In conversation with Kamaan Agarwal, CTO of Metro Ride

Metro Ride is a last-mile connectivity service that aims to connect people to local metro stations in a convenient and affordable manner. Currently operational in Bengaluru and Gurgaon, it makes use of a mobile app to connect passengers to a rickshaw that takes them to the closest metro station. This service is powered by AI that makes it more efficient and affordable. We caught up with Kamaan Agarwal, the CTO of Metro Ride to understand what makes it tick.

Team evo India: Could you tell us about MetroRide and what made you start it?

Kamaan Agarwal: So Girish (Nagpal) and I have been friends for over a decade and half now, 12-15 years. We were roommates in Pune back in 2008-09, stayed together for a few years and always had it in mind that we would do something together. But then life happened. We both got married, I moved to the US, he moved to Bangalore. In 2019 I was visiting India, in January, when I met him, and he was cribbing about his commute to work from where he lives in Bangalore, to the main town where he works and all that. And I thought the metro would solve all of that. But the problem there is, yes the metro cuts down my commute, but getting to the metro station kind of normalises it. It’s actually more painful to get to the metro station itself. And I was able to relate to it, because where I stay in the Bay Area, I have to travel to San Francisco city almost every day, and I use the Bay Area metro, which is called BART (Bay Area Rapid Transit) and it’s exactly the same problem. How do you get to the BART Station? There’s no commute option, like, taking an Uber or an Ola or a Lyft everyday is expensive, not everybody can afford that. That’s why we started in 2019.

We went to the drawing board, we did some market research, we spoke to 300-odd daily commuters in Bangalore, and I spoke to another 100 in the US. And there were some things that really stood out, which were very common across the two countries. Things like the priorities. For a daily commuter, it’s a different class of uses altogether. These are the people who would be using a ride hailing service 2 to 4 times a day. So they are very cost conscious. Anything that they spend is going to multiply by 60, for the whole month. The first priority that everybody said was a service that was affordable, that we wanna use multiple times a day. Second thing was that even if it is affordable, it has to be punctual too, because we wanna use it for work or for school, or college. So these are the two major things that stood out. Another thing that came out was that almost none of them said that we use Uber or Ola or Lyft to get to the metro station. It’s expensive to do that. And that’s where we started. We also participated in some awards and some industry validations, and we got a very good response and we won a couple of awards even before we started. And that kind of validated the idea. And in January of 2021, we were ready with the product and we launched Noida and Bangalore. There were a couple of stations in Noida and a couple of stations in Bangalore that we started with. We went through a couple of fundraise cycles and all of that to get some seed funding in. Finally, right now we’ve served around 200,000 customers in the past year, even when there were a couple of months of 100 per cent shutdown in most of our markets. We were still able to do around 200,00 rides, a little more than that. And we’re currently present in three cities, mainly Noida, Bangalore and Delhi and we’re close to 150-odd drivers on-board our platform today.

Team evo India: How does MetroRide work for a customer?

KA: One of the things we also found when we spoke to these daily commuters, that they’re okay to walk a little to get to a pickup point. So when we started building this solution, there were two main KPIs (Key Performance Indicators) that were important for us. One, make the ride cheaper, two, make the ride faster. So everything that we built there on, has to affect one of these two KPIs. So one of the things we did was we post-mortemed the Uber model. We tried to find out why they're expensive. And what we find out is that they're expensive because a driver spends a lot of… wastage in terms of dead miles and dead minutes. So when you book an Uber ride, there’s a driver who’s a couple of miles or a couple of kilometres away, he’s gonna come to you, look for you and that’s why he’s able to do lesser rides in a day. Which is why, at the end of it, the customer pays for that price. So that was the first thing that we removed. So we asked the commuters, would they be okay to walk a 100-200metres to get to a point where they would be picked up. And they were all okay with that, until it's affordable. That’s the way we designed our system. Like its a fixed route, fixed stop, ride hailing service which is 100 per cent shared. When you open the app, it would show you the closest pickup point from where you are, which might be 100-200 metres away. It will ask you to walk to there, and you would do what you would normally do on Uber, you’d select your drop point and you’d book a ride. That drop point would probably be a metro station, and can be another point also within a five mile radius.

