measure how many people visit a store. So it actually is a combination of three main data sources to get to the conversion rate. One, as I mentioned, the mobile footfall data, the second one is building footprints, actual outlines of buildings that we source from the cadastral agency that we need to use it to cut out the footfall data to get the visitors. And the third one, of course, is sales that we get from the official car registration data. Then we combine those to have the key numbers. >> Is there a facial recognition involved in this? >> There's no facial recognition involved. >> So the tire kickers that come in and kick the tires and leave, but might come back. Is that factored in too, or? >> So there is a lot of pre-processing going on to really only get the signals from visitors. So filtering out people that maybe come into the store after hours, cleaning crews, people that come into the store every day, people that work there, they would be in the footfall data. So we applied some logic to identify exactly those people that are most likely actually visitors interested in buying a car >> Well everyone can relate to buying a car, obviously. I wanted you to step back and you mentioned scale. Can you scope the scale of the problem for us? How big is this observation space? What systems are involved? 'Cause when you say scale, I'm thinking all the dealerships in the aggregate. Or, is it by franchise or is it anonymous data? Can you scale the scope of this thing, or scope the scale? >> So we built this as a prototype for the German market and we used the top 10 car brands in Germany. They have around 10,000 car dealerships, for which we all have data. The actual mobile phone footprint data, it's a lot more. I think it was 30 million data points. >> Are you triangulating that? How does that mobile data work? Signal? >> So the mobile data is coming through apps. So mobile apps where you allow the app to track your location. >> Got it, okay. >> That gets anonymized and then you have these mobile data aggregators, like Play Sense. >> Got it, okay. >> That sell the data on. >> So you have to plug into a lot of systems? >> Yes. >> To make all this work. >> Yes and a lot of different data sources. >> And how easy is that? What's the challenge there? Is it cloud integration? How are you guys pulling this together? >> So we build it as a prototype initially, based on our own internal infrastructure, using basic Python and regular cloud infrastructure to process the data. >> All right, so I'm looking at my notes here. Data sets, you have a lot of data sets. What kind of analytics are you running on that? Can you share some examples? >> So I have to be careful since we filed a patent on this but a lot of the thing is actually in data processing, making sure that the data points we get are accurate and usable for this, and then differentiating between the different types of businesses that people are running. So there is on the one hand, you have the problem of outliers, basically filtering out when numbers don't make sense. On the other hand, there is a lot going on in the business itself. Like what do you do when a car dealership sells cars of multiple brands? You see only one visitor seeing cars of different brands but you see sales for two different types of brands. So this is just two examples of some of the processing that we had to implement to make this happen. >> So where can people find out information on this project? Or is it pretty much not public? Are you sharing anything publicly? >> So currently, we have held off the publication on this because we filed a patent on it. We're now about to go to market, building out a solution for the US as well, to then bring this to clients. >> What do you think about this show here at re:Mars? What's your assessment of the vibe? What's it like? Share with the folks who aren't here, what's your takeaway? >> It's really fun. It's really impressive. And it has a great, really inspiring vibe of cool innovative solutions. >> Yeah, you get the creative geniuses, you got the industrial geniuses, you got the software geniuses, all kind of coming together, and they're real people and they're here as a community. To me, the positive future vibe of this show, really is resonating in the keynotes and the energy. It's a forward thinking, positive message. And it's not marketing, this is the vibe. >> Exactly, I think it's something we really need at the moment. >> Yeah, we can solve all of the global problems by going to the moon and Mars. First the moon, then Mars. Who knows, maybe the breakthrough is there. >> People solve a lot of fundamental issues along the way that'll help in a lot of different areas as well. >> I wonder if I'll be alive when there's tourism in the moon. I was just joking with the folks earlier, "Oh yeah, I left that part on Earth, I have to go get it." Cause there's going to be a whole infrastructure there. Construction, all in good time. Okay, what's next for you guys? Tell me what's next on the project. You got a patent pending, so you're a little bit tight lipped and quiet on the secret sauce, I get that. What's next for the vision of the project? >> So this is just one example of how we can use this. Especially this footfall data set in an innovative way in the automotive industry. What we would like to look into is getting more details. Currently, we only see a single data point for a visitor. What would be interesting to understand, also, like the journey of visitors. Did they visit other car dealerships? Or, where are they from? What demographics do they come from? If you can tie that to a geographic location. And then on the business side as well, linking this for example, for companies to marketing campaigns. Does advertisement catch on? Do discounts catch on? Do they drive more people into the stores? Do they drive more sales? How does it affect conversion rate? Also, benchmark within the network, how different car dealerships are performing, how different brands are performing. And then eventually, everything is going to online. This can also be a foundation to set a baseline for online sales, which is still at the very early stages in the automotive industry. >> Yeah, I think there's a lot of reference implementations here for other applications, not just dealerships, all footfall traffic. That's interesting. The question I have for you, and the final question really before we wrap up, is the convergence of online, offline, physical, virtual. It's pretty clear we're living in a hybrid steady state right now, with all the post pandemic and the innovations pulled forward. So, having a device on me, IOT device or phone, will be a big part of things. So I'm buying online and I'm walking in, I'm one presence, virtually and physical. How do you guys see that around the corner? What's next there? Because I can see that coming together in my mind. >> It is. I mean, we can see it happen at Tesla. Tesla barely has any physical dealerships anymore, they have showrooms and do all the sales online. And I think that has a large impact on the industry at the moment. Driving the more traditional manufacturers also to think about what can be and what can be in a digital and online first world. >> Yeah, well this is happening. Well, Jens, thanks for coming on. I appreciate the commentary on re:Mars. Thanks for sharing your perspective and sharing about your project at Boston Consulting Group, also known as BCG. >> Thank you very much. >> Very reputable firm. Okay, that's the Cube coverage here at re:Mars. I'm John Furrier, your host. Two days of wall to wall coverage here. It's a great show. Machine learning, automation, robotics, and space, Mars. Of course, you got Reinvent, the big show, and at Reinforce, the security show. You got the space-software-robotics show, security. And then of course Reinvent is the big show. The Cube covers it, all three will be here. So keep watching here for more coverage. We'll be right back. (gentle inspiring music)