Team evo India: So these are all pre-determined stops?

KA: Yes, so we use our AI model to identify the right routes, and the right stops. Our objective is to be able to do more number of rides in a day at a cheaper price, only then we’ll be able to meet our economics, so that’s why we planned this system in a way that we make people come to high frequency stops, so that a lot of people can come together over there and the vehicle utilisation is the highest. You would walk for 100-200 metres, a driver would come pick you up, you’d share your OTP just as would on Ola, he’d pick up a couple of other passengers on the way and he’d drop you at your designated stop, and it is the same route, day in and day out for the driver. So they’re extremely familiar and efficient on those routes. That’s the whole idea of removing inefficiency from an unknown route system to creating a known route system where a driver can do many more rides. The outcome is, the driver makes more than what they’d make driving a metered auto, and you as a customer would probably pay half the amount you would pay for an Uber ride. That’s the outcome that we wanted. Second main KPI was faster rides. Our average wait time for this whole year is 2 minutes and 1 second, which is much lower than the industry average. The industry average is 15-20 minutes, and our wait time is 2 minutes and one second. So we delivered what we promised. We had promised 5 minutes but we delivered two minutes of wait. That’s why we became popular in the areas that we launched.

Team evo India: How does it work for the drivers?

KA: That’s another differentiator that came out of the Uber post-mortem, as we call it, and what came out was that one of the main problems that Uber has is capacity planning, which is why these surge prices and peak pricing all that comes. Because they cannot predict how many drivers will be online at a given time, because their drivers have a choice of not logging in at say a peak hour. Our drivers are not cab drivers, they’re auto rickshaw drivers. So these drivers, when they come onboard with us, they’re full time with us. So for the said number of hours which might be a half shift, which is the first half, or it could be the full day shift of ten hours, they drive only with us. And for that we give them a minimum guarantee, so every driver gets a minimum guarantee from us on a daily basis and they also get an incentive on doing a certain number of rides.

Team evo India: Could you tell us more about how the AI works and how it optimises itself for your routes?

KA: We wanted AI to be the backbone or the nervous system of this business, which is why we made our own AI model, a platform called VIKI (Virtual Interactive Kinetic Intelligence). VIKI is like a central system which is used in every process in our business starting from booking a ride. If you want to book a ride, see we're a new app. A lot of customers when they’re putting a new app on their phone, or when there’s an issue with the internet or connectivity, or they don’t have a smartphone, for that we created another AI based platform which is a chatbot, where using an SMS or a Whatsapp number, you can book a ride, you don’t need an app. That was the first AI use case that we built. Once you book a ride, the assignment, who’d be the driver who’d be assigned for the ride, there are close to 35 variables for that, things like distance, number of rides already done, their gender. So if it's a lady rider, we try to club them with a lady driver for safety concerns and also better comfort. That’s the second place where we use AI. Third is how do we identify these routes. The way it optimises these routes are so we identify a metro station and our AI model pulls the data from Google Business Index. When you search anything on Google Maps, on the left side you see a histogram, which tells you the footfall or traffic at that point at various times of the day. So we use that data, and using that, we find out which are those spots which have the highest density near a particular metro station. From there we create our routes. And we also know for customers that are coming to the metro station where they are coming from, where they are spending most of their time, that location intelligence is also a part of this VIKI engine. This was about the customer. Now the biggest problem in this business which we realised after we launched was fraud. The ride hailing has a very high fraud rate, where a lot of fraud rides are booked, the driver comes to and says sir why don’t you cancel the ride and why don’t you give me 40 rupees less or 50 rupees less, that’s a big revenue leakage for companies. We realised early on that we’ll have to control this. That’s why installed this AI enabled IoT cameras in our vehicle, this camera what it does, it does not record a video of people sitting in the vehicle, but what it does is it takes pictures of inside the vehicle every 2-3 minutes and that picture is automatically read by the AI engine and it counts the number of people there in the vehicle and it matches that with the number of rides that have been booked at that time on that vehicle. If the number of people are more, it's a fraud ride. Everything that we do in this business, we try to solve problems as soon as possible and try to find a long term solution for it using VIKI so that it's automated and has less human involvement.

Team evo India: Have there been any concerns from your customer over something like privacy?

KA: If you’re talking about an Uber or an Ola, or even an airport ride, you’re probably sitting there for an hour or maybe a couple of hours. Our ride is a ten minute ride, it doesn’t really matter. For ten minutes, if you’re sitting in a monitored system, our users don’t really object, that’s one. Two we very clearly show that there’s a camera in the vehicle, and that the vehicle is being monitored. And three, we don’t keep any data. So when it takes a picture, it reads the number of people sitting in the vehicle and deletes the image right over there. So we don’t carry those images forward. It’s meant for a particular purpose. It has a two fold purpose actually. One it counts the number of people, two it also shows the people not wearing masks, which helped us a lot during COVID. So the AI helped us identify if the people are wearing a mask or not, or more importantly, is the driver wearing a mask or not, or if a driver is drunk or drowsy. All those things, the camera helps us with. It's a cost benefit analysis right. The benefit that we’re getting for monitoring, with say 5 pictures for 2 minutes, the benefit that you’re getting for safety reasons is much higher. We did not really get a concern yet from any of our users so far, as to why do you have a camera here.

Team evo India: When getting drivers on board, have they been open to the platform, or are they apprehensive?

KA: In the beginning there were challenges, so we had to tweak and re-tweak our model to ensure they got what they wanted. See in the end, even the drivers are our customers. So for the platform, the drivers and the riders both are our customers, and that’s the reason we created this model with a minimum guarantee, because these drivers have to go home with a certain amount of money and they need that predictability with their income. Right now their income was not predictable, there were days they’d be making 1500 rupees a day or 1200 rupees a day. And there are days when they make less than 500 rupees a day, and they cannot even pay the rent for the vehicles through the fare that they collect. But with us, there is a fixed income that they get everyday. That’s why drivers come with us and stay with us. Driver churn is very low.

Team evo India: What would you say was the biggest challenge of setting up MetroRide?

KA: The biggest challenge was the drivers. The miscalculation for us was we planned this system not knowing the difference between a cab driver and an auto rickshaw driver. Their literacy of a digital system itself is very difficult, that’s the biggest challenge that we faced. How this went was, we launched a pilot or a beta in December 2020, a month before our actual launch and we ran it for a month in Noida, I had my tech team sit in these vehicles the whole day to monitor how are those drivers using the app, and the feedback was, if we wanted drivers to use our app, it had to be as simple as a calculator, otherwise they will not be using it. That’s when we went back to the drawing board, we took another couple of weeks, revamped the full driver app, and it was as simple as a calculator

Team evo India: In terms of scalability, is this a model which you can replicate in various cities and how do you plan on growing now?

KA: This one year was more built to get the model right, so we got the model right. We got decent number of rides, people liking us, customers liking us, drivers are liking us, so now we have a template. We just have to replicate this template in different cities and stations. April is when we’re probably starting Hyderabad. This is how we’d be adding cities and metro stations month on month on this platform, it’s a little slower for the first year since we didn't have the template ready. Now we have the template ready so we can work much faster in adding cities and stations on our platform. By the end of this year, we’re also launching US. We started in 2019, I was trying to move back to India. But then we realised there’s a very big opportunity in US itself, and that’s where I stayed back and launched the US operations over here. The plan by the end of this year is that we would start at least two stations in the US.

Team evo India: Now you’ve launched in Noida, Bangalore, Delhi, each of these places have different people, different cultures, the demographics are different, even the geography is different. Does it cause different problems that you have to solve?

KA: The difference is mainly in the people. Comparing Bangalore to Noida, the people, the drivers are different, that’s why we had to take longer to standardise our process. We cannot run a Noida company and a Bangalore company right? Now we have that standard template which is at least 80 per cent common between any launch, 20 per cent is where we take care of… it's not just the people which are different, it's also the regulations. That 20 per cent regulation part is where it is variable right now and the processes have been standardised.

Team evo India: How do you reach out to customers in a hyper-local area?
KA: That’s the easiest part. All my current and prospective customers are coming to the same gate everyday, so all I have to do is be present at that gate and advertise at that gate. We do digital and media activities, but a lot of our advertising right now is either at the metro station or the last point which is probably a tech park or a college or a university.